{"id":"2ome-lm-2025","kind":"source","name":"2OMe-LM: predicting 2′-O-methylation sites in human RNA using a pre-trained RNA language model","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["rna-transcriptomes"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12342186/","version":"journal full text in PMC","retrieved_at":"2026-09-15T23:33:26Z","doi":"10.1093/bioinformatics/btaf417","publication_status":"peer_reviewed","year":2025,"artifact_sha256":"54fe4db6f35c03d0d4f3ef4da720eb26a832199372c56d0956609ff07af750ee","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12342186/fullTextXML","artifact_retrieved_at":"2026-09-16T10:33:38.332Z","legacy_paper":{"id":"2ome-lm-2025","title":"2OMe-LM: predicting 2′-O-methylation sites in human RNA using a pre-trained RNA language model","year":2025,"publication_status":"peer_reviewed","version":"journal full text in PMC","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12342186/","primary_domain":"rna-transcriptomes","retrieved_utc":"2026-09-15T23:33:26Z","doi":"10.1093/bioinformatics/btaf417","notes":"Numeric result checked against Table 1. in primary full-text XML; journal/source: Bioinformatics."},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"adar-gpt-editing-2026","kind":"source","name":"ADAR-GPT: A continually fine-tuned language model for predicting A-to-I RNA editing sites","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["rna-transcriptomes"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12798952/","version":"version of record","retrieved_at":"2026-09-15T23:37:05Z","doi":"10.1073/pnas.2529073123","publication_status":"peer_reviewed","year":2026,"artifact_sha256":"cc8c7eb928f246f1f347a8822f614cd3475381c35eef6d579032ce441580198e","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12798952/fullTextXML","artifact_retrieved_at":"2026-09-16T10:41:06Z","legacy_paper":{"id":"adar-gpt-editing-2026","title":"ADAR-GPT: A continually fine-tuned language model for predicting A-to-I RNA editing sites","year":2026,"publication_status":"peer_reviewed","version":"version of record","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12798952/","primary_domain":"rna-transcriptomes","retrieved_utc":"2026-09-15T23:37:05Z","notes":"Primary full text verified via Europe PMC fullTextXML; venue: Proceedings of the National Academy of Sciences of the United States of America; PMC ID: PMC12798952. RNA editing site benchmark on a restricted liver validation set.","doi":"10.1073/pnas.2529073123"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"akscore-2020","kind":"source","name":"AK-Score: Accurate Protein-Ligand Binding Affinity Prediction Using an Ensemble of 3D-Convolutional Neural Networks","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["molecular-interactions"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC7697539/","version":"version of record","retrieved_at":"2026-09-15T23:29:32Z","doi":"10.3390/ijms21228424","publication_status":"peer_reviewed","year":2020,"artifact_sha256":"40cfd28dcd587599768ec99a6590ec593486475ff01c7b1d1f229b44aa91bf8d","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC7697539/fullTextXML","artifact_retrieved_at":"2026-09-16T10:44:03.436853+00:00","legacy_paper":{"id":"akscore-2020","title":"AK-Score: Accurate Protein-Ligand Binding Affinity Prediction Using an Ensemble of 3D-Convolutional Neural Networks","year":2020,"publication_status":"peer_reviewed","version":"version of record","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC7697539/","primary_domain":"molecular-interactions","retrieved_utc":"2026-09-15T23:29:32Z","notes":"Primary full text verified using Europe PMC XML; venue: International Journal of Molecular Sciences; PMC ID: PMC7697539.","doi":"10.3390/ijms21228424"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"antibody-deamidation-plm-2024","kind":"source","name":"The Accurate Prediction of Antibody Deamidations by Combining High-Throughput Automated Peptide Mapping and Protein Language Model-Based Deep Learning","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["proteins-complexes"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11417914/","version":"journal full text in PMC","retrieved_at":"2026-09-15T23:33:26Z","doi":"10.3390/antib13030074","publication_status":"peer_reviewed","year":2024,"artifact_sha256":"aa049f6d78e29540ba902a0d3b7f53d49e9833e4dad9869f79ac27664e8c150b","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11417914/fullTextXML","artifact_retrieved_at":"2026-09-16T10:33:38.478Z","legacy_paper":{"id":"antibody-deamidation-plm-2024","title":"The Accurate Prediction of Antibody Deamidations by Combining High-Throughput Automated Peptide Mapping and Protein Language Model-Based Deep Learning","year":2024,"publication_status":"peer_reviewed","version":"journal full text in PMC","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11417914/","primary_domain":"proteins-complexes","retrieved_utc":"2026-09-15T23:33:26Z","doi":"10.3390/antib13030074","notes":"Numeric result checked against Table 1 in primary full-text XML; journal/source: Antibodies."},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"antibody-flexibility-2025","kind":"source","name":"Enhancing antibody-antigen interaction prediction with atomic flexibility","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["molecular-interactions"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12530544/","version":"version of record","retrieved_at":"2026-09-15T23:29:32Z","doi":"10.1371/journal.pcbi.1013576","publication_status":"peer_reviewed","year":2025,"artifact_sha256":"57e64694c69052ed0495570e12ebfb4bb6c0ad152219f23827cd4b1cb53450ef","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12530544/fullTextXML","artifact_retrieved_at":"2026-09-16T10:33:55.400Z","legacy_paper":{"id":"antibody-flexibility-2025","title":"Enhancing antibody-antigen interaction prediction with atomic flexibility","year":2025,"publication_status":"peer_reviewed","version":"version of record","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12530544/","primary_domain":"molecular-interactions","retrieved_utc":"2026-09-15T23:29:32Z","notes":"Primary full text verified using Europe PMC XML; venue: PLOS Computational Biology; PMC ID: PMC12530544.","doi":"10.1371/journal.pcbi.1013576"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"arsenal-regulatory-dna-2026","kind":"source","name":"Short-Context Regulatory DNA Language Models with Motif-Discovery Regularization","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["dna-genomes"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12889687/","version":"preprint version in PMC","retrieved_at":"2026-09-15T23:37:05Z","doi":"10.64898/2026.02.05.703637","publication_status":"preprint","year":2026,"artifact_sha256":"4a264956e47fc633aaff6573aac368dc691dd5de709b27c7421c078608ff542a","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12889687/fullTextXML","artifact_retrieved_at":"2026-09-16T10:41:06Z","legacy_paper":{"id":"arsenal-regulatory-dna-2026","title":"Short-Context Regulatory DNA Language Models with Motif-Discovery Regularization","year":2026,"publication_status":"preprint","version":"preprint version in PMC","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12889687/","primary_domain":"dna-genomes","retrieved_utc":"2026-09-15T23:37:05Z","notes":"Primary full text verified via Europe PMC fullTextXML; venue: bioRxiv; PMC ID: PMC12889687. Preprint; result is a supervised downstream model rather than a general-purpose DNA foundation model.","doi":"10.64898/2026.02.05.703637"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"b2-2ome-lm-2025","kind":"result","name":"2OMe-LM · AUC · human RNA 2OMe sites","description":"","status":"source_checked","facets":{"areas":["rna-transcriptomes"],"tasks":["human RNA 2-prime-O-methylation site prediction"]},"source_ids":["2ome-lm-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-b2-2ome-lm-2025"}],"attributes":{"printed_value":"0.919","numeric_value":"0.919","metric":"AUC","metric_direction":"unknown","unit":"fraction","uncertainty":null,"source_locator":"Table 1, 2OMe-LM row, AUC column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:38.332Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness.","evidence":"Table 1, 2OMe-LM row, AUC column; cell: 0.919","artifact_sha256":"54fe4db6f35c03d0d4f3ef4da720eb26a832199372c56d0956609ff07af750ee","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12342186/fullTextXML"},"legacy_id":"b2-2ome-lm-2025","legacy_row":{"id":"b2-2ome-lm-2025","paper_id":"2ome-lm-2025","domain_id":"rna-transcriptomes","task":"human RNA 2-prime-O-methylation site prediction","model":"2OMe-LM","model_version":"not stated in table","dataset":"human RNA 2OMe sites","dataset_version":"","split":"5-fold cross-validation","metric":"AUC","value":"0.919","unit":"fraction","uncertainty":"","protocol":"pretrained RNA language model predictor","source_locator":"Table 1, 2OMe-LM row, AUC column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12342186/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:33:26Z"},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"b2-antibody-deamidation-plm-2024","kind":"result","name":"ESM-2 650M embeddings + classifier · accuracy · antibody peptide-mapping training dataset","description":"","status":"source_checked","facets":{"areas":["proteins-complexes"],"tasks":["antibody deamidation-site prediction"]},"source_ids":["antibody-deamidation-plm-2024"],"links":[{"relation":"evaluation","target_id":"evaluation-b2-antibody-deamidation-plm-2024"}],"attributes":{"printed_value":"0.944","numeric_value":"0.944","metric":"accuracy","metric_direction":"unknown","unit":"fraction","uncertainty":"± 0.012","source_locator":"Table 1, Global embeddings only row, Accuracy column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:38.478Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness.","evidence":"Table 1, Global embeddings only row, Accuracy column; cell: 0.944 ± 0.012","artifact_sha256":"aa049f6d78e29540ba902a0d3b7f53d49e9833e4dad9869f79ac27664e8c150b","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11417914/fullTextXML"},"legacy_id":"b2-antibody-deamidation-plm-2024","legacy_row":{"id":"b2-antibody-deamidation-plm-2024","paper_id":"antibody-deamidation-plm-2024","domain_id":"proteins-complexes","task":"antibody deamidation-site prediction","model":"ESM-2 650M embeddings + classifier","model_version":"esm2_t33_650m_UR50D","dataset":"antibody peptide-mapping training dataset","dataset_version":"","split":"fivefold stratified CV","metric":"accuracy","value":"0.944","unit":"fraction","uncertainty":"± 0.012","protocol":"global contextual embeddings only","source_locator":"Table 1, Global embeddings only row, Accuracy column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11417914/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:33:26Z"},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract"}}}
{"id":"b2-barcodebert-2026","kind":"result","name":"BarcodeBERT (4–4-4) · accuracy · DNA barcodes of unseen species","description":"","status":"source_checked","facets":{"areas":["dna-genomes"],"tasks":["unseen-species genus classification"]},"source_ids":["barcodebert-2026"],"links":[{"relation":"evaluation","target_id":"evaluation-b2-barcodebert-2026"}],"attributes":{"printed_value":"78.5","numeric_value":"78.5","metric":"accuracy","metric_direction":"unknown","unit":"percent","uncertainty":null,"source_locator":"Table 1, BarcodeBERT (4–4-4) row, unseen-species genus-level 1-NN Acc (%) column","review":{"method":"independent_ai_table_review","reviewer":"Codex secondary table review","reviewed_at":"2026-09-16T10:38:57.558051+00:00","notes":"Resolved the two-level column header: Acc (%) falls under genus-level 1-NN probe of unseen species, not seen-species classification or BIN reconstruction. BarcodeBERT (4–4-4) has 78.5 in this cell.","evidence":"Original PMC XML table headers, row groups and caption inspected; printed value 78.5.","artifact_sha256":"493f9fe70b483780ba76d51ccf217d3ca83539c82b89917fd3ccebe2b6eb831d","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC13008329/fullTextXML"},"legacy_id":"b2-barcodebert-2026","legacy_row":{"id":"b2-barcodebert-2026","paper_id":"barcodebert-2026","domain_id":"dna-genomes","task":"unseen-species genus classification","model":"BarcodeBERT (4–4-4)","model_version":"4–4–4","dataset":"DNA barcodes of unseen species","dataset_version":"","split":"1-NN probe","metric":"accuracy","value":"78.5","unit":"percent","uncertainty":"","protocol":"genus-level nearest-neighbor probe on species unseen in training","source_locator":"Table 1, BarcodeBERT (4–4-4) row, unseen-species genus-level 1-NN Acc (%) column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC13008329/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:33:26Z"},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"b2-birna-bert-2025","kind":"result","name":"BiRNA-BERT · F1 · extremely long-sequence species classification","description":"","status":"source_checked","facets":{"areas":["rna-transcriptomes"],"tasks":["extremely long RNA species classification"]},"source_ids":["birna-bert-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-b2-birna-bert-2025"}],"attributes":{"printed_value":"0.804","numeric_value":"0.804","metric":"F1","metric_direction":"unknown","unit":"fraction","uncertainty":null,"source_locator":"Table 2, BiRNA-BERT row, F1 Score column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:38.292Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness.","evidence":"Table 2, BiRNA-BERT row, F1 Score column; cell: 0.804","artifact_sha256":"bf7dbc52b6515301c77010c513f13e676c38395ddc82c20310171f518690c152","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12635123/fullTextXML"},"legacy_id":"b2-birna-bert-2025","legacy_row":{"id":"b2-birna-bert-2025","paper_id":"birna-bert-2025","domain_id":"rna-transcriptomes","task":"extremely long RNA species classification","model":"BiRNA-BERT","model_version":"not stated in table","dataset":"extremely long-sequence species classification","dataset_version":"","split":"paper evaluation","metric":"F1","value":"0.804","unit":"fraction","uncertainty":"","protocol":"adaptive tokenization on full-length long RNA sequences","source_locator":"Table 2, BiRNA-BERT row, F1 Score column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12635123/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:33:26Z"},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"b2-cathe2-2025","kind":"result","name":"CATHe2 + ProstT5 · F1 · CATH superfamily benchmark","description":"","status":"source_checked","facets":{"areas":["proteins-complexes"],"tasks":["CATH superfamily annotation"]},"source_ids":["cathe2-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-b2-cathe2-2025"}],"attributes":{"printed_value":"82.3","numeric_value":"82.3","metric":"F1","metric_direction":"unknown","unit":"percent","uncertainty":"± 1.3 percentage points","source_locator":"Table 3, ProstT5 full row, F1 score column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:38.366Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness.","evidence":"Table 3, ProstT5 full row, F1 score column; cell: 82.3% ± 1.3%","artifact_sha256":"713dbfb6ec1cc1aa85c0543eb93aafa0b45b8873df28053b765dd0a1b6d9b563","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12631783/fullTextXML"},"legacy_id":"b2-cathe2-2025","legacy_row":{"id":"b2-cathe2-2025","paper_id":"cathe2-2025","domain_id":"proteins-complexes","task":"CATH superfamily annotation","model":"CATHe2 + ProstT5","model_version":"full ProstT5","dataset":"CATH superfamily benchmark","dataset_version":"","split":"paper evaluation","metric":"F1","value":"82.3","unit":"percent","uncertainty":"± 1.3 percentage points","protocol":"amino-acid and structural alphabet embedding classifier","source_locator":"Table 3, ProstT5 full row, F1 score column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12631783/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:33:26Z"},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract"}}}
{"id":"b2-clathrin-plm-2025","kind":"result","name":"ESM-2 embedding + paper classifier · accuracy · CLA-IND0.6","description":"","status":"source_checked","facets":{"areas":["proteins-complexes"],"tasks":["clathrin protein classification"]},"source_ids":["clathrin-plm-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-b2-clathrin-plm-2025"}],"attributes":{"printed_value":"0.916","numeric_value":"0.916","metric":"accuracy","metric_direction":"unknown","unit":"fraction","uncertainty":null,"source_locator":"Table 2, Independent test / ESM-2 row, ACC column","review":{"method":"independent_ai_table_review","reviewer":"Codex secondary table review","reviewed_at":"2026-09-16T10:38:57.558194+00:00","notes":"Resolved the blank evaluation-strategy cells by their independent-test row group. ESM-2 ACC is 0.916 there; the cross-validation ESM-2 ACC is instead 0.873. This is the paper classifier using embeddings, not a standalone checkpoint.","evidence":"Original PMC XML table headers, row groups and caption inspected; printed value 0.916.","artifact_sha256":"2edc86b25707c1b737d26117093ce8d856e79cc5d0b335f27c1c341f887f1c7e","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12238356/fullTextXML"},"legacy_id":"b2-clathrin-plm-2025","legacy_row":{"id":"b2-clathrin-plm-2025","paper_id":"clathrin-plm-2025","domain_id":"proteins-complexes","task":"clathrin protein classification","model":"ESM-2 embedding + paper classifier","model_version":"not stated in table","dataset":"CLA-IND0.6","dataset_version":"","split":"independent test","metric":"accuracy","value":"0.916","unit":"fraction","uncertainty":"","protocol":"single-feature ESM-2 embedding comparison","source_locator":"Table 2, Independent test / ESM-2 row, ACC column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12238356/","evaluation_origin":"independent_paper","reviewed_utc":"2026-09-15T23:33:26Z"},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"b2-cobra-rna-binding-2026","kind":"result","name":"ERNIE-RNA + CoBRA · MCC · CoBRA compound-binding test set","description":"","status":"source_checked","facets":{"areas":["rna-transcriptomes"],"tasks":["RNA compound-binding site prediction"]},"source_ids":["cobra-rna-binding-2026"],"links":[{"relation":"evaluation","target_id":"evaluation-b2-cobra-rna-binding-2026"}],"attributes":{"printed_value":"0.657","numeric_value":"0.657","metric":"MCC","metric_direction":"unknown","unit":"unitless","uncertainty":null,"source_locator":"Table 2, ERNIE-RNA / TCL focal row, MCC column","review":{"method":"independent_ai_table_review","reviewer":"Codex secondary table review","reviewed_at":"2026-09-16T10:38:57.558197+00:00","notes":"Matched ERNIE-RNA jointly with TCL focal loss, then the MCC column. Table 2 explicitly reports test-set models. The cell is 0.657, distinct from AUROC 0.868.","evidence":"Original PMC XML table headers, row groups and caption inspected; printed value 0.657.","artifact_sha256":"8c6a6f00f5fa5f62acf301a66e9e6fa9ef11c7a05ad9b7447d2ade2ce8eba793","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12790621/fullTextXML"},"legacy_id":"b2-cobra-rna-binding-2026","legacy_row":{"id":"b2-cobra-rna-binding-2026","paper_id":"cobra-rna-binding-2026","domain_id":"rna-transcriptomes","task":"RNA compound-binding site prediction","model":"ERNIE-RNA + CoBRA","model_version":"not stated in table","dataset":"CoBRA compound-binding test set","dataset_version":"","split":"test set","metric":"MCC","value":"0.657","unit":"unitless","uncertainty":"","protocol":"ERNIE-RNA embedding with TCL focal loss","source_locator":"Table 2, ERNIE-RNA / TCL focal row, MCC column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12790621/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:33:26Z"},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"b2-codonbert-vaccines-2024","kind":"result","name":"CodonBERT · Spearman rho · flu-vaccine sequences","description":"","status":"source_checked","facets":{"areas":["rna-transcriptomes"],"tasks":["flu-vaccine mRNA property prediction"]},"source_ids":["codonbert-vaccines-2024"],"links":[{"relation":"evaluation","target_id":"evaluation-b2-codonbert-vaccines-2024"}],"attributes":{"printed_value":"0.81","numeric_value":"0.81","metric":"Spearman rho","metric_direction":"unknown","unit":"unitless","uncertainty":null,"source_locator":"Table 2, CodonBERT row, Flu vaccines Spearman correlation column","review":{"method":"independent_ai_table_review","reviewer":"Codex secondary table review","reviewed_at":"2026-09-16T10:38:57.558201+00:00","notes":"Matched the CodonBERT row and Flu vaccines column (0.81). The table footnote identifies regression columns as Spearman rank correlation and singles out E. coli as classification; this is not a flu-vaccine accuracy score.","evidence":"Original PMC XML table headers, row groups and caption inspected; printed value 0.81.","artifact_sha256":"2968073753e6d44feff9c08b131edf23145e95b171434539dddf77bb92847033","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11368176/fullTextXML"},"legacy_id":"b2-codonbert-vaccines-2024","legacy_row":{"id":"b2-codonbert-vaccines-2024","paper_id":"codonbert-vaccines-2024","domain_id":"rna-transcriptomes","task":"flu-vaccine mRNA property prediction","model":"CodonBERT","model_version":"not stated in table","dataset":"flu-vaccine sequences","dataset_version":"","split":"paper evaluation","metric":"Spearman rho","value":"0.81","unit":"unitless","uncertainty":"","protocol":"codon-based model fine-tuned for downstream regression","source_locator":"Table 2, CodonBERT row, Flu vaccines Spearman correlation column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11368176/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:33:26Z"},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"b2-dart-eval-regulatory-2024","kind":"result","name":"DNABERT-2 · accuracy · DART-Eval cCREs versus matched shuffled controls","description":"","status":"source_checked","facets":{"areas":["dna-genomes"],"tasks":["regulatory element identification"]},"source_ids":["dart-eval-regulatory-2024"],"links":[{"relation":"evaluation","target_id":"evaluation-b2-dart-eval-regulatory-2024"}],"attributes":{"printed_value":"0.876","numeric_value":"0.876","metric":"accuracy","metric_direction":"unknown","unit":"fraction","uncertainty":null,"source_locator":"Table 3 (PDF page 5), DNABERT-2 row, Zero-Shot Accuracy column","review":{"method":"independent_ai_table_review","reviewer":"Codex secondary table review","reviewed_at":"2026-09-16T10:38:57.558203+00:00","notes":"Inspected the pinned NeurIPS primary PDF table and explanatory text. The DNABERT-2 row reports 0.876 under Zero-Shot Accuracy. The caption defines this as pairwise prioritization of positives over matched controls, distinct from supervised absolute accuracy.","evidence":"DNABERT-2; zero-shot accuracy 0.876; probed absolute/paired 0.847/0.943; fine-tuned absolute/paired 0.913/0.973.","artifact_sha256":"e5aee5b1f7cc6fd961b1d2a131d02cf243b79e091d5e418fbabee7fde9b39b22","retrieval_url":"https://proceedings.neurips.cc/paper_files/paper/2024/file/71998bfc3217ffe1cca1ee084dfadadd-Paper-Datasets_and_Benchmarks_Track.pdf"},"legacy_id":"b2-dart-eval-regulatory-2024","legacy_row":{"id":"b2-dart-eval-regulatory-2024","paper_id":"dart-eval-regulatory-2024","domain_id":"dna-genomes","task":"regulatory element identification","model":"DNABERT-2","model_version":"not stated in table","dataset":"DART-Eval cCREs versus matched shuffled controls","dataset_version":"","split":"paper evaluation","metric":"accuracy","value":"0.876","unit":"fraction","uncertainty":"","protocol":"zero-shot likelihood ranking: higher likelihood for cCRE than matched control","source_locator":"Table 3, DNABERT-2 row, Zero-Shot Accuracy column","source_url":"https://proceedings.neurips.cc/paper_files/paper/2024/file/71998bfc3217ffe1cca1ee084dfadadd-Paper-Datasets_and_Benchmarks_Track.pdf","evaluation_origin":"independent_paper","reviewed_utc":"2026-09-15T23:33:26Z"},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"b2-dnabert2-enhancer-2025","kind":"result","name":"DNABERT2-Enhancer · AUC · Liu training dataset","description":"","status":"source_checked","facets":{"areas":["dna-genomes"],"tasks":["enhancer recognition"]},"source_ids":["dnabert2-enhancer-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-b2-dnabert2-enhancer-2025"}],"attributes":{"printed_value":"0.965","numeric_value":"0.965","metric":"AUC","metric_direction":"unknown","unit":"fraction","uncertainty":null,"source_locator":"Table 4, first-layer DNABERT2-Enhancer row, AUC column","review":{"method":"independent_ai_table_review","reviewer":"Codex secondary table review","reviewed_at":"2026-09-16T10:38:57.558204+00:00","notes":"Resolved the first-layer row group. DNABERT2-Enhancer AUC is 0.965, whereas second-layer AUC is 0.933. The caption explicitly describes 5-fold cross-validation on Liu training data, not an independent held-out test.","evidence":"Original PMC XML table headers, row groups and caption inspected; printed value 0.965.","artifact_sha256":"d052b80efe7bfc1380994ad28503a5575f04ef940f74d5c9c137cb4ba6827863","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11981215/fullTextXML"},"legacy_id":"b2-dnabert2-enhancer-2025","legacy_row":{"id":"b2-dnabert2-enhancer-2025","paper_id":"dnabert2-enhancer-2025","domain_id":"dna-genomes","task":"enhancer recognition","model":"DNABERT2-Enhancer","model_version":"not stated in table","dataset":"Liu training dataset","dataset_version":"","split":"5-fold cross-validation","metric":"AUC","value":"0.965","unit":"fraction","uncertainty":"","protocol":"first-layer enhancer versus non-enhancer classifier","source_locator":"Table 4, first-layer DNABERT2-Enhancer row, AUC column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11981215/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:33:26Z"},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"b2-eden-genomic-classification-2026","kind":"result","name":"DNABERT-2 · MCC · GUE H-CPD","description":"","status":"source_checked","facets":{"areas":["dna-genomes"],"tasks":["human core-promoter classification"]},"source_ids":["eden-genomic-classification-2026"],"links":[{"relation":"evaluation","target_id":"evaluation-b2-eden-genomic-classification-2026"}],"attributes":{"printed_value":"70.52","numeric_value":"70.52","metric":"MCC","metric_direction":"unknown","unit":"percent","uncertainty":null,"source_locator":"Table 5, DNABERT-2 row, H-CPD (MCC) column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:37.531Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness.","evidence":"Table 5, DNABERT-2 row, H-CPD (MCC) column; cell: 70.52","artifact_sha256":"38a6e26b3caffe8e021a2b0b672218e783aca9ee42046765e323946813015e65","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12879454/fullTextXML"},"legacy_id":"b2-eden-genomic-classification-2026","legacy_row":{"id":"b2-eden-genomic-classification-2026","paper_id":"eden-genomic-classification-2026","domain_id":"dna-genomes","task":"human core-promoter classification","model":"DNABERT-2","model_version":"not stated in table","dataset":"GUE H-CPD","dataset_version":"","split":"paper evaluation","metric":"MCC","value":"70.52","unit":"percent","uncertainty":"","protocol":"DNABERT-2 comparator in consolidated H-CPD table; rerun provenance not explicit","source_locator":"Table 5, DNABERT-2 row, H-CPD (MCC) column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12879454/","evaluation_origin":"paper_compilation","reviewed_utc":"2026-09-15T23:33:26Z"},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"b2-ernie-rna-2025","kind":"result","name":"ERNIE-RNA · binary F1 · bpRNA-new","description":"","status":"source_checked","facets":{"areas":["rna-transcriptomes"],"tasks":["RNA secondary-structure prediction"]},"source_ids":["ernie-rna-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-b2-ernie-rna-2025"}],"attributes":{"printed_value":"0.575","numeric_value":"0.575","metric":"binary F1","metric_direction":"unknown","unit":"fraction","uncertainty":null,"source_locator":"Table 2, ERNIE-RNA zero-shot row, bpRNA-new F1-Score (binary) column","review":{"method":"independent_ai_table_review","reviewer":"Codex secondary table review","reviewed_at":"2026-09-16T10:38:57.558206+00:00","notes":"Resolved bpRNA-new as the first three-column dataset group and F1-Score (binary) as its third metric. ERNIE-RNA zero shot is 86M and reports 0.575; RNA3DB-2D F1 is instead 0.542.","evidence":"Original PMC XML table headers, row groups and caption inspected; printed value 0.575.","artifact_sha256":"0bd1d4b3cbf5d59d452cec4864614947861efcee050ba07e7de395cd90630047","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12627772/fullTextXML"},"legacy_id":"b2-ernie-rna-2025","legacy_row":{"id":"b2-ernie-rna-2025","paper_id":"ernie-rna-2025","domain_id":"rna-transcriptomes","task":"RNA secondary-structure prediction","model":"ERNIE-RNA","model_version":"86M","dataset":"bpRNA-new","dataset_version":"","split":"cross-family test","metric":"binary F1","value":"0.575","unit":"fraction","uncertainty":"","protocol":"zero-shot attention-derived base-pair prediction","source_locator":"Table 2, ERNIE-RNA zero-shot row, bpRNA-new F1-Score (binary) column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12627772/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:33:26Z"},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"b2-esm2-ofs-fitness-2025","kind":"result","name":"ESM2 OFS pseudo-perplexity · Spearman rho · ProteinGym substitutions","description":"","status":"source_checked","facets":{"areas":["proteins-complexes"],"tasks":["protein variant fitness prediction"]},"source_ids":["esm2-ofs-fitness-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-b2-esm2-ofs-fitness-2025"}],"attributes":{"printed_value":"0.403","numeric_value":"0.403","metric":"Spearman rho","metric_direction":"unknown","unit":"unitless","uncertainty":null,"source_locator":"Published PDF page 6 (033014-6), Table I, ESM2: OFS PP row, Aggregate mean column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source review, not human review","reviewed_at":"2026-09-16T10:45:41.099916+00:00","notes":"Publisher PDF retrieved through official APS harvest endpoint after direct download returned403. Table I is ProteinGym substitutions, not indels TableII. Last column aggregate mean0.403; separate function categories precede it. This verifies reported score, not experimental reproduction. Comparator rows in this table are sourced from ProteinGym; OFS PP is authors own method.","evidence":"Headers: Activity(43), Binding(14), Expression(17), Organismal fitness(77), Stability(66), Aggregate mean. ESM2:OFS PP row:0.393,0.279,0.397,0.331,0.507,0.403. Verified publisher PDF layout extraction against web-rendered primary PDF table.","artifact_sha256":"085ef646f11b8e5335c4b3d86b15fb6c7bf5edf4a80a8b622753ac69d9991a67","retrieval_url":"https://harvest.aps.org/v2/journals/articles/10.1103/zhx7-hcmm/fulltext"},"legacy_id":"b2-esm2-ofs-fitness-2025","legacy_row":{"id":"b2-esm2-ofs-fitness-2025","paper_id":"esm2-ofs-fitness-2025","domain_id":"proteins-complexes","task":"protein variant fitness prediction","model":"ESM2 OFS pseudo-perplexity","model_version":"not stated in table","dataset":"ProteinGym substitutions","dataset_version":"","split":"aggregate across assays","metric":"Spearman rho","value":"0.403","unit":"unitless","uncertainty":"","protocol":"authors’ zero-shot ESM2 OFS pseudo-perplexity evaluation; aggregate mean across ProteinGym substitution assays","source_locator":"Table I, ESM2: OFS PP row, Aggregate Mean Spearman correlation column","source_url":"https://journals.aps.org/prxlife/pdf/10.1103/zhx7-hcmm","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:33:26Z"},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"b2-fusion-breakpoint-foundation-models-2026","kind":"result","name":"Nucleotide Transformer + NN (middle) · ROC AUC · gene fusion breakpoint DNA sequences","description":"","status":"source_checked","facets":{"areas":["dna-genomes"],"tasks":["gene fusion breakpoint classification"]},"source_ids":["fusion-breakpoint-foundation-models-2026"],"links":[{"relation":"evaluation","target_id":"evaluation-b2-fusion-breakpoint-foundation-models-2026"}],"attributes":{"printed_value":"0.994","numeric_value":"0.994","metric":"ROC AUC","metric_direction":"unknown","unit":"fraction","uncertainty":null,"source_locator":"Table 2, NT / NN (middle) row, ROC AUC column","review":{"method":"independent_ai_table_review","reviewer":"Codex secondary table review","reviewed_at":"2026-09-16T10:38:57.558209+00:00","notes":"Matched NT jointly with NN (middle) and ROC AUC 0.994 in the full-test-set table. NT with SVM reports 0.995 and is a separate pipeline.","evidence":"Original PMC XML table headers, row groups and caption inspected; printed value 0.994.","artifact_sha256":"0f4d9de77f1e39cfd2164a20653d86370767da684dc22d17e09f589761abeb5f","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC13182013/fullTextXML"},"legacy_id":"b2-fusion-breakpoint-foundation-models-2026","legacy_row":{"id":"b2-fusion-breakpoint-foundation-models-2026","paper_id":"fusion-breakpoint-foundation-models-2026","domain_id":"dna-genomes","task":"gene fusion breakpoint classification","model":"Nucleotide Transformer + NN (middle)","model_version":"not stated in table","dataset":"gene fusion breakpoint DNA sequences","dataset_version":"","split":"full test set","metric":"ROC AUC","value":"0.994","unit":"fraction","uncertainty":"","protocol":"middle embedding with neural-network classifier","source_locator":"Table 2, NT / NN (middle) row, ROC AUC column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC13182013/","evaluation_origin":"independent_paper","reviewed_utc":"2026-09-15T23:33:26Z"},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"b2-genomic-tokenizer-selection-2025","kind":"result","name":"Caduceus (character tokens) · MCC · genomic benchmark categories","description":"","status":"source_checked","facets":{"areas":["dna-genomes"],"tasks":["regulatory sequence classification"]},"source_ids":["genomic-tokenizer-selection-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-b2-genomic-tokenizer-selection-2025"}],"attributes":{"printed_value":"0.778","numeric_value":"0.778","metric":"MCC","metric_direction":"unknown","unit":"unitless","uncertainty":null,"source_locator":"Table 2, Regulatory row, Caduceus (char) MCC column","review":{"method":"independent_ai_table_review","reviewer":"Codex secondary table review","reviewed_at":"2026-09-16T10:38:57.558210+00:00","notes":"Matched Regulatory row with Caduceus (char) column, 0.778. Caption establishes these as MCC summaries by category; model-size row identifies 3.9M parameters. This is an aggregated category result, not a single unspecified split.","evidence":"Original PMC XML table headers, row groups and caption inspected; printed value 0.778.","artifact_sha256":"0a01c36fdd63f3f6db509777e61c3f87e8a298c810f8aef7974915aaa0655342","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12453675/fullTextXML"},"legacy_id":"b2-genomic-tokenizer-selection-2025","legacy_row":{"id":"b2-genomic-tokenizer-selection-2025","paper_id":"genomic-tokenizer-selection-2025","domain_id":"dna-genomes","task":"regulatory sequence classification","model":"Caduceus (character tokens)","model_version":"3.9M parameter variant","dataset":"genomic benchmark categories","dataset_version":"","split":"paper benchmark summary","metric":"MCC","value":"0.778","unit":"unitless","uncertainty":"","protocol":"task-category MCC across benchmark datasets","source_locator":"Table 2, Regulatory row, Caduceus (char) MCC column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12453675/","evaluation_origin":"independent_paper","reviewed_utc":"2026-09-15T23:33:26Z"},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"b2-gsmformer-ppi-2026","kind":"result","name":"GSMFormer-PPI + ProstT5 · AUROC · paper PPI test set","description":"","status":"source_checked","facets":{"areas":["proteins-complexes"],"tasks":["protein-protein interaction prediction"]},"source_ids":["gsmformer-ppi-2026"],"links":[{"relation":"evaluation","target_id":"evaluation-b2-gsmformer-ppi-2026"}],"attributes":{"printed_value":"0.988","numeric_value":"0.988","metric":"AUROC","metric_direction":"unknown","unit":"fraction","uncertainty":null,"source_locator":"Table 6, ProstT5 embedding row, AUROC column","review":{"method":"independent_ai_table_review","reviewer":"Codex secondary table review","reviewed_at":"2026-09-16T10:38:57.558212+00:00","notes":"Matched ProstT5 embedding row and AUROC column, 0.988. Caption explicitly describes GSMFormer-PPI using embeddings as node features, not standalone ProstT5 prediction.","evidence":"Original PMC XML table headers, row groups and caption inspected; printed value 0.988.","artifact_sha256":"9b364b5d73d16f2787f93f78f17dbe98b954ab9c2c64c1df960eec2e615eb3b4","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12873117/fullTextXML"},"legacy_id":"b2-gsmformer-ppi-2026","legacy_row":{"id":"b2-gsmformer-ppi-2026","paper_id":"gsmformer-ppi-2026","domain_id":"proteins-complexes","task":"protein-protein interaction prediction","model":"GSMFormer-PPI + ProstT5","model_version":"not stated in table","dataset":"paper PPI test set","dataset_version":"","split":"test set","metric":"AUROC","value":"0.988","unit":"fraction","uncertainty":"","protocol":"ProstT5 embeddings as graph node features","source_locator":"Table 6, ProstT5 embedding row, AUROC column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12873117/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:33:26Z"},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"b2-megsite-2025","kind":"result","name":"MegSite + ESM3 · AUC · DNA-129_Test","description":"","status":"source_checked","facets":{"areas":["proteins-complexes"],"tasks":["DNA-binding residue prediction"]},"source_ids":["megsite-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-b2-megsite-2025"}],"attributes":{"printed_value":"0.948","numeric_value":"0.948","metric":"AUC","metric_direction":"unknown","unit":"fraction","uncertainty":null,"source_locator":"Table 2, DNA-129_Test / ESM3 row, AUC column","review":{"method":"independent_ai_table_review","reviewer":"Codex secondary table review","reviewed_at":"2026-09-16T10:38:57.558213+00:00","notes":"Resolved DNA-129_Test row group and ESM3 row. AUC is 0.948; the next numeric cell 0.582 is AP. Caption states an embedding comparison within MegSite.","evidence":"Original PMC XML table headers, row groups and caption inspected; printed value 0.948.","artifact_sha256":"10d13122331813243d83b84fe6f9294eac7e7c03cde082ebed276191ac41089c","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12496013/fullTextXML"},"legacy_id":"b2-megsite-2025","legacy_row":{"id":"b2-megsite-2025","paper_id":"megsite-2025","domain_id":"proteins-complexes","task":"DNA-binding residue prediction","model":"MegSite + ESM3","model_version":"not stated in table","dataset":"DNA-129_Test","dataset_version":"","split":"independent test","metric":"AUC","value":"0.948","unit":"fraction","uncertainty":"","protocol":"ESM3 multimodal embedding ablation in MegSite","source_locator":"Table 2, DNA-129_Test / ESM3 row, AUC column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12496013/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:33:26Z"},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"b2-mrna-lm-2025","kind":"result","name":"mRNA-LM · Spearman rho · mRNA half-life","description":"","status":"source_checked","facets":{"areas":["rna-transcriptomes"],"tasks":["mRNA half-life prediction"]},"source_ids":["mrna-lm-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-b2-mrna-lm-2025"}],"attributes":{"printed_value":"0.696","numeric_value":"0.696","metric":"Spearman rho","metric_direction":"unknown","unit":"unitless","uncertainty":null,"source_locator":"Table 1, mRNA-LM row, mRNA half-life Spearman column","review":{"method":"independent_ai_table_review","reviewer":"Codex secondary table review","reviewed_at":"2026-09-16T10:38:57.558214+00:00","notes":"Resolved mRNA half-life column under the Spearman header spanning three tasks. mRNA-LM gives 0.696. Caption identifies average test performance across cross-validation splits; protein-expression AUROC is a different column.","evidence":"Original PMC XML table headers, row groups and caption inspected; printed value 0.696.","artifact_sha256":"3a23de3c672ec162d13561c483f180a73b550d717256deffdc9099accec205fd","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11962594/fullTextXML"},"legacy_id":"b2-mrna-lm-2025","legacy_row":{"id":"b2-mrna-lm-2025","paper_id":"mrna-lm-2025","domain_id":"rna-transcriptomes","task":"mRNA half-life prediction","model":"mRNA-LM","model_version":"not stated in table","dataset":"mRNA half-life","dataset_version":"","split":"test set across CV splits","metric":"Spearman rho","value":"0.696","unit":"unitless","uncertainty":"","protocol":"average test performance across cross-validation splits","source_locator":"Table 1, mRNA-LM row, mRNA half-life Spearman column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11962594/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:33:26Z"},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"b2-mrnabert-2025","kind":"result","name":"mRNABERT · R-squared · human ultra-long mRNAs","description":"","status":"source_checked","facets":{"areas":["rna-transcriptomes"],"tasks":["translation-efficiency prediction"]},"source_ids":["mrnabert-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-b2-mrnabert-2025"}],"attributes":{"printed_value":"0.669","numeric_value":"0.669","metric":"R-squared","metric_direction":"unknown","unit":"unitless","uncertainty":null,"source_locator":"Table 2, mRNABERT (3066) row, Human R-squared column","review":{"method":"independent_ai_table_review","reviewer":"Codex secondary table review","reviewed_at":"2026-09-16T10:38:57.558216+00:00","notes":"Resolved Human group and its R-squared subcolumn. mRNABERT (3066) reports 0.669; Human Spearman is 0.814 and Mouse R-squared is 0.649. Caption specifies ultra-long mRNA translation-efficiency prediction.","evidence":"Original PMC XML table headers, row groups and caption inspected; printed value 0.669.","artifact_sha256":"ff08ba895b7080446c08a930548b48a0041ae990c222ebb07e6ba7dcaf48ad44","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12644827/fullTextXML"},"legacy_id":"b2-mrnabert-2025","legacy_row":{"id":"b2-mrnabert-2025","paper_id":"mrnabert-2025","domain_id":"rna-transcriptomes","task":"translation-efficiency prediction","model":"mRNABERT","model_version":"3066-nt input","dataset":"human ultra-long mRNAs","dataset_version":"","split":"paper evaluation","metric":"R-squared","value":"0.669","unit":"unitless","uncertainty":"","protocol":"human translation-efficiency regression at 3066-nt input","source_locator":"Table 2, mRNABERT (3066) row, Human R-squared column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12644827/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:33:26Z"},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"b2-mulan-2025","kind":"result","name":"MULAN-ESM2 S · AUC · HumanPPI","description":"","status":"source_checked","facets":{"areas":["proteins-complexes"],"tasks":["human protein-protein interaction prediction"]},"source_ids":["mulan-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-b2-mulan-2025"}],"attributes":{"printed_value":"0.717","numeric_value":"0.717","metric":"AUC","metric_direction":"unknown","unit":"fraction","uncertainty":null,"source_locator":"Table 2, MULAN-ESM2 S row, HumanPPI AUC column","review":{"method":"independent_ai_table_review","reviewer":"Codex secondary table review","reviewed_at":"2026-09-16T10:38:57.558217+00:00","notes":"Resolved the multirow header: HumanPPI uses AUC. MULAN-ESM2 S has 0.717; this is the small-model group, distinct from M and L variants.","evidence":"Original PMC XML table headers, row groups and caption inspected; printed value 0.717.","artifact_sha256":"771a9a26ebda6f49ea266540e8dd6e6de0cbaef724de818ca6124a5f9c50d350","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12452268/fullTextXML"},"legacy_id":"b2-mulan-2025","legacy_row":{"id":"b2-mulan-2025","paper_id":"mulan-2025","domain_id":"proteins-complexes","task":"human protein-protein interaction prediction","model":"MULAN-ESM2 S","model_version":"small ESM2 backbone","dataset":"HumanPPI","dataset_version":"","split":"paper evaluation","metric":"AUC","value":"0.717","unit":"fraction","uncertainty":"","protocol":"MULAN sequence-structure model based on ESM2 8M","source_locator":"Table 2, MULAN-ESM2 S row, HumanPPI AUC column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12452268/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:33:26Z"},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"b2-phylogpn-2025","kind":"result","name":"PhyloGPN · AUROC · ClinVar 3-prime UTR variants","description":"","status":"source_checked","facets":{"areas":["dna-genomes"],"tasks":["ClinVar 3-prime UTR variant classification"]},"source_ids":["phylogpn-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-b2-phylogpn-2025"}],"attributes":{"printed_value":"0.94","numeric_value":"0.94","metric":"AUROC","metric_direction":"unknown","unit":"fraction","uncertainty":null,"source_locator":"Table 1, 3-prime UTR row, PhyloGPN AUROC column","review":{"method":"independent_ai_table_review","reviewer":"Codex secondary table review","reviewed_at":"2026-09-16T10:38:57.558218+00:00","notes":"Matched 3-prime UTR row and PhyloGPN column (0.94). Caption specifies log-likelihood-ratio predictions of ClinVar classes and explicitly defines each cell as AUROC.","evidence":"Original PMC XML table headers, row groups and caption inspected; printed value 0.94.","artifact_sha256":"807f3a26cbfa9b5ce238d92164bd523302c67d1c5794b08273c51cca1acd4224","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11908359/fullTextXML"},"legacy_id":"b2-phylogpn-2025","legacy_row":{"id":"b2-phylogpn-2025","paper_id":"phylogpn-2025","domain_id":"dna-genomes","task":"ClinVar 3-prime UTR variant classification","model":"PhyloGPN","model_version":"not stated in table","dataset":"ClinVar 3-prime UTR variants","dataset_version":"","split":"paper evaluation","metric":"AUROC","value":"0.94","unit":"fraction","uncertainty":"","protocol":"log-likelihood-ratio scoring","source_locator":"Table 1, 3-prime UTR row, PhyloGPN AUROC column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11908359/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:33:26Z"},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"b2-polya-glm-2025","kind":"result","name":"HyenaDNA · AUC · poly(A) Gene-Gene","description":"","status":"source_checked","facets":{"areas":["dna-genomes"],"tasks":["polyadenylation site detection"]},"source_ids":["polya-glm-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-b2-polya-glm-2025"}],"attributes":{"printed_value":"0.7510","numeric_value":"0.7510","metric":"AUC","metric_direction":"unknown","unit":"fraction","uncertainty":null,"source_locator":"Table 1, Few-shot HyenaDNA row, G-G AUC column","review":{"method":"independent_ai_table_review","reviewer":"Codex secondary table review","reviewed_at":"2026-09-16T10:38:57.558220+00:00","notes":"Resolved Few-shot group, HyenaDNA row, and G-G subcolumn under AUC (0.7510). IG-G AUC is 0.7541. Caption states averages over five-fold cross-validation and distinguishes negative sampling regions.","evidence":"Original PMC XML table headers, row groups and caption inspected; printed value 0.7510.","artifact_sha256":"e9ebd53d88837ad8d457881ffee918d2734dcae87d3c5cd03135947b6cf5dbde","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12799945/fullTextXML"},"legacy_id":"b2-polya-glm-2025","legacy_row":{"id":"b2-polya-glm-2025","paper_id":"polya-glm-2025","domain_id":"dna-genomes","task":"polyadenylation site detection","model":"HyenaDNA","model_version":"not stated in table","dataset":"poly(A) Gene-Gene","dataset_version":"","split":"5-fold cross-validation","metric":"AUC","value":"0.7510","unit":"fraction","uncertainty":"","protocol":"few-shot Gene-Gene negative-set comparison","source_locator":"Table 1, Few-shot HyenaDNA row, G-G AUC column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12799945/","evaluation_origin":"independent_paper","reviewed_utc":"2026-09-15T23:33:26Z"},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"b2-rlsite-rna-binding-2025","kind":"result","name":"RLsite · AUC · T18","description":"","status":"source_checked","facets":{"areas":["rna-transcriptomes"],"tasks":["RNA-small-molecule binding-site prediction"]},"source_ids":["rlsite-rna-binding-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-b2-rlsite-rna-binding-2025"}],"attributes":{"printed_value":"0.828","numeric_value":"0.828","metric":"AUC","metric_direction":"unknown","unit":"fraction","uncertainty":null,"source_locator":"Table 1, RLsite row, T18 AUC column","review":{"method":"independent_ai_table_review","reviewer":"Codex secondary table review","reviewed_at":"2026-09-16T10:38:57.558222+00:00","notes":"Matched RLsite and AUC (0.828). Caption explicitly identifies dataset T18; MCC 0.474 is a different metric.","evidence":"Original PMC XML table headers, row groups and caption inspected; printed value 0.828.","artifact_sha256":"a50f344e253162ae43f51d7120cfb35a1d0f6114fd8176d760aceb6d05fd95bd","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12417085/fullTextXML"},"legacy_id":"b2-rlsite-rna-binding-2025","legacy_row":{"id":"b2-rlsite-rna-binding-2025","paper_id":"rlsite-rna-binding-2025","domain_id":"rna-transcriptomes","task":"RNA-small-molecule binding-site prediction","model":"RLsite","model_version":"not stated in table","dataset":"T18","dataset_version":"","split":"paper evaluation","metric":"AUC","value":"0.828","unit":"fraction","uncertainty":"","protocol":"RNA language-model plus graph-attention classifier","source_locator":"Table 1, RLsite row, T18 AUC column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12417085/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:33:26Z"},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"b2-rnaret-2026","kind":"result","name":"RNAret · F1 · MirTarRAW","description":"","status":"source_checked","facets":{"areas":["rna-transcriptomes"],"tasks":["miRNA-mRNA interaction prediction"]},"source_ids":["rnaret-2026"],"links":[{"relation":"evaluation","target_id":"evaluation-b2-rnaret-2026"}],"attributes":{"printed_value":"0.9622","numeric_value":"0.9622","metric":"F1","metric_direction":"unknown","unit":"fraction","uncertainty":null,"source_locator":"Table 1, MirTarRAW / 5-mer RNAret row, F1 column","review":{"method":"independent_ai_table_review","reviewer":"Codex secondary table review","reviewed_at":"2026-09-16T10:38:57.558224+00:00","notes":"Resolved the MirTarRAW section, 5-mer RNAret row, and F1 column (0.9622), distinct from DeepMirTarLeft F1 0.9728. Methods confirm 72/8/20 train/validation/test partition for MirTarRAW.","evidence":"Original PMC XML table headers, row groups and caption inspected; printed value 0.9622.","artifact_sha256":"e970e7322e07fb3c9d12efd315691cc5de5575a3f2616f4b788614c8c706dd0b","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC13111708/fullTextXML"},"legacy_id":"b2-rnaret-2026","legacy_row":{"id":"b2-rnaret-2026","paper_id":"rnaret-2026","domain_id":"rna-transcriptomes","task":"miRNA-mRNA interaction prediction","model":"RNAret","model_version":"5-mer","dataset":"MirTarRAW","dataset_version":"","split":"held-out test","metric":"F1","value":"0.9622","unit":"fraction","uncertainty":"","protocol":"5-mer RNAret classifier; 72/8/20 train/validation/test split","source_locator":"Table 1, MirTarRAW / 5-mer RNAret row, F1 column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC13111708/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:33:26Z"},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"b2-spin-protein-function-2026","kind":"result","name":"SPIN + ESM2-35M · F1 macro-weighted · TRX","description":"","status":"source_checked","facets":{"areas":["proteins-complexes"],"tasks":["protein function annotation"]},"source_ids":["spin-protein-function-2026"],"links":[{"relation":"evaluation","target_id":"evaluation-b2-spin-protein-function-2026"}],"attributes":{"printed_value":"0.796","numeric_value":"0.796","metric":"F1 macro-weighted","metric_direction":"unknown","unit":"fraction","uncertainty":null,"source_locator":"Table 1, ESM2-35M Test row, F1_m-w column","review":{"method":"independent_ai_table_review","reviewer":"Codex secondary table review","reviewed_at":"2026-09-16T10:38:57.558225+00:00","notes":"Resolved Test group and macro-weighted F1 subcolumn (0.796) for frozen ESM2-35M in SPIN. Test weighted accuracy is 0.798. Methods define inverse-frequency class weighting for macro-weighted F1.","evidence":"Original PMC XML table headers, row groups and caption inspected; printed value 0.796.","artifact_sha256":"9701843e93bf7fa3ead71e19693fb07d483f1022379871adfb04486783722a9d","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12970593/fullTextXML"},"legacy_id":"b2-spin-protein-function-2026","legacy_row":{"id":"b2-spin-protein-function-2026","paper_id":"spin-protein-function-2026","domain_id":"proteins-complexes","task":"protein function annotation","model":"SPIN + ESM2-35M","model_version":"ESM2-35M frozen","dataset":"TRX","dataset_version":"","split":"test set","metric":"F1 macro-weighted","value":"0.796","unit":"fraction","uncertainty":"","protocol":"frozen ESM2-35M backbone in SPIN","source_locator":"Table 1, ESM2-35M Test row, F1_m-w column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12970593/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:33:26Z"},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"b2-structure-informed-plm-2025","kind":"result","name":"structure-informed pLM · AUROC · variant-effects benchmark","description":"","status":"source_checked","facets":{"areas":["proteins-complexes"],"tasks":["protein variant-effect classification"]},"source_ids":["structure-informed-plm-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-b2-structure-informed-plm-2025"}],"attributes":{"printed_value":".803","numeric_value":"0.803","metric":"AUROC","metric_direction":"unknown","unit":"fraction","uncertainty":null,"source_locator":"PMC12068927 HTML, Table4, AA+SS+RSA+CM row, AUROC column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source review, not human review","reviewed_at":"2026-09-16T10:45:41.099916+00:00","notes":"Full-text HTML succeeds although EuropePMC XMLreturned404. Row is mutation-site variables AA+SS+RSA+CM, not neighbouring environment variant. AUROC .803 is numerically equivalent to preserved legacy0.803. Source check, not experimental reproduction; do not claim original source printed leading zero.","evidence":"Table4 headers: Type, Variable(s), Spearman rho, AUROC, AUPRC. Parsed HTML row: AA+SS+RSA+CM | .552 | .803 | .792. Primary web rendering independently confirms columns.","artifact_sha256":"76082e1cd992d2c09c38f86d05aba575cc76c5022b53a297123b713bb1ce9267","retrieval_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12068927/"},"legacy_id":"b2-structure-informed-plm-2025","legacy_row":{"id":"b2-structure-informed-plm-2025","paper_id":"structure-informed-plm-2025","domain_id":"proteins-complexes","task":"protein variant-effect classification","model":"structure-informed pLM","model_version":"not stated in table","dataset":"variant-effects benchmark","dataset_version":"","split":"paper evaluation","metric":"AUROC","value":"0.803","unit":"fraction","uncertainty":"","protocol":"combined amino-acid, secondary structure, solvent accessibility and contact-map scoring","source_locator":"Table 4, AA+SS+RSA+CM row, AUROC column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12068927/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:33:26Z"},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"barcodebert-2026","kind":"source","name":"BarcodeBERT: transformers for biodiversity analyses","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["dna-genomes"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC13008329/","version":"journal full text in PMC","retrieved_at":"2026-09-15T23:33:26Z","doi":"10.1093/bioadv/vbag054","publication_status":"peer_reviewed","year":2026,"artifact_sha256":"493f9fe70b483780ba76d51ccf217d3ca83539c82b89917fd3ccebe2b6eb831d","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC13008329/fullTextXML","artifact_retrieved_at":"2026-09-16T10:38:57.558051+00:00","legacy_paper":{"id":"barcodebert-2026","title":"BarcodeBERT: transformers for biodiversity analyses","year":2026,"publication_status":"peer_reviewed","version":"journal full text in PMC","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC13008329/","primary_domain":"dna-genomes","retrieved_utc":"2026-09-15T23:33:26Z","doi":"10.1093/bioadv/vbag054","notes":"Numeric result checked against Table 1 in primary full-text XML; journal/source: Bioinformatics Advances."},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"birna-bert-2025","kind":"source","name":"BiRNA-BERT allows efficient RNA language modeling with adaptive tokenization","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["rna-transcriptomes"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12635123/","version":"journal full text in PMC","retrieved_at":"2026-09-15T23:33:26Z","doi":"10.1038/s42003-025-08982-0","publication_status":"peer_reviewed","year":2025,"artifact_sha256":"bf7dbc52b6515301c77010c513f13e676c38395ddc82c20310171f518690c152","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12635123/fullTextXML","artifact_retrieved_at":"2026-09-16T10:33:38.292Z","legacy_paper":{"id":"birna-bert-2025","title":"BiRNA-BERT allows efficient RNA language modeling with adaptive tokenization","year":2025,"publication_status":"peer_reviewed","version":"journal full text in PMC","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12635123/","primary_domain":"rna-transcriptomes","retrieved_utc":"2026-09-15T23:33:26Z","doi":"10.1038/s42003-025-08982-0","notes":"Numeric result checked against Table 2 in primary full-text XML; journal/source: Communications Biology."},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"boltz-stereochemistry-2025","kind":"source","name":"Improving Stereochemical Limitations in Protein–Ligand Complex Structure Prediction","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["molecular-interactions"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12658688/","version":"version of record","retrieved_at":"2026-09-15T23:25:00Z","doi":"10.1021/acsomega.5c07675","publication_status":"peer_reviewed","year":2025,"artifact_sha256":"78a77b9a0ab8bfa371f5b9baef3f443f4590d6e71cf864d67e90e9ebdfa7fc1b","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12658688/fullTextXML","artifact_retrieved_at":"2026-09-16T10:41:16.555674+00:00","legacy_paper":{"id":"boltz-stereochemistry-2025","title":"Improving Stereochemical Limitations in Protein–Ligand Complex Structure Prediction","year":2025,"publication_status":"peer_reviewed","version":"version of record","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12658688/","primary_domain":"molecular-interactions","retrieved_utc":"2026-09-15T23:25:00Z","notes":"Primary full text via Europe PMC XML; venue: ACS Omega; PMC ID: PMC12658688.","doi":"10.1021/acsomega.5c07675"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"boltz1-2025","kind":"source","name":"Boltz-1 Democratizing Biomolecular Interaction Modeling","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["molecular-interactions"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11601547/","version":"PMC archival version PMC11601547.4","retrieved_at":"2026-09-15T23:29:32Z","doi":"10.1101/2024.11.19.624167","publication_status":"preprint","year":2025,"artifact_sha256":"1ebf712314d9a1c678ded989cc95a0c00c0331e5ad8c9f63194bc9780971d214","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11601547/fullTextXML","artifact_retrieved_at":"2026-09-16T10:44:03.424864+00:00","legacy_paper":{"id":"boltz1-2025","title":"Boltz-1 Democratizing Biomolecular Interaction Modeling","year":2025,"publication_status":"preprint","version":"PMC archival version PMC11601547.4","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11601547/","primary_domain":"molecular-interactions","retrieved_utc":"2026-09-15T23:29:32Z","notes":"Primary full text verified using Europe PMC XML; venue: bioRxiv; PMC ID: PMC11601547.","doi":"10.1101/2024.11.19.624167"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"bpfold-2025","kind":"source","name":"Deep generalizable prediction of RNA secondary structure via base pair motif energy","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["rna-transcriptomes"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12216785/","version":"version of record","retrieved_at":"2026-09-15T23:25:00Z","doi":"10.1038/s41467-025-60048-1","publication_status":"peer_reviewed","year":2025,"artifact_sha256":"976218bd172998a1a6e7ed1609ecb8cb2ee380fb48a8dc7b25bc05ea8b0a49af","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12216785/fullTextXML","artifact_retrieved_at":"2026-09-16T10:41:16.502000+00:00","legacy_paper":{"id":"bpfold-2025","title":"Deep generalizable prediction of RNA secondary structure via base pair motif energy","year":2025,"publication_status":"peer_reviewed","version":"version of record","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12216785/","primary_domain":"rna-transcriptomes","retrieved_utc":"2026-09-15T23:25:00Z","notes":"Primary full text via Europe PMC XML; venue: Nature Communications; PMC ID: PMC12216785.","doi":"10.1038/s41467-025-60048-1"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"cammiq-2022","kind":"source","name":"Strain level microbial detection and quantification with applications to single cell metagenomics","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["microbes-communities"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC9616933/","version":"version of record","retrieved_at":"2026-09-15T23:29:32Z","doi":"10.1038/s41467-022-33869-7","publication_status":"peer_reviewed","year":2022,"artifact_sha256":"f0939647ed3de995d58254f79472a612c21b0e1b2560a82783302aa1a148dde3","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC9616933/fullTextXML","artifact_retrieved_at":"2026-09-16T10:44:03.408237+00:00","legacy_paper":{"id":"cammiq-2022","title":"Strain level microbial detection and quantification with applications to single cell metagenomics","year":2022,"publication_status":"peer_reviewed","version":"version of record","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC9616933/","primary_domain":"microbes-communities","retrieved_utc":"2026-09-15T23:29:32Z","notes":"Primary full text verified using Europe PMC XML; venue: Nature Communications; PMC ID: PMC9616933.","doi":"10.1038/s41467-022-33869-7"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"catalog-baseline-kraken2","kind":"baseline","name":"Kraken2","description":"","status":"discovered","facets":{"areas":["microbes-communities"]},"source_ids":["catalog-source-kraken2"],"links":[{"relation":"model","target_id":"catalog-model-kraken2"},{"relation":"applicable_to","target_id":"catalog-task-heldout-clade"},{"relation":"applicable_to","target_id":"catalog-task-phage-pathogen-reads"}],"attributes":{"baseline_type":"established_method","applicability":"proposed","requirements":"Public classifier; database build/version must be pinned separately.","missing_metadata":{"exact_protocol":"not_yet_extracted"}}}
{"id":"catalog-baseline-scvi","kind":"baseline","name":"scVI","description":"","status":"discovered","facets":{"areas":["cells-tissues"]},"source_ids":["catalog-source-scvi"],"links":[{"relation":"model","target_id":"catalog-model-scvi"},{"relation":"applicable_to","target_id":"catalog-task-cell-reference-mapping"},{"relation":"applicable_to","target_id":"catalog-task-cell-batch-integration"}],"attributes":{"baseline_type":"established_method","applicability":"proposed","requirements":"Public software; train a task-specific model on the permitted split.","missing_metadata":{"exact_protocol":"not_yet_extracted"}}}
{"id":"catalog-baseline-vina","kind":"baseline","name":"AutoDock Vina","description":"","status":"discovered","facets":{"areas":["molecular-interactions"]},"source_ids":["catalog-source-vina"],"links":[{"relation":"model","target_id":"catalog-model-vina"},{"relation":"applicable_to","target_id":"catalog-task-ligand-pose"}],"attributes":{"baseline_type":"established_method","applicability":"proposed","requirements":"Public docking software; receptor and ligand preparation required.","missing_metadata":{"exact_protocol":"not_yet_extracted"}}}
{"id":"catalog-model-alphafold-3-server","kind":"model","name":"AlphaFold 3 Server","description":"Existing candidate catalogue entry; exact checkpoint and metadata require primary-source review.","status":"discovered","facets":{"areas":["molecular-interactions"],"method_types":["foundation model"]},"source_ids":["catalog-source-alphafold-3-server"],"links":[{"relation":"uses_model","target_id":"discovery-model-alphafold-3"}],"attributes":{"entity_level":"service","version":"hosted server","reported_name":"AlphaFold 3 Server","access":"Manual, non-commercial server access; output terms restrict automated docking combinations.","method_type":"foundation model","historical_missing_metadata":{"checkpoint_revision":"not_yet_extracted","training_data":"not_yet_extracted","licence":"not_yet_extracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"AlphaFold Server is Google DeepMind’s hosted interface to AlphaFold 3. It predicts molecular complex structures without a local installation and returns confidence estimates. Its service limits, supported inputs and output terms are separate from the downloadable model.","summary_source_ids":["evidence-alphafold-server-faq","evidence-alphafold-server-terms"],"summary_source_locator":"FAQ: supported molecules, job size, outputs; Terms: Overview","sections":[{"title":"Joint structure prediction","body":"Sequence, chemical and evolutionary features feed a Pairformer, which builds representations of individual tokens and their relationships. A diffusion module then predicts atomic coordinates. Separate heads estimate confidence. The paper describes 48 Pairformer blocks; the architecture models complexes jointly rather than treating every partner as a separately folded structure.","source_ids":["evidence-alphafold-paper"],"source_locator":"Main text: Model architecture; Fig. 1d and Fig. 2"},{"title":"What the server adds","body":"The service handles the input preparation and presents ranked structures and confidence outputs for download. The current FAQ supports custom protein MSAs and templates, as well as ligands specified by CCD code. These options must be recorded when interpreting a result; a server run is not automatically the same configuration as the original paper.","source_ids":["evidence-alphafold-server-faq"],"source_locator":"What structure templates and MSA are used?; How do I add a ligand using its CCD code?; How many predictions are returned?"},{"title":"Service identity and reproducibility","body":"The FAQ states that the server and released AlphaFold 3 use the same weights and equivalent model code. Their genetic-search implementations can nevertheless produce different alignments, so matching weights does not make every run equivalent. The hosted checkpoint digest is not disclosed, and compiler changes can affect exact repeatability. Preserve job inputs, seeds and outputs when citing an evaluation.","source_ids":["evidence-alphafold-server-faq"],"source_locator":"Why might Multiple Sequence Alignments differ between AlphaFold 3 and AlphaFold Server?; job repeatability and seed questions"}],"facts":[{"label":"Model type","value":"Hosted AlphaFold 3 structure-prediction service.","status":"source_checked","source_ids":["evidence-alphafold-server-terms"],"source_locator":"Overview"},{"label":"Architecture","value":"A 48-block Pairformer builds token and pair representations; a diffusion module predicts atomic coordinates and separate heads estimate confidence.","status":"source_checked","source_ids":["evidence-alphafold-paper"],"source_locator":"Model architecture; Fig. 1d and Fig. 2"},{"label":"Known versions","value":"The FAQ states that the released AlphaFold 3 and server use the same weights and equivalent model code. It does not supply a public digest identifying the hosted checkpoint.","status":"source_checked","source_ids":["evidence-alphafold-server-faq"],"source_locator":"Why might Multiple Sequence Alignments differ between AlphaFold 3 and AlphaFold Server?"},{"label":"Inputs","value":"Protein, DNA and RNA sequences, supported modifications, ions and ligands. Current FAQ also permits additional ligands by CCD code and custom protein MSAs/templates.","status":"source_checked","source_ids":["evidence-alphafold-server-faq"],"source_locator":"Supported molecule types; CCD code; template and MSA questions"},{"label":"Outputs","value":"Five predictions per seed, downloadable structures and confidence JSON; the top-ranked prediction is shown in the interface.","status":"source_checked","source_ids":["evidence-alphafold-server-faq"],"source_locator":"How many predictions are returned?; downloaded JSON files"},{"label":"Training data","value":"Uses AlphaFold 3. The paper describes PDB-based training and distillation; the server FAQ does not provide a separate hosted-checkpoint training inventory.","status":"source_checked","source_ids":["evidence-alphafold-paper","evidence-alphafold-server-faq"],"source_locator":"Paper Methods: Training regime; FAQ: templates and MSA"},{"label":"Training cutoff","value":"The paper gives 2021-09-30 for the standard model’s structural training data. The FAQ does not publish a complete dated training manifest for the hosted deployment; its template-search cutoff is a separate input setting.","status":"source_checked","source_ids":["evidence-alphafold-paper","evidence-alphafold-server-faq"],"source_locator":"Paper Methods: Training regime; FAQ: template and MSA settings and comparison with released AlphaFold 3"},{"label":"Context limits","value":"5,000 tokens per job. Residues, nucleotides and molecular atoms count differently; this is a hosted-service limit.","status":"source_checked","source_ids":["evidence-alphafold-server-faq"],"source_locator":"What is the maximum job size allowed?"},{"label":"Access","value":"Google-account web service for non-commercial use. FAQ reports 30 jobs per day as checked on 2026-09-16; quotas may change.","status":"source_checked","source_ids":["evidence-alphafold-server-faq","evidence-alphafold-server-terms"],"source_locator":"How many jobs can I run?; Terms: Key things to know"},{"label":"Code licence","value":"The service is governed by its Terms of Service. Apache-2.0 applies to the separate downloadable inference code, not to ownership or licensing of the hosted service.","status":"source_checked","source_ids":["evidence-alphafold-server-terms","evidence-alphafold-license"],"source_locator":"Terms: Overview; separate implementation LICENSE"},{"label":"Weights licence","value":"Hosted use does not supply a weights licence. Downloadable model parameters are a separate offering under their own terms.","status":"inapplicable","source_ids":["evidence-alphafold-server-terms","evidence-alphafold-weights-terms-of-use"],"source_locator":"Service Terms; Model Parameters Terms"},{"label":"Output terms","value":"Non-commercial use restrictions apply. Terms prohibit use with automated protein–ligand/peptide interaction prediction systems and training similar structure-prediction models; downstream notices are required.","status":"source_checked","source_ids":["evidence-alphafold-server-output-terms"],"source_locator":"Use restrictions"},{"label":"Parameters","value":"The hosted checkpoint’s total parameter count is not supplied in the inspected FAQ. No number is inferred from the model name.","status":"unreported","source_ids":["evidence-alphafold-server-faq"],"source_locator":"FAQ: model identity and access; no total parameter count"}],"strengths":[{"text":"Provides a hosted route to predictions and downloadable confidence outputs without maintaining the local inference installation.","source_ids":["evidence-alphafold-server-faq","evidence-alphafold-server-terms"],"source_locator":"Terms: Overview; FAQ: downloaded outputs"},{"text":"Current input controls expose custom protein MSAs and templates, which helps users document those inputs.","source_ids":["evidence-alphafold-server-faq"],"source_locator":"What structure templates and MSA are used?"}],"limitations":[{"text":"Predictions can contain incorrect chirality, atomic clashes or spurious structure in disordered regions. Confidence and structural plausibility need separate inspection.","source_ids":["evidence-alphafold-paper"],"source_locator":"Model limitations; Fig. 5"},{"text":"Sampled structures are not a calibrated solution-state ensemble. Prediction confidence does not establish binding affinity or experimental function.","source_ids":["evidence-alphafold-paper"],"source_locator":"Model limitations: dynamics and conformational states; confidence outputs are structure-quality estimates"},{"text":"Usage quotas and server terms constrain access and downstream reuse; the service is not an unrestricted batch prediction API.","source_ids":["evidence-alphafold-server-faq","evidence-alphafold-server-terms","evidence-alphafold-server-output-terms"],"source_locator":"Daily quota; Terms and Output Terms: use restrictions"},{"text":"A hosted result cannot establish performance for an unspecified local checkpoint or paper evaluation variant.","source_ids":["evidence-alphafold-server-faq","evidence-alphafold-paper"],"source_locator":"FAQ reproducibility; paper Methods: separate evaluation variants"}],"diagram":{"title":"AlphaFold Server workflow","steps":["Specify molecules and input settings","Hosted preparation of features","AlphaFold 3 structure prediction","Rank samples and inspect confidence","Download structures and job records"],"caption":"Conceptual service workflow. The deployed checkpoint is not pinned by the public FAQ.","source_ids":["evidence-alphafold-server-faq"],"source_locator":"Input, templates/MSAs, returned predictions and download questions"},"coverage":"reviewed","gaps":["Exact hosted checkpoint digest and parameter count remain unreported in the inspected public FAQ.","Service quotas and input options are time-dependent; checked on 2026-09-16."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Read primary paper XML, pinned official repository documentation and licences. Reviewed public server FAQ separately. No model run, independent performance replication or human review. A second automated reviewer checked the AlphaFold source claims and service/model distinction; this is not human review or experimental reproduction."}}}}
{"id":"catalog-model-alphagenome","kind":"model","name":"AlphaGenome","description":"Existing candidate catalogue entry; exact checkpoint and metadata require primary-source review.","status":"discovered","facets":{"areas":["dna-genomes"],"method_types":["foundation model"]},"source_ids":["catalog-source-alphagenome"],"links":[],"attributes":{"entity_level":"family","version":"API / released weights","reported_name":"AlphaGenome","access":"Rate-limited, non-commercial API requires a key. Downloadable weights require accepting non-commercial model terms; local inference recommends an H100 GPU.","method_type":"foundation model","historical_missing_metadata":{"checkpoint_revision":"not_yet_extracted","training_data":"not_yet_extracted","licence":"not_yet_extracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"AlphaGenome predicts regulatory activity and variant effects from long DNA sequences, with outputs for expression, splicing, chromatin and contact maps.","summary_source_ids":["evidence-official-578439da3f3a3a476b7f","evidence-official-56e5abfb5f12f1cd3b20","evidence-official-bc0327ef2ea6107e1773"],"summary_source_locator":"Paper: Unifying DNA sequence-to-function model and Discussion; README.md: Model weights and licensing; Supplementary Methods: Model (p.9), ENCODE RNA-seq Data (p.3), Contact maps","sections":[{"title":"How it works","body":"AlphaGenome progressively downsamples DNA with convolutional blocks, then uses a transformer tower and pairwise interaction blocks to represent long-range context. A U-Net-style decoder restores sequence resolution using skip connections. Modality-specific heads predict one-dimensional genomic tracks, splicing outputs and two-dimensional contact maps.","source_ids":["evidence-official-578439da3f3a3a476b7f","evidence-official-56e5abfb5f12f1cd3b20","evidence-official-bc0327ef2ea6107e1773"],"source_locator":"Paper: Unifying DNA sequence-to-function model and Discussion; README.md: Model weights and licensing; Supplementary Methods: Model (p.9), ENCODE RNA-seq Data (p.3), Contact maps"},{"title":"Versions and reproducibility","body":"The paper distinguishes fold-specific evaluation models, all-fold teachers and distilled students; a family name alone does not choose one of these configurations. Up to 1 million base pairs; single-base outputs for the modalities described in the README.","source_ids":["evidence-official-578439da3f3a3a476b7f","evidence-official-56e5abfb5f12f1cd3b20","evidence-official-bc0327ef2ea6107e1773"],"source_locator":"Paper: Unifying DNA sequence-to-function model and Discussion; README.md: Model weights and licensing; Supplementary Methods: Model (p.9), ENCODE RNA-seq Data (p.3), Contact maps"}],"facts":[{"label":"Model type","value":"Sequence-to-function convolutional/transformer model","status":"source_checked","source_ids":["evidence-official-578439da3f3a3a476b7f","evidence-official-56e5abfb5f12f1cd3b20","evidence-official-bc0327ef2ea6107e1773"],"source_locator":"Paper: Unifying DNA sequence-to-function model and Discussion; README.md: Model weights and licensing; Supplementary Methods: Model (p.9), ENCODE RNA-seq Data (p.3), Contact maps"},{"label":"Architecture","value":"U-Net-inspired sequence backbone combining convolutional local processing with transformer blocks for longer-range interactions; one-dimensional track heads and two-dimensional contact-map representations.","status":"source_checked","source_ids":["evidence-official-578439da3f3a3a476b7f","evidence-official-56e5abfb5f12f1cd3b20","evidence-official-bc0327ef2ea6107e1773"],"source_locator":"Paper: Unifying DNA sequence-to-function model and Discussion; README.md: Model weights and licensing; Supplementary Methods: Model (p.9), ENCODE RNA-seq Data (p.3), Contact maps"},{"label":"Inputs","value":"DNA sequence and optional variant information.","status":"source_checked","source_ids":["evidence-official-578439da3f3a3a476b7f","evidence-official-56e5abfb5f12f1cd3b20","evidence-official-bc0327ef2ea6107e1773"],"source_locator":"Paper: Unifying DNA sequence-to-function model and Discussion; README.md: Model weights and licensing; Supplementary Methods: Model (p.9), ENCODE RNA-seq Data (p.3), Contact maps"},{"label":"Outputs","value":"Predicted expression, splicing, chromatin features and contact maps.","status":"source_checked","source_ids":["evidence-official-578439da3f3a3a476b7f","evidence-official-56e5abfb5f12f1cd3b20","evidence-official-bc0327ef2ea6107e1773"],"source_locator":"Paper: Unifying DNA sequence-to-function model and Discussion; README.md: Model weights and licensing; Supplementary Methods: Model (p.9), ENCODE RNA-seq Data (p.3), Contact maps"},{"label":"Parameters","value":"Approximately 450M trainable parameters, including encoder, sequence transformer, pairwise blocks, decoder and prediction heads.","status":"source_checked","source_ids":["evidence-official-578439da3f3a3a476b7f","evidence-official-56e5abfb5f12f1cd3b20","evidence-official-bc0327ef2ea6107e1773"],"source_locator":"Paper: Unifying DNA sequence-to-function model and Discussion; README.md: Model weights and licensing; Supplementary Methods: Model (p.9), ENCODE RNA-seq Data (p.3), Contact maps"},{"label":"Known versions","value":"The paper distinguishes fold-specific evaluation models, all-fold teachers and distilled students; a family name alone does not choose one of these configurations.","status":"source_checked","source_ids":["evidence-official-578439da3f3a3a476b7f","evidence-official-56e5abfb5f12f1cd3b20","evidence-official-bc0327ef2ea6107e1773"],"source_locator":"Paper: Unifying DNA sequence-to-function model and Discussion; README.md: Model weights and licensing; Supplementary Methods: Model (p.9), ENCODE RNA-seq Data (p.3), Contact maps"},{"label":"Training data","value":"Human and mouse molecular datasets; separate fold-specific models for held-out reference-interval evaluation and all-fold teachers for student distillation.","status":"source_checked","source_ids":["evidence-official-578439da3f3a3a476b7f","evidence-official-56e5abfb5f12f1cd3b20","evidence-official-bc0327ef2ea6107e1773"],"source_locator":"Paper: Unifying DNA sequence-to-function model and Discussion; README.md: Model weights and licensing; Supplementary Methods: Model (p.9), ENCODE RNA-seq Data (p.3), Contact maps"},{"label":"Training cutoff","value":"ENCODE RNA-seq and chromatin metadata were downloaded 9–17 January 2025; the contact-map source was accessed 4 March 2021. These are component retrieval dates, not one universal latest-experiment cutoff.","status":"source_checked","source_ids":["evidence-official-578439da3f3a3a476b7f","evidence-official-56e5abfb5f12f1cd3b20","evidence-official-bc0327ef2ea6107e1773"],"source_locator":"Paper: Unifying DNA sequence-to-function model and Discussion; README.md: Model weights and licensing; Supplementary Methods: Model (p.9), ENCODE RNA-seq Data (p.3), Contact maps"},{"label":"Context limits","value":"Up to 1 million base pairs; single-base outputs for the modalities described in the README.","status":"source_checked","source_ids":["evidence-official-578439da3f3a3a476b7f","evidence-official-56e5abfb5f12f1cd3b20","evidence-official-bc0327ef2ea6107e1773"],"source_locator":"Paper: Unifying DNA sequence-to-function model and Discussion; README.md: Model weights and licensing; Supplementary Methods: Model (p.9), ENCODE RNA-seq Data (p.3), Contact maps"},{"label":"Weights licence","value":"Non-commercial AlphaGenome model terms; not the Apache licence covering source code.","status":"source_checked","source_ids":["evidence-official-578439da3f3a3a476b7f","evidence-official-56e5abfb5f12f1cd3b20","evidence-official-bc0327ef2ea6107e1773"],"source_locator":"Paper: Unifying DNA sequence-to-function model and Discussion; README.md: Model weights and licensing; Supplementary Methods: Model (p.9), ENCODE RNA-seq Data (p.3), Contact maps"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/google-deepmind/alphagenome_research","status":"source_checked","source_ids":["evidence-official-578439da3f3a3a476b7f","evidence-official-56e5abfb5f12f1cd3b20","evidence-official-bc0327ef2ea6107e1773"],"source_locator":"Paper: Unifying DNA sequence-to-function model and Discussion; README.md: Model weights and licensing; Supplementary Methods: Model (p.9), ENCODE RNA-seq Data (p.3), Contact maps"},{"label":"Code licence","value":"Apache-2.0","status":"source_checked","source_ids":["evidence-official-c3d7a64294dcad7c6405"],"source_locator":"LICENSE: licence text"}],"strengths":[{"text":"Combines several molecular readouts in one sequence model, with variant scoring utilities.","source_ids":["evidence-official-578439da3f3a3a476b7f","evidence-official-56e5abfb5f12f1cd3b20","evidence-official-bc0327ef2ea6107e1773"],"source_locator":"Paper: Unifying DNA sequence-to-function model and Discussion; README.md: Model weights and licensing; Supplementary Methods: Model (p.9), ENCODE RNA-seq Data (p.3), Contact maps"}],"limitations":[{"text":"The paper identifies remaining challenges beyond 100kb, context-specific variant effects and non-coding genes. Species coverage is human/mouse and personal-genome prediction was not benchmarked in the study.","source_ids":["evidence-official-578439da3f3a3a476b7f","evidence-official-56e5abfb5f12f1cd3b20","evidence-official-bc0327ef2ea6107e1773"],"source_locator":"Paper: Unifying DNA sequence-to-function model and Discussion; README.md: Model weights and licensing; Supplementary Methods: Model (p.9), ENCODE RNA-seq Data (p.3), Contact maps"}],"diagram":{"title":"AlphaGenome workflow","steps":["DNA and species identifier","Convolutional sequence encoder","Transformer and pairwise blocks","Decoder with skip connections","Track, splicing and contact-map heads"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-578439da3f3a3a476b7f","evidence-official-56e5abfb5f12f1cd3b20","evidence-official-bc0327ef2ea6107e1773"],"source_locator":"Paper: Unifying DNA sequence-to-function model and Discussion; README.md: Model weights and licensing; Supplementary Methods: Model (p.9), ENCODE RNA-seq Data (p.3), Contact maps"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}}}
{"id":"catalog-model-boltz-2","kind":"model","name":"Boltz-2","description":"Existing candidate catalogue entry; exact checkpoint and metadata require primary-source review.","status":"discovered","facets":{"areas":["molecular-interactions"],"method_types":["foundation model"]},"source_ids":["catalog-source-boltz-2"],"links":[{"relation":"variant_of","target_id":"discovery-model-boltz"}],"attributes":{"entity_level":"family","version":"released weights","reported_name":"Boltz-2","access":"Public MIT code and weights; substantial compute required.","method_type":"foundation model","historical_missing_metadata":{"checkpoint_revision":"not_yet_extracted","training_data":"not_yet_extracted","licence":"not_yet_extracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Boltz predicts biomolecular complex structures; Boltz-2 also predicts binding affinity.","summary_source_ids":["evidence-official-f780521ccebbb0691f9e","evidence-official-5a9e55abdf288d9dd8da","evidence-official-39a15ed8072aeea14a22","evidence-official-734117ed8501eb26fefc","evidence-official-7cb8cb7f091536b92c11","evidence-official-bd6ec1f819e22f075c32"],"summary_source_locator":"src/boltz/model/models/boltz2.py: Boltz2.__init__, forward, PairformerModule, AtomDiffusion and AffinityModule; docs/prediction.md: affinity outputs; Boltz-2 report Sections 2 Data, 3 Architecture, 4 Training and 6 Limitations","sections":[{"title":"How it works","body":"Boltz-2 first encodes the molecular inputs, alignments and optional templates into token and pair features. A Pairformer trunk updates those features and conditions atom-coordinate diffusion to generate a complex. Separate confidence and affinity modules assess the prediction; affinity classification and regression outputs answer different questions.","source_ids":["evidence-official-f780521ccebbb0691f9e","evidence-official-5a9e55abdf288d9dd8da","evidence-official-39a15ed8072aeea14a22","evidence-official-734117ed8501eb26fefc","evidence-official-7cb8cb7f091536b92c11","evidence-official-bd6ec1f819e22f075c32"],"source_locator":"src/boltz/model/models/boltz2.py: Boltz2.__init__, forward, PairformerModule, AtomDiffusion and AffinityModule; docs/prediction.md: affinity outputs; Boltz-2 report Sections 2 Data, 3 Architecture, 4 Training and 6 Limitations"},{"title":"Versions and reproducibility","body":"Boltz-1 and Boltz-2 are distinct released generations; the catalogue does not select an evaluated checkpoint. The Boltz-2 report describes training crops up to 768 tokens. This is a training-crop size rather than a universal inference maximum; affinity additionally uses a pocket crop.","source_ids":["evidence-official-f780521ccebbb0691f9e","evidence-official-5a9e55abdf288d9dd8da","evidence-official-39a15ed8072aeea14a22","evidence-official-734117ed8501eb26fefc","evidence-official-7cb8cb7f091536b92c11","evidence-official-bd6ec1f819e22f075c32"],"source_locator":"src/boltz/model/models/boltz2.py: Boltz2.__init__, forward, PairformerModule, AtomDiffusion and AffinityModule; docs/prediction.md: affinity outputs; Boltz-2 report Sections 2 Data, 3 Architecture, 4 Training and 6 Limitations"}],"facts":[{"label":"Model type","value":"Biomolecular structure and affinity predictor","status":"source_checked","source_ids":["evidence-official-f780521ccebbb0691f9e","evidence-official-5a9e55abdf288d9dd8da","evidence-official-39a15ed8072aeea14a22","evidence-official-734117ed8501eb26fefc","evidence-official-7cb8cb7f091536b92c11","evidence-official-bd6ec1f819e22f075c32"],"source_locator":"src/boltz/model/models/boltz2.py: Boltz2.__init__, forward, PairformerModule, AtomDiffusion and AffinityModule; docs/prediction.md: affinity outputs; Boltz-2 report Sections 2 Data, 3 Architecture, 4 Training and 6 Limitations"},{"label":"Architecture","value":"Molecular input embeddings, MSA and optional template modules build single-token and pair representations. Pairformer blocks refine these features; atom-coordinate diffusion generates structures, with separate confidence and binding-affinity modules.","status":"source_checked","source_ids":["evidence-official-f780521ccebbb0691f9e","evidence-official-5a9e55abdf288d9dd8da","evidence-official-39a15ed8072aeea14a22","evidence-official-734117ed8501eb26fefc","evidence-official-7cb8cb7f091536b92c11","evidence-official-bd6ec1f819e22f075c32"],"source_locator":"src/boltz/model/models/boltz2.py: Boltz2.__init__, forward, PairformerModule, AtomDiffusion and AffinityModule; docs/prediction.md: affinity outputs; Boltz-2 report Sections 2 Data, 3 Architecture, 4 Training and 6 Limitations"},{"label":"Inputs","value":"Protein, nucleic-acid and ligand specifications in prediction input files.","status":"source_checked","source_ids":["evidence-official-f780521ccebbb0691f9e","evidence-official-5a9e55abdf288d9dd8da","evidence-official-39a15ed8072aeea14a22","evidence-official-734117ed8501eb26fefc","evidence-official-7cb8cb7f091536b92c11","evidence-official-bd6ec1f819e22f075c32"],"source_locator":"src/boltz/model/models/boltz2.py: Boltz2.__init__, forward, PairformerModule, AtomDiffusion and AffinityModule; docs/prediction.md: affinity outputs; Boltz-2 report Sections 2 Data, 3 Architecture, 4 Training and 6 Limitations"},{"label":"Outputs","value":"Predicted complex structures and, for supported Boltz-2 inputs, binding-affinity predictions.","status":"source_checked","source_ids":["evidence-official-f780521ccebbb0691f9e","evidence-official-5a9e55abdf288d9dd8da","evidence-official-39a15ed8072aeea14a22","evidence-official-734117ed8501eb26fefc","evidence-official-7cb8cb7f091536b92c11","evidence-official-bd6ec1f819e22f075c32"],"source_locator":"src/boltz/model/models/boltz2.py: Boltz2.__init__, forward, PairformerModule, AtomDiffusion and AffinityModule; docs/prediction.md: affinity outputs; Boltz-2 report Sections 2 Data, 3 Architecture, 4 Training and 6 Limitations"},{"label":"Parameters","value":"A complete parameter total is not stated in the reviewed Boltz-2 architecture report or model constructor; structure, confidence and affinity are separate modules.","status":"unreported","source_ids":["evidence-official-f780521ccebbb0691f9e","evidence-official-5a9e55abdf288d9dd8da","evidence-official-39a15ed8072aeea14a22","evidence-official-734117ed8501eb26fefc","evidence-official-7cb8cb7f091536b92c11","evidence-official-bd6ec1f819e22f075c32"],"source_locator":"src/boltz/model/models/boltz2.py: Boltz2.__init__, forward, PairformerModule, AtomDiffusion and AffinityModule; docs/prediction.md: affinity outputs; Boltz-2 report Sections 2 Data, 3 Architecture, 4 Training and 6 Limitations"},{"label":"Known versions","value":"Boltz-1 and Boltz-2 are distinct released generations; the catalogue does not select an evaluated checkpoint.","status":"source_checked","source_ids":["evidence-official-f780521ccebbb0691f9e","evidence-official-5a9e55abdf288d9dd8da","evidence-official-39a15ed8072aeea14a22","evidence-official-734117ed8501eb26fefc","evidence-official-7cb8cb7f091536b92c11","evidence-official-bd6ec1f819e22f075c32"],"source_locator":"src/boltz/model/models/boltz2.py: Boltz2.__init__, forward, PairformerModule, AtomDiffusion and AffinityModule; docs/prediction.md: affinity outputs; Boltz-2 report Sections 2 Data, 3 Architecture, 4 Training and 6 Limitations"},{"label":"Training data","value":"Boltz-2 structure training combines pre-June-2023 PDB entries, MISATO/ATLAS/mdCATH molecular dynamics, and AlphaFold2/Boltz-1 distillation. Separate affinity training uses curated PubChem, ChEMBL, BindingDB, HTS, CeMM and MIDAS evidence with different regression/classification labels.","status":"source_checked","source_ids":["evidence-official-f780521ccebbb0691f9e","evidence-official-5a9e55abdf288d9dd8da","evidence-official-39a15ed8072aeea14a22","evidence-official-734117ed8501eb26fefc","evidence-official-7cb8cb7f091536b92c11","evidence-official-bd6ec1f819e22f075c32"],"source_locator":"src/boltz/model/models/boltz2.py: Boltz2.__init__, forward, PairformerModule, AtomDiffusion and AffinityModule; docs/prediction.md: affinity outputs; Boltz-2 report Sections 2 Data, 3 Architecture, 4 Training and 6 Limitations"},{"label":"Training cutoff","value":"Boltz-2 experimental PDB structures were released before 2023-06-01. This is not a shared cutoff for every affinity, MD or distilled resource, nor a Boltz-1 training cutoff.","status":"source_checked","source_ids":["evidence-official-f780521ccebbb0691f9e","evidence-official-5a9e55abdf288d9dd8da","evidence-official-39a15ed8072aeea14a22","evidence-official-734117ed8501eb26fefc","evidence-official-7cb8cb7f091536b92c11","evidence-official-bd6ec1f819e22f075c32"],"source_locator":"src/boltz/model/models/boltz2.py: Boltz2.__init__, forward, PairformerModule, AtomDiffusion and AffinityModule; docs/prediction.md: affinity outputs; Boltz-2 report Sections 2 Data, 3 Architecture, 4 Training and 6 Limitations"},{"label":"Context limits","value":"The Boltz-2 report describes training crops up to 768 tokens. This is a training-crop size rather than a universal inference maximum; affinity additionally uses a pocket crop.","status":"source_checked","source_ids":["evidence-official-f780521ccebbb0691f9e","evidence-official-5a9e55abdf288d9dd8da","evidence-official-39a15ed8072aeea14a22","evidence-official-734117ed8501eb26fefc","evidence-official-7cb8cb7f091536b92c11","evidence-official-bd6ec1f819e22f075c32"],"source_locator":"src/boltz/model/models/boltz2.py: Boltz2.__init__, forward, PairformerModule, AtomDiffusion and AffinityModule; docs/prediction.md: affinity outputs; Boltz-2 report Sections 2 Data, 3 Architecture, 4 Training and 6 Limitations"},{"label":"Weights licence","value":"MIT; the README explicitly applies this licence to code and model weights.","status":"source_checked","source_ids":["evidence-official-f780521ccebbb0691f9e","evidence-official-5a9e55abdf288d9dd8da","evidence-official-39a15ed8072aeea14a22","evidence-official-734117ed8501eb26fefc","evidence-official-7cb8cb7f091536b92c11","evidence-official-bd6ec1f819e22f075c32"],"source_locator":"src/boltz/model/models/boltz2.py: Boltz2.__init__, forward, PairformerModule, AtomDiffusion and AffinityModule; docs/prediction.md: affinity outputs; Boltz-2 report Sections 2 Data, 3 Architecture, 4 Training and 6 Limitations"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/jwohlwend/boltz","status":"source_checked","source_ids":["evidence-official-f780521ccebbb0691f9e","evidence-official-5a9e55abdf288d9dd8da","evidence-official-39a15ed8072aeea14a22","evidence-official-734117ed8501eb26fefc","evidence-official-7cb8cb7f091536b92c11","evidence-official-bd6ec1f819e22f075c32"],"source_locator":"src/boltz/model/models/boltz2.py: Boltz2.__init__, forward, PairformerModule, AtomDiffusion and AffinityModule; docs/prediction.md: affinity outputs; Boltz-2 report Sections 2 Data, 3 Architecture, 4 Training and 6 Limitations"},{"label":"Code licence","value":"MIT","status":"source_checked","source_ids":["evidence-official-8a985eabfd054f0dec4d"],"source_locator":"LICENSE: licence text"}],"strengths":[{"text":"The project distributes prediction code, model weights and training instructions.","source_ids":["evidence-official-f780521ccebbb0691f9e","evidence-official-5a9e55abdf288d9dd8da","evidence-official-39a15ed8072aeea14a22","evidence-official-734117ed8501eb26fefc","evidence-official-7cb8cb7f091536b92c11","evidence-official-bd6ec1f819e22f075c32"],"source_locator":"src/boltz/model/models/boltz2.py: Boltz2.__init__, forward, PairformerModule, AtomDiffusion and AffinityModule; docs/prediction.md: affinity outputs; Boltz-2 report Sections 2 Data, 3 Architecture, 4 Training and 6 Limitations"}],"limitations":[{"text":"Affinity predictions depend on a plausible binding pose and mix biochemical endpoint types. The report notes limited handling of cofactors, water and multimeric binding partners, and substantial variation between assays.","source_ids":["evidence-official-f780521ccebbb0691f9e","evidence-official-5a9e55abdf288d9dd8da","evidence-official-39a15ed8072aeea14a22","evidence-official-734117ed8501eb26fefc","evidence-official-7cb8cb7f091536b92c11","evidence-official-bd6ec1f819e22f075c32"],"source_locator":"src/boltz/model/models/boltz2.py: Boltz2.__init__, forward, PairformerModule, AtomDiffusion and AffinityModule; docs/prediction.md: affinity outputs; Boltz-2 report Sections 2 Data, 3 Architecture, 4 Training and 6 Limitations"}],"diagram":{"title":"Boltz-2 workflow","steps":["Molecular inputs, MSA and templates","Token and pair embeddings","Pairformer trunk","Atom-coordinate diffusion","Structure, confidence and affinity"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-f780521ccebbb0691f9e","evidence-official-5a9e55abdf288d9dd8da","evidence-official-39a15ed8072aeea14a22","evidence-official-734117ed8501eb26fefc","evidence-official-7cb8cb7f091536b92c11","evidence-official-bd6ec1f819e22f075c32"],"source_locator":"src/boltz/model/models/boltz2.py: Boltz2.__init__, forward, PairformerModule, AtomDiffusion and AffinityModule; docs/prediction.md: affinity outputs; Boltz-2 report Sections 2 Data, 3 Architecture, 4 Training and 6 Limitations"},"coverage":"limited","gaps":["Parameters: A complete parameter total is not stated in the reviewed Boltz-2 architecture report or model constructor; structure, confidence and affinity are separate modules."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}}}
{"id":"catalog-model-chai-1","kind":"model","name":"Chai-1","description":"Existing candidate catalogue entry; exact checkpoint and metadata require primary-source review.","status":"discovered","facets":{"areas":["proteins-complexes","molecular-interactions"],"method_types":["foundation model"]},"source_ids":["catalog-source-chai-1"],"links":[],"attributes":{"entity_level":"family","version":"released weights","reported_name":"Chai-1","access":"Public code and weights under Apache 2.0; substantial compute required.","method_type":"foundation model","historical_missing_metadata":{"checkpoint_revision":"not_yet_extracted","training_data":"not_yet_extracted","licence":"not_yet_extracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Chai-1 predicts the structures of biomolecular complexes containing proteins, nucleic acids and small molecules.","summary_source_ids":["evidence-official-affbe2d511ff2a80f457","evidence-official-b91fc1ec601eec6598c3"],"summary_source_locator":"Chai-1 Technical Report v1 (9 September 2024), Sections 2.1, 2.7–2.8 and 4.1–4.3; current repository README Licence","sections":[{"title":"How it works","body":"Chai-1 predicts the structures of biomolecular complexes containing proteins, nucleic acids and small molecules. An AlphaFold3-like structure architecture dominated by pair-biased self-attention, with additional protein-language-model embeddings and optional inter-chain constraint features. The documented inputs are FASTA sequences, ligand SMILES and optional alignments, templates, contacts or covalent-bond restraints. The output consists of sampled complex structures; the default command produces five predictions.","source_ids":["evidence-official-affbe2d511ff2a80f457","evidence-official-b91fc1ec601eec6598c3"],"source_locator":"Chai-1 Technical Report v1 (9 September 2024), Sections 2.1, 2.7–2.8 and 4.1–4.3; current repository README Licence"},{"title":"Versions and reproducibility","body":"Chai-1; README installation example pins chai_lab 0.6.1. The applicable input limits require configuration-specific checking.","source_ids":["evidence-official-affbe2d511ff2a80f457","evidence-official-b91fc1ec601eec6598c3"],"source_locator":"Chai-1 Technical Report v1 (9 September 2024), Sections 2.1, 2.7–2.8 and 4.1–4.3; current repository README Licence"}],"facts":[{"label":"Model type","value":"Multimodal biomolecular structure predictor","status":"source_checked","source_ids":["evidence-official-affbe2d511ff2a80f457","evidence-official-b91fc1ec601eec6598c3"],"source_locator":"Chai-1 Technical Report v1 (9 September 2024), Sections 2.1, 2.7–2.8 and 4.1–4.3; current repository README Licence"},{"label":"Architecture","value":"An AlphaFold3-like structure architecture dominated by pair-biased self-attention, with additional protein-language-model embeddings and optional inter-chain constraint features.","status":"source_checked","source_ids":["evidence-official-affbe2d511ff2a80f457","evidence-official-b91fc1ec601eec6598c3"],"source_locator":"Chai-1 Technical Report v1 (9 September 2024), Sections 2.1, 2.7–2.8 and 4.1–4.3; current repository README Licence"},{"label":"Inputs","value":"FASTA sequences, ligand SMILES and optional alignments, templates, contacts or covalent-bond restraints.","status":"source_checked","source_ids":["evidence-official-affbe2d511ff2a80f457","evidence-official-b91fc1ec601eec6598c3"],"source_locator":"Chai-1 Technical Report v1 (9 September 2024), Sections 2.1, 2.7–2.8 and 4.1–4.3; current repository README Licence"},{"label":"Outputs","value":"Sampled complex structures; the default command produces five predictions.","status":"source_checked","source_ids":["evidence-official-affbe2d511ff2a80f457","evidence-official-b91fc1ec601eec6598c3"],"source_locator":"Chai-1 Technical Report v1 (9 September 2024), Sections 2.1, 2.7–2.8 and 4.1–4.3; current repository README Licence"},{"label":"Parameters","value":"The September 2024 report specifies a 3B protein-language-model component but does not state a complete predictor total in the reviewed architecture sections.","status":"unreported","source_ids":["evidence-official-affbe2d511ff2a80f457","evidence-official-b91fc1ec601eec6598c3"],"source_locator":"Chai-1 Technical Report v1 (9 September 2024), Sections 2.1, 2.7–2.8 and 4.1–4.3; current repository README Licence"},{"label":"Known versions","value":"Chai-1; README installation example pins chai_lab 0.6.1.","status":"source_checked","source_ids":["evidence-official-affbe2d511ff2a80f457","evidence-official-b91fc1ec601eec6598c3"],"source_locator":"Chai-1 Technical Report v1 (9 September 2024), Sections 2.1, 2.7–2.8 and 4.1–4.3; current repository README Licence"},{"label":"Training data","value":"PDB structures and AlphaFoldDB distillation, with optional MSAs/templates. The technical report describes no other AlphaFold3 distillation datasets; server MSA search differs from the paper evaluation pipeline.","status":"source_checked","source_ids":["evidence-official-affbe2d511ff2a80f457","evidence-official-b91fc1ec601eec6598c3"],"source_locator":"Chai-1 Technical Report v1 (9 September 2024), Sections 2.1, 2.7–2.8 and 4.1–4.3; current repository README Licence"},{"label":"Training cutoff","value":"PDB structure and PDB70-template release cutoff: 2021-01-12, according to the September 2024 technical report.","status":"source_checked","source_ids":["evidence-official-affbe2d511ff2a80f457","evidence-official-b91fc1ec601eec6598c3"],"source_locator":"Chai-1 Technical Report v1 (9 September 2024), Sections 2.1, 2.7–2.8 and 4.1–4.3; current repository README Licence"},{"label":"Context limits","value":"The reviewed report and inference README do not specify one validated maximum for all protein, nucleic-acid and ligand inputs; resource requirements and molecular composition remain relevant.","status":"unreported","source_ids":["evidence-official-affbe2d511ff2a80f457","evidence-official-b91fc1ec601eec6598c3"],"source_locator":"Chai-1 Technical Report v1 (9 September 2024), Sections 2.1, 2.7–2.8 and 4.1–4.3; current repository README Licence"},{"label":"Weights licence","value":"Apache-2.0; README Licence explicitly covers code and weights.","status":"source_checked","source_ids":["evidence-official-affbe2d511ff2a80f457","evidence-official-b91fc1ec601eec6598c3"],"source_locator":"Chai-1 Technical Report v1 (9 September 2024), Sections 2.1, 2.7–2.8 and 4.1–4.3; current repository README Licence"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/chaidiscovery/chai-lab","status":"source_checked","source_ids":["evidence-official-affbe2d511ff2a80f457","evidence-official-b91fc1ec601eec6598c3"],"source_locator":"Chai-1 Technical Report v1 (9 September 2024), Sections 2.1, 2.7–2.8 and 4.1–4.3; current repository README Licence"},{"label":"Code licence","value":"Apache-2.0","status":"source_checked","source_ids":["evidence-official-290b60bb932abe9929a2"],"source_locator":"LICENSE: licence text"}],"strengths":[{"text":"Supports user-provided restraints and covalent bonds, alongside MSA and template inputs.","source_ids":["evidence-official-affbe2d511ff2a80f457","evidence-official-b91fc1ec601eec6598c3"],"source_locator":"Chai-1 Technical Report v1 (9 September 2024), Sections 2.1, 2.7–2.8 and 4.1–4.3; current repository README Licence"}],"limitations":[{"text":"The report documents failures of relative chain placement and sensitivity to modified residues. Inputs, MSA/template evidence and sampled structures must remain explicit when comparing runs.","source_ids":["evidence-official-affbe2d511ff2a80f457","evidence-official-b91fc1ec601eec6598c3"],"source_locator":"Chai-1 Technical Report v1 (9 September 2024), Sections 2.1, 2.7–2.8 and 4.1–4.3; current repository README Licence"}],"diagram":{"title":"Chai-1 workflow","steps":["Sequences and molecules","Optional MSA, template or restraints","Chai-1 inference","Complex structures"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-affbe2d511ff2a80f457","evidence-official-b91fc1ec601eec6598c3"],"source_locator":"Chai-1 Technical Report v1 (9 September 2024), Sections 2.1, 2.7–2.8 and 4.1–4.3; current repository README Licence"},"coverage":"limited","gaps":["Parameters: The September 2024 report specifies a 3B protein-language-model component but does not state a complete predictor total in the reviewed architecture sections.","Context limits: The reviewed report and inference README do not specify one validated maximum for all protein, nucleic-acid and ligand inputs; resource requirements and molecular composition remain relevant."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}}}
{"id":"catalog-model-diffdock-l","kind":"model","name":"DiffDock-L","description":"Existing candidate catalogue entry; exact checkpoint and metadata require primary-source review.","status":"discovered","facets":{"areas":["molecular-interactions"],"method_types":["specialist"]},"source_ids":["catalog-source-diffdock-l"],"links":[],"attributes":{"entity_level":"family","version":"2024 release","reported_name":"DiffDock-L","access":"Public pose-prediction code and weights; no native affinity prediction.","method_type":"specialist","historical_missing_metadata":{"checkpoint_revision":"not_yet_extracted","training_data":"not_yet_extracted","licence":"not_yet_extracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"DiffDock-L places small-molecule ligands in protein structures using a diffusion docking model.","summary_source_ids":["evidence-official-fa7a6c37148223466dee","evidence-official-fffd4026cb8f4018fe2e"],"summary_source_locator":"DiffDock-L paper Section 5.1 and Figure 3; official README: February 2024 update and License","sections":[{"title":"How it works","body":"DiffDock-L places small-molecule ligands in protein structures using a diffusion docking model. Diffusion-based molecular docking; the repository defaults to DiffDock-L rather than the original DiffDock model. The documented inputs are protein structure and a small-molecule ligand. The output consists of candidate ligand poses and confidence scores.","source_ids":["evidence-official-fa7a6c37148223466dee","evidence-official-fffd4026cb8f4018fe2e"],"source_locator":"DiffDock-L paper Section 5.1 and Figure 3; official README: February 2024 update and License"},{"title":"Versions and reproducibility","body":"DiffDock-L released February 2024; reproducing the original DiffDock requires its historical commit. The applicable input limits require configuration-specific checking.","source_ids":["evidence-official-fa7a6c37148223466dee","evidence-official-fffd4026cb8f4018fe2e"],"source_locator":"DiffDock-L paper Section 5.1 and Figure 3; official README: February 2024 update and License"}],"facts":[{"label":"Model type","value":"Diffusion-based molecular docking model","status":"source_checked","source_ids":["evidence-official-fa7a6c37148223466dee","evidence-official-fffd4026cb8f4018fe2e"],"source_locator":"DiffDock-L paper Section 5.1 and Figure 3; official README: February 2024 update and License"},{"label":"Architecture","value":"Diffusion-based molecular docking; the repository defaults to DiffDock-L rather than the original DiffDock model.","status":"source_checked","source_ids":["evidence-official-fa7a6c37148223466dee","evidence-official-fffd4026cb8f4018fe2e"],"source_locator":"DiffDock-L paper Section 5.1 and Figure 3; official README: February 2024 update and License"},{"label":"Inputs","value":"Protein structure and a small-molecule ligand.","status":"source_checked","source_ids":["evidence-official-fa7a6c37148223466dee","evidence-official-fffd4026cb8f4018fe2e"],"source_locator":"DiffDock-L paper Section 5.1 and Figure 3; official README: February 2024 update and License"},{"label":"Outputs","value":"Candidate ligand poses and confidence scores.","status":"source_checked","source_ids":["evidence-official-fa7a6c37148223466dee","evidence-official-fffd4026cb8f4018fe2e"],"source_locator":"DiffDock-L paper Section 5.1 and Figure 3; official README: February 2024 update and License"},{"label":"Parameters","value":"Approximately 30M in the larger score model; the confidence model is a separate component, so this is not a verified total for the complete pipeline.","status":"source_checked","source_ids":["evidence-official-fa7a6c37148223466dee","evidence-official-fffd4026cb8f4018fe2e"],"source_locator":"DiffDock-L paper Section 5.1 and Figure 3; official README: February 2024 update and License"},{"label":"Known versions","value":"DiffDock-L released February 2024; reproducing the original DiffDock requires its historical commit.","status":"source_checked","source_ids":["evidence-official-fa7a6c37148223466dee","evidence-official-fffd4026cb8f4018fe2e"],"source_locator":"DiffDock-L paper Section 5.1 and Figure 3; official README: February 2024 update and License"},{"label":"Training data","value":"PDBBind plus filtered pre-2019 Binding MOAD complexes from training/validation protein-domain clusters, with synthetic sidechain-as-ligand augmentation for additional pocket diversity.","status":"source_checked","source_ids":["evidence-official-fa7a6c37148223466dee","evidence-official-fffd4026cb8f4018fe2e"],"source_locator":"DiffDock-L paper Section 5.1 and Figure 3; official README: February 2024 update and License"},{"label":"Training cutoff","value":"The added Binding MOAD complexes were released before 2019; this is a component-specific restriction, not a universal cutoff for every input resource.","status":"source_checked","source_ids":["evidence-official-fa7a6c37148223466dee","evidence-official-fffd4026cb8f4018fe2e"],"source_locator":"DiffDock-L paper Section 5.1 and Figure 3; official README: February 2024 update and License"},{"label":"Context limits","value":"This is a protein–ligand graph model rather than a fixed text-token window. The reviewed paper and README do not establish one maximum for arbitrary receptor and ligand sizes.","status":"unreported","source_ids":["evidence-official-fa7a6c37148223466dee","evidence-official-fffd4026cb8f4018fe2e"],"source_locator":"DiffDock-L paper Section 5.1 and Figure 3; official README: February 2024 update and License"},{"label":"Weights licence","value":"MIT; README License explicitly includes code and model weights.","status":"source_checked","source_ids":["evidence-official-fa7a6c37148223466dee","evidence-official-fffd4026cb8f4018fe2e"],"source_locator":"DiffDock-L paper Section 5.1 and Figure 3; official README: February 2024 update and License"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/gcorso/DiffDock","status":"source_checked","source_ids":["evidence-official-fa7a6c37148223466dee","evidence-official-fffd4026cb8f4018fe2e"],"source_locator":"DiffDock-L paper Section 5.1 and Figure 3; official README: February 2024 update and License"},{"label":"Code licence","value":"MIT","status":"source_checked","source_ids":["evidence-official-af60e327735834e48fd7"],"source_locator":"LICENSE: licence text"}],"strengths":[{"text":"The repository provides evaluation splits and instructions for PDBBind, Binding MOAD and PoseBusters.","source_ids":["evidence-official-fa7a6c37148223466dee","evidence-official-fffd4026cb8f4018fe2e"],"source_locator":"DiffDock-L paper Section 5.1 and Figure 3; official README: February 2024 update and License"}],"limitations":[{"text":"The authors restrict the intended use to small-molecule docking. Large ligands, large protein complexes and unbound receptor conformations require particular caution.","source_ids":["evidence-official-fa7a6c37148223466dee","evidence-official-fffd4026cb8f4018fe2e"],"source_locator":"DiffDock-L paper Section 5.1 and Figure 3; official README: February 2024 update and License"}],"diagram":{"title":"DiffDock-L workflow","steps":["Protein and ligand","Diffusion pose sampling","Confidence scoring","Candidate docking poses"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-fa7a6c37148223466dee","evidence-official-fffd4026cb8f4018fe2e"],"source_locator":"DiffDock-L paper Section 5.1 and Figure 3; official README: February 2024 update and License"},"coverage":"limited","gaps":["Context limits: This is a protein–ligand graph model rather than a fixed text-token window. The reviewed paper and README do not establish one maximum for arbitrary receptor and ligand sizes."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}}}
{"id":"catalog-model-dnabert-2","kind":"model","name":"DNABERT-2","description":"Existing candidate catalogue entry; exact checkpoint and metadata require primary-source review.","status":"discovered","facets":{"areas":["dna-genomes"],"method_types":["foundation model"]},"source_ids":["catalog-source-dnabert-2"],"links":[{"relation":"variant_of","target_id":"discovery-model-dnabert-2"}],"attributes":{"entity_level":"family","version":"117M","reported_name":"DNABERT-2","access":"Public checkpoint; remote model code needs review before local use.","method_type":"foundation model","historical_missing_metadata":{"checkpoint_revision":"not_yet_extracted","training_data":"not_yet_extracted","licence":"not_yet_extracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"DNABERT-2 learns DNA representations that can be adapted to genomic prediction tasks.","summary_source_ids":["evidence-official-7267eabca7c4a7945878","evidence-official-2cf0a41fec83ee9c5cc9","evidence-official-96b5c3a31a50f7c259d6","evidence-official-a48ae27aa000bdab7442","evidence-official-3aa749d3205c0a768b22"],"summary_source_locator":"DNABERT-2 paper Sections 3.2, 4.1 and 5.2 (Further Pre-Training); official model card and implementation README; official DNABERT-2-117M checkpoint LICENSE at revision 7bce263b15377fc15361f52cfab88f8b586abda0","sections":[{"title":"How it works","body":"DNABERT-2 merges recurring DNA substrings into byte-pair tokens, then processes those tokens with a masked-language-model transformer. ALiBi supplies distance-dependent attention biases, while FlashAttention changes how attention is computed. The resulting contextual embeddings need an explicit pooling rule and prediction head for a downstream task.","source_ids":["evidence-official-7267eabca7c4a7945878","evidence-official-2cf0a41fec83ee9c5cc9","evidence-official-96b5c3a31a50f7c259d6","evidence-official-a48ae27aa000bdab7442","evidence-official-3aa749d3205c0a768b22"],"source_locator":"DNABERT-2 paper Sections 3.2, 4.1 and 5.2 (Further Pre-Training); official model card and implementation README; official DNABERT-2-117M checkpoint LICENSE at revision 7bce263b15377fc15361f52cfab88f8b586abda0"},{"title":"Versions and reproducibility","body":"DNABERT-2-117M model card and official DNABERT_2 implementation. ALiBi permits inference beyond the pretraining sequence length, subject to attention/memory cost; this does not establish unlimited biological context or validated accuracy at arbitrary lengths.","source_ids":["evidence-official-7267eabca7c4a7945878","evidence-official-2cf0a41fec83ee9c5cc9","evidence-official-96b5c3a31a50f7c259d6","evidence-official-a48ae27aa000bdab7442","evidence-official-3aa749d3205c0a768b22"],"source_locator":"DNABERT-2 paper Sections 3.2, 4.1 and 5.2 (Further Pre-Training); official model card and implementation README; official DNABERT-2-117M checkpoint LICENSE at revision 7bce263b15377fc15361f52cfab88f8b586abda0"}],"facts":[{"label":"Model type","value":"Masked-token DNA transformer encoder","status":"source_checked","source_ids":["evidence-official-7267eabca7c4a7945878","evidence-official-2cf0a41fec83ee9c5cc9","evidence-official-96b5c3a31a50f7c259d6","evidence-official-a48ae27aa000bdab7442","evidence-official-3aa749d3205c0a768b22"],"source_locator":"DNABERT-2 paper Sections 3.2, 4.1 and 5.2 (Further Pre-Training); official model card and implementation README; official DNABERT-2-117M checkpoint LICENSE at revision 7bce263b15377fc15361f52cfab88f8b586abda0"},{"label":"Architecture","value":"BERT-style DNA encoder with byte-pair tokenization, ALiBi relative attention biases and FlashAttention; task heads and pooling are separately configured.","status":"source_checked","source_ids":["evidence-official-7267eabca7c4a7945878","evidence-official-2cf0a41fec83ee9c5cc9","evidence-official-96b5c3a31a50f7c259d6","evidence-official-a48ae27aa000bdab7442","evidence-official-3aa749d3205c0a768b22"],"source_locator":"DNABERT-2 paper Sections 3.2, 4.1 and 5.2 (Further Pre-Training); official model card and implementation README; official DNABERT-2-117M checkpoint LICENSE at revision 7bce263b15377fc15361f52cfab88f8b586abda0"},{"label":"Inputs","value":"DNA sequence tokenized with the supplied tokenizer.","status":"source_checked","source_ids":["evidence-official-7267eabca7c4a7945878","evidence-official-2cf0a41fec83ee9c5cc9","evidence-official-96b5c3a31a50f7c259d6","evidence-official-a48ae27aa000bdab7442","evidence-official-3aa749d3205c0a768b22"],"source_locator":"DNABERT-2 paper Sections 3.2, 4.1 and 5.2 (Further Pre-Training); official model card and implementation README; official DNABERT-2-117M checkpoint LICENSE at revision 7bce263b15377fc15361f52cfab88f8b586abda0"},{"label":"Outputs","value":"Token representations and, after a specified adaptation, task predictions.","status":"source_checked","source_ids":["evidence-official-7267eabca7c4a7945878","evidence-official-2cf0a41fec83ee9c5cc9","evidence-official-96b5c3a31a50f7c259d6","evidence-official-a48ae27aa000bdab7442","evidence-official-3aa749d3205c0a768b22"],"source_locator":"DNABERT-2 paper Sections 3.2, 4.1 and 5.2 (Further Pre-Training); official model card and implementation README; official DNABERT-2-117M checkpoint LICENSE at revision 7bce263b15377fc15361f52cfab88f8b586abda0"},{"label":"Parameters","value":"117 million for DNABERT-2-117M; family names do not establish a particular checkpoint.","status":"source_checked","source_ids":["evidence-official-7267eabca7c4a7945878","evidence-official-2cf0a41fec83ee9c5cc9","evidence-official-96b5c3a31a50f7c259d6","evidence-official-a48ae27aa000bdab7442","evidence-official-3aa749d3205c0a768b22"],"source_locator":"DNABERT-2 paper Sections 3.2, 4.1 and 5.2 (Further Pre-Training); official model card and implementation README; official DNABERT-2-117M checkpoint LICENSE at revision 7bce263b15377fc15361f52cfab88f8b586abda0"},{"label":"Known versions","value":"DNABERT-2-117M model card and official DNABERT_2 implementation.","status":"source_checked","source_ids":["evidence-official-7267eabca7c4a7945878","evidence-official-2cf0a41fec83ee9c5cc9","evidence-official-96b5c3a31a50f7c259d6","evidence-official-a48ae27aa000bdab7442","evidence-official-3aa749d3205c0a768b22"],"source_locator":"DNABERT-2 paper Sections 3.2, 4.1 and 5.2 (Further Pre-Training); official model card and implementation README; official DNABERT-2-117M checkpoint LICENSE at revision 7bce263b15377fc15361f52cfab88f8b586abda0"},{"label":"Training data","value":"The paper describes a 32.49-billion-base corpus covering 135 species in six groups, alongside a 2.75-billion-base human corpus. Further GUE-domain pretraining is a separately reported model variant.","status":"source_checked","source_ids":["evidence-official-7267eabca7c4a7945878","evidence-official-2cf0a41fec83ee9c5cc9","evidence-official-96b5c3a31a50f7c259d6","evidence-official-a48ae27aa000bdab7442","evidence-official-3aa749d3205c0a768b22"],"source_locator":"DNABERT-2 paper Sections 3.2, 4.1 and 5.2 (Further Pre-Training); official model card and implementation README; official DNABERT-2-117M checkpoint LICENSE at revision 7bce263b15377fc15361f52cfab88f8b586abda0"},{"label":"Training cutoff","value":"The paper identifies the human and multispecies genome corpora but does not state one latest-sequence deposition date in its reviewed pretraining-data sections.","status":"unreported","source_ids":["evidence-official-7267eabca7c4a7945878","evidence-official-2cf0a41fec83ee9c5cc9","evidence-official-96b5c3a31a50f7c259d6","evidence-official-a48ae27aa000bdab7442","evidence-official-3aa749d3205c0a768b22"],"source_locator":"DNABERT-2 paper Sections 3.2, 4.1 and 5.2 (Further Pre-Training); official model card and implementation README; official DNABERT-2-117M checkpoint LICENSE at revision 7bce263b15377fc15361f52cfab88f8b586abda0"},{"label":"Context limits","value":"ALiBi permits inference beyond the pretraining sequence length, subject to attention/memory cost; this does not establish unlimited biological context or validated accuracy at arbitrary lengths.","status":"source_checked","source_ids":["evidence-official-7267eabca7c4a7945878","evidence-official-2cf0a41fec83ee9c5cc9","evidence-official-96b5c3a31a50f7c259d6","evidence-official-a48ae27aa000bdab7442","evidence-official-3aa749d3205c0a768b22"],"source_locator":"DNABERT-2 paper Sections 3.2, 4.1 and 5.2 (Further Pre-Training); official model card and implementation README; official DNABERT-2-117M checkpoint LICENSE at revision 7bce263b15377fc15361f52cfab88f8b586abda0"},{"label":"Weights licence","value":"The official zhihan1996/DNABERT-2-117M checkpoint repository carries Apache-2.0 in its pinned LICENSE. This does not assign terms to a separately fitted downstream predictor.","status":"source_checked","source_ids":["evidence-official-7267eabca7c4a7945878","evidence-official-2cf0a41fec83ee9c5cc9","evidence-official-96b5c3a31a50f7c259d6","evidence-official-a48ae27aa000bdab7442","evidence-official-3aa749d3205c0a768b22"],"source_locator":"DNABERT-2 paper Sections 3.2, 4.1 and 5.2 (Further Pre-Training); official model card and implementation README; official DNABERT-2-117M checkpoint LICENSE at revision 7bce263b15377fc15361f52cfab88f8b586abda0"},{"label":"Access","value":"Official downloadable model/card and usage examples: https://huggingface.co/zhihan1996/DNABERT-2-117M","status":"source_checked","source_ids":["evidence-official-7267eabca7c4a7945878","evidence-official-2cf0a41fec83ee9c5cc9","evidence-official-96b5c3a31a50f7c259d6","evidence-official-a48ae27aa000bdab7442","evidence-official-3aa749d3205c0a768b22"],"source_locator":"DNABERT-2 paper Sections 3.2, 4.1 and 5.2 (Further Pre-Training); official model card and implementation README; official DNABERT-2-117M checkpoint LICENSE at revision 7bce263b15377fc15361f52cfab88f8b586abda0"},{"label":"Code licence","value":"Apache-2.0","status":"source_checked","source_ids":["evidence-official-25b222d11900e0e88a51"],"source_locator":"LICENSE: licence text"}],"strengths":[{"text":"The released model supports embedding extraction and task-specific fine-tuning.","source_ids":["evidence-official-7267eabca7c4a7945878","evidence-official-2cf0a41fec83ee9c5cc9","evidence-official-96b5c3a31a50f7c259d6","evidence-official-a48ae27aa000bdab7442","evidence-official-3aa749d3205c0a768b22"],"source_locator":"DNABERT-2 paper Sections 3.2, 4.1 and 5.2 (Further Pre-Training); official model card and implementation README; official DNABERT-2-117M checkpoint LICENSE at revision 7bce263b15377fc15361f52cfab88f8b586abda0"}],"limitations":[{"text":"An embedding model alone is not the same evaluated pipeline as frozen embeddings followed by logistic regression. Tokenization and pooling choices must be preserved.","source_ids":["evidence-official-7267eabca7c4a7945878","evidence-official-2cf0a41fec83ee9c5cc9","evidence-official-96b5c3a31a50f7c259d6","evidence-official-a48ae27aa000bdab7442","evidence-official-3aa749d3205c0a768b22"],"source_locator":"DNABERT-2 paper Sections 3.2, 4.1 and 5.2 (Further Pre-Training); official model card and implementation README; official DNABERT-2-117M checkpoint LICENSE at revision 7bce263b15377fc15361f52cfab88f8b586abda0"}],"diagram":{"title":"DNABERT-2 workflow","steps":["DNA sequence","BPE tokens","Transformer encoder","Representations","Specified task head"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-7267eabca7c4a7945878","evidence-official-2cf0a41fec83ee9c5cc9","evidence-official-96b5c3a31a50f7c259d6","evidence-official-a48ae27aa000bdab7442","evidence-official-3aa749d3205c0a768b22"],"source_locator":"DNABERT-2 paper Sections 3.2, 4.1 and 5.2 (Further Pre-Training); official model card and implementation README; official DNABERT-2-117M checkpoint LICENSE at revision 7bce263b15377fc15361f52cfab88f8b586abda0"},"coverage":"limited","gaps":["Training cutoff: The paper identifies the human and multispecies genome corpora but does not state one latest-sequence deposition date in its reviewed pretraining-data sections."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}}}
{"id":"catalog-model-esm-2","kind":"model","name":"ESM-2","description":"Existing candidate catalogue entry; exact checkpoint and metadata require primary-source review.","status":"discovered","facets":{"areas":["proteins-complexes"],"method_types":["foundation model"]},"source_ids":["catalog-source-esm-2"],"links":[{"relation":"variant_of","target_id":"discovery-model-esm-2"}],"attributes":{"entity_level":"family","version":"8M","reported_name":"ESM-2","access":"Public checkpoint; small 8M variant suits a local pilot.","method_type":"foundation model","historical_missing_metadata":{"checkpoint_revision":"not_yet_extracted","training_data":"not_yet_extracted","licence":"not_yet_extracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"ESM-2 is a family of protein sequence encoders that produce representations for downstream protein analyses.","summary_source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-d7e5c63feb0e0e627625","evidence-official-28fc17fbe0c1d99da219"],"summary_source_locator":"ESM README: Pre-trained Models and Main models; official facebook/esm2_t33_650M_UR50D README licence metadata and config.json","sections":[{"title":"How it works","body":"ESM-2 tokenizes an amino-acid sequence and uses a transformer encoder trained to recover masked residues. Self-attention lets each residue representation depend on its sequence context. The released model returns token probabilities and embeddings; a specified pooling rule, task head or complete folding pipeline is needed for a particular biological prediction.","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-d7e5c63feb0e0e627625","evidence-official-28fc17fbe0c1d99da219"],"source_locator":"ESM README: Pre-trained Models and Main models; official facebook/esm2_t33_650M_UR50D README licence metadata and config.json"},{"title":"Versions and reproducibility","body":"ESM-2 checkpoint identifiers encode layer count, parameter scale and training-data tag. The checked esm2_t33_650M_UR50D configuration lists max_position_embeddings=1,026. This configuration field includes model positions and is not a claim of training or validated inference on 1,026 amino acids.","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-d7e5c63feb0e0e627625","evidence-official-28fc17fbe0c1d99da219"],"source_locator":"ESM README: Pre-trained Models and Main models; official facebook/esm2_t33_650M_UR50D README licence metadata and config.json"}],"facts":[{"label":"Model type","value":"Masked-token protein transformer encoder","status":"source_checked","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-d7e5c63feb0e0e627625","evidence-official-28fc17fbe0c1d99da219"],"source_locator":"ESM README: Pre-trained Models and Main models; official facebook/esm2_t33_650M_UR50D README licence metadata and config.json"},{"label":"Architecture","value":"Masked-token protein transformer encoder; the checked 650M checkpoint has 33 layers, hidden width 1,280, 20 attention heads and rotary positional encoding.","status":"source_checked","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-d7e5c63feb0e0e627625","evidence-official-28fc17fbe0c1d99da219"],"source_locator":"ESM README: Pre-trained Models and Main models; official facebook/esm2_t33_650M_UR50D README licence metadata and config.json"},{"label":"Inputs","value":"Single amino-acid sequences.","status":"source_checked","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-d7e5c63feb0e0e627625","evidence-official-28fc17fbe0c1d99da219"],"source_locator":"ESM README: Pre-trained Models and Main models; official facebook/esm2_t33_650M_UR50D README licence metadata and config.json"},{"label":"Outputs","value":"Residue embeddings, sequence representations and masked-token predictions.","status":"source_checked","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-d7e5c63feb0e0e627625","evidence-official-28fc17fbe0c1d99da219"],"source_locator":"ESM README: Pre-trained Models and Main models; official facebook/esm2_t33_650M_UR50D README licence metadata and config.json"},{"label":"Parameters","value":"Released scales: 8M, 35M, 150M, 650M, 3B and 15B.","status":"source_checked","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-d7e5c63feb0e0e627625","evidence-official-28fc17fbe0c1d99da219"],"source_locator":"ESM README: Pre-trained Models and Main models; official facebook/esm2_t33_650M_UR50D README licence metadata and config.json"},{"label":"Known versions","value":"ESM-2 checkpoint identifiers encode layer count, parameter scale and training-data tag.","status":"source_checked","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-d7e5c63feb0e0e627625","evidence-official-28fc17fbe0c1d99da219"],"source_locator":"ESM README: Pre-trained Models and Main models; official facebook/esm2_t33_650M_UR50D README licence metadata and config.json"},{"label":"Training data","value":"UniRef50 clusters with UniRef90 sampling; the pretrained-model table labels UR50/D 2021_04.","status":"source_checked","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-d7e5c63feb0e0e627625","evidence-official-28fc17fbe0c1d99da219"],"source_locator":"ESM README: Pre-trained Models and Main models; official facebook/esm2_t33_650M_UR50D README licence metadata and config.json"},{"label":"Training cutoff","value":"The pretrained-model table identifies training-data release UR50/D 2021_04; a corpus release date is not necessarily a last-deposited-sequence cutoff.","status":"source_checked","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-d7e5c63feb0e0e627625","evidence-official-28fc17fbe0c1d99da219"],"source_locator":"ESM README: Pre-trained Models and Main models; official facebook/esm2_t33_650M_UR50D README licence metadata and config.json"},{"label":"Context limits","value":"The checked esm2_t33_650M_UR50D configuration lists max_position_embeddings=1,026. This configuration field includes model positions and is not a claim of training or validated inference on 1,026 amino acids.","status":"source_checked","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-d7e5c63feb0e0e627625","evidence-official-28fc17fbe0c1d99da219"],"source_locator":"ESM README: Pre-trained Models and Main models; official facebook/esm2_t33_650M_UR50D README licence metadata and config.json"},{"label":"Weights licence","value":"The official facebook/esm2_t33_650M_UR50D model card declares MIT; this is the inspected checkpoint, not a licence inference from source code.","status":"source_checked","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-d7e5c63feb0e0e627625","evidence-official-28fc17fbe0c1d99da219"],"source_locator":"ESM README: Pre-trained Models and Main models; official facebook/esm2_t33_650M_UR50D README licence metadata and config.json"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/facebookresearch/esm","status":"source_checked","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-d7e5c63feb0e0e627625","evidence-official-28fc17fbe0c1d99da219"],"source_locator":"ESM README: Pre-trained Models and Main models; official facebook/esm2_t33_650M_UR50D README licence metadata and config.json"},{"label":"Code licence","value":"MIT","status":"source_checked","source_ids":["evidence-official-2e7c7649620407f50f6b"],"source_locator":"LICENSE: licence text"}],"strengths":[{"text":"Released checkpoints span several sizes and can be used without constructing a multiple sequence alignment.","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-d7e5c63feb0e0e627625","evidence-official-28fc17fbe0c1d99da219"],"source_locator":"ESM README: Pre-trained Models and Main models; official facebook/esm2_t33_650M_UR50D README licence metadata and config.json"}],"limitations":[{"text":"A general embedding is not a directly measured function or structure. Fine-tuning, pooling and downstream heads remain part of each evaluated configuration.","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-d7e5c63feb0e0e627625","evidence-official-28fc17fbe0c1d99da219"],"source_locator":"ESM README: Pre-trained Models and Main models; official facebook/esm2_t33_650M_UR50D README licence metadata and config.json"}],"diagram":{"title":"ESM-2 workflow","steps":["Protein sequence","Transformer layers","Residue embeddings","Specified downstream analysis"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-d7e5c63feb0e0e627625","evidence-official-28fc17fbe0c1d99da219"],"source_locator":"ESM README: Pre-trained Models and Main models; official facebook/esm2_t33_650M_UR50D README licence metadata and config.json"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}}}
{"id":"catalog-model-esmfold","kind":"model","name":"ESMFold","description":"Existing candidate catalogue entry; exact checkpoint and metadata require primary-source review.","status":"discovered","facets":{"areas":["proteins-complexes"],"method_types":["foundation model"]},"source_ids":["catalog-source-esmfold"],"links":[{"relation":"variant_of","target_id":"discovery-model-esmfold"}],"attributes":{"entity_level":"family","version":"v1","reported_name":"ESMFold","access":"Public checkpoint; materially larger than ESM-2 8M.","method_type":"foundation model","historical_missing_metadata":{"checkpoint_revision":"not_yet_extracted","training_data":"not_yet_extracted","licence":"not_yet_extracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"ESMFold predicts protein structures directly from amino-acid sequence using ESM-2 representations.","summary_source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-fa53ab621979fb5f37c8","evidence-official-867f2d1dcb4084248daa"],"summary_source_locator":"ESM README: ESMFold Structure Prediction; official facebook/esmfold_v1 model card and config.json esmfold_config","sections":[{"title":"How it works","body":"ESMFold predicts protein structures directly from amino-acid sequence using ESM-2 representations. ESM-2 sequence representations feed a folding trunk and structure module. The checked v1 configuration has 48 trunk blocks, eight structure-module blocks and up to four recycles. The documented inputs are protein amino-acid sequence; the ESMFold interface also accepts chains separated by a colon. The output consists of predicted PDB structure and confidence values.","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-fa53ab621979fb5f37c8","evidence-official-867f2d1dcb4084248daa"],"source_locator":"ESM README: ESMFold Structure Prediction; official facebook/esmfold_v1 model card and config.json esmfold_config"},{"title":"Versions and reproducibility","body":"esmfold_v0 and esmfold_v1; v1 is the repository recommendation. Inference length is constrained by memory; the repository documents chunking and CPU offload. Backbone position settings do not alone establish the full folding pipeline limit.","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-fa53ab621979fb5f37c8","evidence-official-867f2d1dcb4084248daa"],"source_locator":"ESM README: ESMFold Structure Prediction; official facebook/esmfold_v1 model card and config.json esmfold_config"}],"facts":[{"label":"Model type","value":"Sequence-to-structure protein prediction pipeline","status":"source_checked","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-fa53ab621979fb5f37c8","evidence-official-867f2d1dcb4084248daa"],"source_locator":"ESM README: ESMFold Structure Prediction; official facebook/esmfold_v1 model card and config.json esmfold_config"},{"label":"Architecture","value":"ESM-2 sequence representations feed a folding trunk and structure module. The checked v1 configuration has 48 trunk blocks, eight structure-module blocks and up to four recycles.","status":"source_checked","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-fa53ab621979fb5f37c8","evidence-official-867f2d1dcb4084248daa"],"source_locator":"ESM README: ESMFold Structure Prediction; official facebook/esmfold_v1 model card and config.json esmfold_config"},{"label":"Inputs","value":"Protein amino-acid sequence; the ESMFold interface also accepts chains separated by a colon.","status":"source_checked","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-fa53ab621979fb5f37c8","evidence-official-867f2d1dcb4084248daa"],"source_locator":"ESM README: ESMFold Structure Prediction; official facebook/esmfold_v1 model card and config.json esmfold_config"},{"label":"Outputs","value":"Predicted PDB structure and confidence values.","status":"source_checked","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-fa53ab621979fb5f37c8","evidence-official-867f2d1dcb4084248daa"],"source_locator":"ESM README: ESMFold Structure Prediction; official facebook/esmfold_v1 model card and config.json esmfold_config"},{"label":"Parameters","value":"The released v1 configuration identifies an ESM-2 3B backbone plus a folding trunk and structure module. The 3B figure is not the total size of the complete predictor.","status":"source_checked","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-fa53ab621979fb5f37c8","evidence-official-867f2d1dcb4084248daa"],"source_locator":"ESM README: ESMFold Structure Prediction; official facebook/esmfold_v1 model card and config.json esmfold_config"},{"label":"Known versions","value":"esmfold_v0 and esmfold_v1; v1 is the repository recommendation.","status":"source_checked","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-fa53ab621979fb5f37c8","evidence-official-867f2d1dcb4084248daa"],"source_locator":"ESM README: ESMFold Structure Prediction; official facebook/esmfold_v1 model card and config.json esmfold_config"},{"label":"Training data","value":"PDB and UniRef50 are listed for ESMFold. Full structural training-cutoff verification remains outstanding.","status":"source_checked","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-fa53ab621979fb5f37c8","evidence-official-867f2d1dcb4084248daa"],"source_locator":"ESM README: ESMFold Structure Prediction; official facebook/esmfold_v1 model card and config.json esmfold_config"},{"label":"Training cutoff","value":"PDB and UniRef50 are identified in the official model table; the inspected ESMFold-v1 card and configuration do not supply a shared latest-data date for both components.","status":"unreported","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-fa53ab621979fb5f37c8","evidence-official-867f2d1dcb4084248daa"],"source_locator":"ESM README: ESMFold Structure Prediction; official facebook/esmfold_v1 model card and config.json esmfold_config"},{"label":"Context limits","value":"Inference length is constrained by memory; the repository documents chunking and CPU offload. Backbone position settings do not alone establish the full folding pipeline limit.","status":"source_checked","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-fa53ab621979fb5f37c8","evidence-official-867f2d1dcb4084248daa"],"source_locator":"ESM README: ESMFold Structure Prediction; official facebook/esmfold_v1 model card and config.json esmfold_config"},{"label":"Weights licence","value":"MIT declared by the official facebook/esmfold_v1 model card.","status":"source_checked","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-fa53ab621979fb5f37c8","evidence-official-867f2d1dcb4084248daa"],"source_locator":"ESM README: ESMFold Structure Prediction; official facebook/esmfold_v1 model card and config.json esmfold_config"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/facebookresearch/esm","status":"source_checked","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-fa53ab621979fb5f37c8","evidence-official-867f2d1dcb4084248daa"],"source_locator":"ESM README: ESMFold Structure Prediction; official facebook/esmfold_v1 model card and config.json esmfold_config"},{"label":"Code licence","value":"MIT","status":"source_checked","source_ids":["evidence-official-2e7c7649620407f50f6b"],"source_locator":"LICENSE: licence text"}],"strengths":[{"text":"Supports sequence-only structure prediction without an MSA search.","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-fa53ab621979fb5f37c8","evidence-official-867f2d1dcb4084248daa"],"source_locator":"ESM README: ESMFold Structure Prediction; official facebook/esmfold_v1 model card and config.json esmfold_config"}],"limitations":[{"text":"ESMFold v0 and v1 are different releases. The repository discourages using structure-module-only ablation models as the standard predictor.","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-fa53ab621979fb5f37c8","evidence-official-867f2d1dcb4084248daa"],"source_locator":"ESM README: ESMFold Structure Prediction; official facebook/esmfold_v1 model card and config.json esmfold_config"}],"diagram":{"title":"ESMFold workflow","steps":["Protein sequence","ESM-2 representations","Folding module","Predicted structure and confidence"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-fa53ab621979fb5f37c8","evidence-official-867f2d1dcb4084248daa"],"source_locator":"ESM README: ESMFold Structure Prediction; official facebook/esmfold_v1 model card and config.json esmfold_config"},"coverage":"limited","gaps":["Training cutoff: PDB and UniRef50 are identified in the official model table; the inspected ESMFold-v1 card and configuration do not supply a shared latest-data date for both components."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}}}
{"id":"catalog-model-evo-2","kind":"model","name":"Evo 2","description":"Existing candidate catalogue entry; exact checkpoint and metadata require primary-source review.","status":"discovered","facets":{"areas":["dna-genomes","microbes-communities"],"method_types":["foundation model"]},"source_ids":["catalog-source-evo-2"],"links":[],"attributes":{"entity_level":"family","version":"7B","reported_name":"Evo 2","access":"Public checkpoints; official local inference needs CUDA hardware and substantial memory.","method_type":"foundation model","historical_missing_metadata":{"checkpoint_revision":"not_yet_extracted","training_data":"not_yet_extracted","licence":"not_yet_extracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Evo 2 models and generates DNA over long contexts at single-nucleotide resolution.","summary_source_ids":["evidence-official-a248990b44d481814f47","evidence-official-849b751ed4bf88622749","evidence-official-b8df803e8f55324a4eb5","evidence-official-8717ffb993cc8f7eddec"],"summary_source_locator":"Evo 2 paper: architecture, training, and data, Figure 1 and Discussion; official checkpoints table and evo2_7b licence metadata","sections":[{"title":"How it works","body":"Evo 2 models and generates DNA over long contexts at single-nucleotide resolution. StripedHyena 2 hybrid architecture combining short, medium and long convolution operators with attention, trained autoregressively at single-base resolution. The documented inputs are DNA sequences represented at single-base resolution. The output consists of next-token outputs, embeddings and generated DNA sequences.","source_ids":["evidence-official-a248990b44d481814f47","evidence-official-849b751ed4bf88622749","evidence-official-b8df803e8f55324a4eb5","evidence-official-8717ffb993cc8f7eddec"],"source_locator":"Evo 2 paper: architecture, training, and data, Figure 1 and Discussion; official checkpoints table and evo2_7b licence metadata"},{"title":"Versions and reproducibility","body":"Base 8K models, long-context 1M models, 7B 262K model and separately fine-tuned Microviridae model. Checkpoint-dependent: 8K, 262K or 1M bases as listed in the Checkpoints table.","source_ids":["evidence-official-a248990b44d481814f47","evidence-official-849b751ed4bf88622749","evidence-official-b8df803e8f55324a4eb5","evidence-official-8717ffb993cc8f7eddec"],"source_locator":"Evo 2 paper: architecture, training, and data, Figure 1 and Discussion; official checkpoints table and evo2_7b licence metadata"}],"facts":[{"label":"Model type","value":"Autoregressive DNA model with StripedHyena 2","status":"source_checked","source_ids":["evidence-official-a248990b44d481814f47","evidence-official-849b751ed4bf88622749","evidence-official-b8df803e8f55324a4eb5","evidence-official-8717ffb993cc8f7eddec"],"source_locator":"Evo 2 paper: architecture, training, and data, Figure 1 and Discussion; official checkpoints table and evo2_7b licence metadata"},{"label":"Architecture","value":"StripedHyena 2 hybrid architecture combining short, medium and long convolution operators with attention, trained autoregressively at single-base resolution.","status":"source_checked","source_ids":["evidence-official-a248990b44d481814f47","evidence-official-849b751ed4bf88622749","evidence-official-b8df803e8f55324a4eb5","evidence-official-8717ffb993cc8f7eddec"],"source_locator":"Evo 2 paper: architecture, training, and data, Figure 1 and Discussion; official checkpoints table and evo2_7b licence metadata"},{"label":"Inputs","value":"DNA sequences represented at single-base resolution.","status":"source_checked","source_ids":["evidence-official-a248990b44d481814f47","evidence-official-849b751ed4bf88622749","evidence-official-b8df803e8f55324a4eb5","evidence-official-8717ffb993cc8f7eddec"],"source_locator":"Evo 2 paper: architecture, training, and data, Figure 1 and Discussion; official checkpoints table and evo2_7b licence metadata"},{"label":"Outputs","value":"Next-token outputs, embeddings and generated DNA sequences.","status":"source_checked","source_ids":["evidence-official-a248990b44d481814f47","evidence-official-849b751ed4bf88622749","evidence-official-b8df803e8f55324a4eb5","evidence-official-8717ffb993cc8f7eddec"],"source_locator":"Evo 2 paper: architecture, training, and data, Figure 1 and Discussion; official checkpoints table and evo2_7b licence metadata"},{"label":"Parameters","value":"1B, 7B, 20B and 40B checkpoints are listed.","status":"source_checked","source_ids":["evidence-official-a248990b44d481814f47","evidence-official-849b751ed4bf88622749","evidence-official-b8df803e8f55324a4eb5","evidence-official-8717ffb993cc8f7eddec"],"source_locator":"Evo 2 paper: architecture, training, and data, Figure 1 and Discussion; official checkpoints table and evo2_7b licence metadata"},{"label":"Known versions","value":"Base 8K models, long-context 1M models, 7B 262K model and separately fine-tuned Microviridae model.","status":"source_checked","source_ids":["evidence-official-a248990b44d481814f47","evidence-official-849b751ed4bf88622749","evidence-official-b8df803e8f55324a4eb5","evidence-official-8717ffb993cc8f7eddec"],"source_locator":"Evo 2 paper: architecture, training, and data, Figure 1 and Discussion; official checkpoints table and evo2_7b licence metadata"},{"label":"Training data","value":"OpenGenome2 contains more than 8.8T curated nucleotides across bacteria, archaea, eukaryotes and bacteriophage. The paper separates 2.4T tokens of training exposure for 7B from 9.3T for 40B; eukaryotic-host viral sequences were excluded.","status":"source_checked","source_ids":["evidence-official-a248990b44d481814f47","evidence-official-849b751ed4bf88622749","evidence-official-b8df803e8f55324a4eb5","evidence-official-8717ffb993cc8f7eddec"],"source_locator":"Evo 2 paper: architecture, training, and data, Figure 1 and Discussion; official checkpoints table and evo2_7b licence metadata"},{"label":"Training cutoff","value":"OpenGenome2 combines multiple nucleotide collections. The inspected paper and released checkpoint documentation do not provide one latest-deposition date that covers every component.","status":"unreported","source_ids":["evidence-official-a248990b44d481814f47","evidence-official-849b751ed4bf88622749","evidence-official-b8df803e8f55324a4eb5","evidence-official-8717ffb993cc8f7eddec"],"source_locator":"Evo 2 paper: architecture, training, and data, Figure 1 and Discussion; official checkpoints table and evo2_7b licence metadata"},{"label":"Context limits","value":"Checkpoint-dependent: 8K, 262K or 1M bases as listed in the Checkpoints table.","status":"source_checked","source_ids":["evidence-official-a248990b44d481814f47","evidence-official-849b751ed4bf88622749","evidence-official-b8df803e8f55324a4eb5","evidence-official-8717ffb993cc8f7eddec"],"source_locator":"Evo 2 paper: architecture, training, and data, Figure 1 and Discussion; official checkpoints table and evo2_7b licence metadata"},{"label":"Weights licence","value":"Apache-2.0 is declared in the inspected ArcInstitute/evo2_7b model card; other checkpoints require their own pinned terms.","status":"source_checked","source_ids":["evidence-official-a248990b44d481814f47","evidence-official-849b751ed4bf88622749","evidence-official-b8df803e8f55324a4eb5","evidence-official-8717ffb993cc8f7eddec"],"source_locator":"Evo 2 paper: architecture, training, and data, Figure 1 and Discussion; official checkpoints table and evo2_7b licence metadata"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/ArcInstitute/evo2","status":"source_checked","source_ids":["evidence-official-a248990b44d481814f47","evidence-official-849b751ed4bf88622749","evidence-official-b8df803e8f55324a4eb5","evidence-official-8717ffb993cc8f7eddec"],"source_locator":"Evo 2 paper: architecture, training, and data, Figure 1 and Discussion; official checkpoints table and evo2_7b licence metadata"},{"label":"Code licence","value":"Apache-2.0","status":"source_checked","source_ids":["evidence-official-67ee1cc31060ba8c9569"],"source_locator":"LICENSE: licence text"}],"strengths":[{"text":"Different released context lengths and model scales support a range of sequence modeling workflows.","source_ids":["evidence-official-a248990b44d481814f47","evidence-official-849b751ed4bf88622749","evidence-official-b8df803e8f55324a4eb5","evidence-official-8717ffb993cc8f7eddec"],"source_locator":"Evo 2 paper: architecture, training, and data, Figure 1 and Discussion; official checkpoints table and evo2_7b licence metadata"}],"limitations":[{"text":"Hardware requirements differ by checkpoint: the README requires FP8/Transformer Engine and Hopper GPUs for some scales, while 7B supports bfloat 16 on a wider set of GPUs.","source_ids":["evidence-official-a248990b44d481814f47","evidence-official-849b751ed4bf88622749","evidence-official-b8df803e8f55324a4eb5","evidence-official-8717ffb993cc8f7eddec"],"source_locator":"Evo 2 paper: architecture, training, and data, Figure 1 and Discussion; official checkpoints table and evo2_7b licence metadata"}],"diagram":{"title":"Evo 2 workflow","steps":["DNA bases","StripedHyena 2","Autoregressive outputs","Sequence scoring or generation"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-a248990b44d481814f47","evidence-official-849b751ed4bf88622749","evidence-official-b8df803e8f55324a4eb5","evidence-official-8717ffb993cc8f7eddec"],"source_locator":"Evo 2 paper: architecture, training, and data, Figure 1 and Discussion; official checkpoints table and evo2_7b licence metadata"},"coverage":"limited","gaps":["Training cutoff: OpenGenome2 combines multiple nucleotide collections. The inspected paper and released checkpoint documentation do not provide one latest-deposition date that covers every component."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}}}
{"id":"catalog-model-gears","kind":"model","name":"GEARS","description":"Existing candidate catalogue entry; exact checkpoint and metadata require primary-source review.","status":"discovered","facets":{"areas":["cells-tissues"],"method_types":["specialist"]},"source_ids":["catalog-source-gears"],"links":[{"relation":"family","target_id":"discovery-model-gears"}],"attributes":{"entity_level":"family","version":"published implementation","reported_name":"GEARS","access":"Public code; task-specific training data required.","method_type":"specialist","historical_missing_metadata":{"checkpoint_revision":"not_yet_extracted","training_data":"not_yet_extracted","licence":"not_yet_extracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"GEARS predicts transcriptional responses to single- and multi-gene perturbations from single-cell perturbation screens.","summary_source_ids":["evidence-official-c037e3419c04936262a0","evidence-official-b5e81b2c497214b18f8a","evidence-official-f464cd407a6a9d031720"],"summary_source_locator":"GEARS paper Methods: Overview, Gene coexpression graph encoder, GO graph, Cross-gene effects and gene-specific decoder; gears/model.py: GEARS_Model","sections":[{"title":"How it works","body":"GEARS predicts transcriptional responses to single- and multi-gene perturbations from single-cell perturbation screens. Two graph encoders represent gene coexpression and Gene Ontology perturbation similarity. Perturbation embeddings are composed with gene embeddings, then a cross-gene network and gene-specific decoders predict expression changes. The documented inputs are single-cell expression data, perturbation labels and the graph resources used by the configured model. The output consists of predicted gene-expression responses and genetic-interaction analyses.","source_ids":["evidence-official-c037e3419c04936262a0","evidence-official-b5e81b2c497214b18f8a","evidence-official-f464cd407a6a9d031720"],"source_locator":"GEARS paper Methods: Overview, Gene coexpression graph encoder, GO graph, Cross-gene effects and gene-specific decoder; gears/model.py: GEARS_Model"},{"title":"Versions and reproducibility","body":"The README describes v0.1.1 updates; a specific trained checkpoint must be recorded separately. A gene-expression vector and perturbation set over the configured gene inventory; no fixed nucleotide or amino-acid token window.","source_ids":["evidence-official-c037e3419c04936262a0","evidence-official-b5e81b2c497214b18f8a","evidence-official-f464cd407a6a9d031720"],"source_locator":"GEARS paper Methods: Overview, Gene coexpression graph encoder, GO graph, Cross-gene effects and gene-specific decoder; gears/model.py: GEARS_Model"}],"facts":[{"label":"Model type","value":"Graph-based perturbation-response predictor","status":"source_checked","source_ids":["evidence-official-c037e3419c04936262a0","evidence-official-b5e81b2c497214b18f8a","evidence-official-f464cd407a6a9d031720"],"source_locator":"GEARS paper Methods: Overview, Gene coexpression graph encoder, GO graph, Cross-gene effects and gene-specific decoder; gears/model.py: GEARS_Model"},{"label":"Architecture","value":"Two graph encoders represent gene coexpression and Gene Ontology perturbation similarity. Perturbation embeddings are composed with gene embeddings, then a cross-gene network and gene-specific decoders predict expression changes.","status":"source_checked","source_ids":["evidence-official-c037e3419c04936262a0","evidence-official-b5e81b2c497214b18f8a","evidence-official-f464cd407a6a9d031720"],"source_locator":"GEARS paper Methods: Overview, Gene coexpression graph encoder, GO graph, Cross-gene effects and gene-specific decoder; gears/model.py: GEARS_Model"},{"label":"Inputs","value":"Single-cell expression data, perturbation labels and the graph resources used by the configured model.","status":"source_checked","source_ids":["evidence-official-c037e3419c04936262a0","evidence-official-b5e81b2c497214b18f8a","evidence-official-f464cd407a6a9d031720"],"source_locator":"GEARS paper Methods: Overview, Gene coexpression graph encoder, GO graph, Cross-gene effects and gene-specific decoder; gears/model.py: GEARS_Model"},{"label":"Outputs","value":"Predicted gene-expression responses and genetic-interaction analyses.","status":"source_checked","source_ids":["evidence-official-c037e3419c04936262a0","evidence-official-b5e81b2c497214b18f8a","evidence-official-f464cd407a6a9d031720"],"source_locator":"GEARS paper Methods: Overview, Gene coexpression graph encoder, GO graph, Cross-gene effects and gene-specific decoder; gears/model.py: GEARS_Model"},{"label":"Parameters","value":"Configuration-dependent: gene and perturbation embedding tables grow with the selected gene/perturbation inventory, alongside graph and decoder parameters.","status":"source_checked","source_ids":["evidence-official-c037e3419c04936262a0","evidence-official-b5e81b2c497214b18f8a","evidence-official-f464cd407a6a9d031720"],"source_locator":"GEARS paper Methods: Overview, Gene coexpression graph encoder, GO graph, Cross-gene effects and gene-specific decoder; gears/model.py: GEARS_Model"},{"label":"Known versions","value":"The README describes v0.1.1 updates; a specific trained checkpoint must be recorded separately.","status":"source_checked","source_ids":["evidence-official-c037e3419c04936262a0","evidence-official-b5e81b2c497214b18f8a","evidence-official-f464cd407a6a9d031720"],"source_locator":"GEARS paper Methods: Overview, Gene coexpression graph encoder, GO graph, Cross-gene effects and gene-specific decoder; gears/model.py: GEARS_Model"},{"label":"Training data","value":"Fitted to the selected perturbation screen. Examples include Norman, Adamson and Dixit; the repository also lists Replogle RPE1/K562 loaders.","status":"source_checked","source_ids":["evidence-official-c037e3419c04936262a0","evidence-official-b5e81b2c497214b18f8a","evidence-official-f464cd407a6a9d031720"],"source_locator":"GEARS paper Methods: Overview, Gene coexpression graph encoder, GO graph, Cross-gene effects and gene-specific decoder; gears/model.py: GEARS_Model"},{"label":"Training cutoff","value":"Inapplicable as a universal pretrained-model cutoff: GEARS fits the provided perturbation training set and builds its coexpression graph from that set.","status":"inapplicable","source_ids":["evidence-official-c037e3419c04936262a0","evidence-official-b5e81b2c497214b18f8a","evidence-official-f464cd407a6a9d031720"],"source_locator":"GEARS paper Methods: Overview, Gene coexpression graph encoder, GO graph, Cross-gene effects and gene-specific decoder; gears/model.py: GEARS_Model"},{"label":"Context limits","value":"A gene-expression vector and perturbation set over the configured gene inventory; no fixed nucleotide or amino-acid token window.","status":"source_checked","source_ids":["evidence-official-c037e3419c04936262a0","evidence-official-b5e81b2c497214b18f8a","evidence-official-f464cd407a6a9d031720"],"source_locator":"GEARS paper Methods: Overview, Gene coexpression graph encoder, GO graph, Cross-gene effects and gene-specific decoder; gears/model.py: GEARS_Model"},{"label":"Weights licence","value":"Separate checkpoint-distribution terms are not stated in the inspected release documentation and licence material. The source-code licence alone is not recorded as an explicit weight grant.","status":"unreported","source_ids":["evidence-official-c037e3419c04936262a0","evidence-official-b5e81b2c497214b18f8a","evidence-official-f464cd407a6a9d031720","evidence-official-f06a8695b2915f86a45a"],"source_locator":"GEARS paper Methods: Overview, Gene coexpression graph encoder, GO graph, Cross-gene effects and gene-specific decoder; gears/model.py: GEARS_Model; LICENSE: licence text"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/snap-stanford/GEARS","status":"source_checked","source_ids":["evidence-official-c037e3419c04936262a0","evidence-official-b5e81b2c497214b18f8a","evidence-official-f464cd407a6a9d031720"],"source_locator":"GEARS paper Methods: Overview, Gene coexpression graph encoder, GO graph, Cross-gene effects and gene-specific decoder; gears/model.py: GEARS_Model"},{"label":"Code licence","value":"MIT","status":"source_checked","source_ids":["evidence-official-f06a8695b2915f86a45a"],"source_locator":"LICENSE: licence text"}],"strengths":[{"text":"Provides documented dataset/split handling and training interfaces for single and combinatorial perturbations.","source_ids":["evidence-official-c037e3419c04936262a0","evidence-official-b5e81b2c497214b18f8a","evidence-official-f464cd407a6a9d031720"],"source_locator":"GEARS paper Methods: Overview, Gene coexpression graph encoder, GO graph, Cross-gene effects and gene-specific decoder; gears/model.py: GEARS_Model"}],"limitations":[{"text":"The authors explicitly warn against cross-cell-type transfer, bulk-RNA assumptions and predicting combinations after training only on single perturbations.","source_ids":["evidence-official-c037e3419c04936262a0","evidence-official-b5e81b2c497214b18f8a","evidence-official-f464cd407a6a9d031720"],"source_locator":"GEARS paper Methods: Overview, Gene coexpression graph encoder, GO graph, Cross-gene effects and gene-specific decoder; gears/model.py: GEARS_Model"}],"diagram":{"title":"GEARS workflow","steps":["Perturbation screen","Gene and perturbation graph representations","GEARS prediction","Expression response"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-c037e3419c04936262a0","evidence-official-b5e81b2c497214b18f8a","evidence-official-f464cd407a6a9d031720"],"source_locator":"GEARS paper Methods: Overview, Gene coexpression graph encoder, GO graph, Cross-gene effects and gene-specific decoder; gears/model.py: GEARS_Model"},"coverage":"limited","gaps":["Weights licence: Separate checkpoint-distribution terms are not stated in the inspected release documentation and licence material. The source-code licence alone is not recorded as an explicit weight grant."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}}}
{"id":"catalog-model-geneformer","kind":"model","name":"Geneformer","description":"Existing candidate catalogue entry; exact checkpoint and metadata require primary-source review.","status":"discovered","facets":{"areas":["cells-tissues"],"method_types":["foundation model"]},"source_ids":["catalog-source-geneformer"],"links":[],"attributes":{"entity_level":"family","version":"published checkpoints","reported_name":"Geneformer","access":"Public checkpoints; specify exact version before evaluation.","method_type":"foundation model","historical_missing_metadata":{"checkpoint_revision":"not_yet_extracted","training_data":"not_yet_extracted","licence":"not_yet_extracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Geneformer represents single-cell transcriptomes as ranked genes and learns contextual gene and cell representations.","summary_source_ids":["evidence-official-d3d9afa5c6172982ed01","evidence-official-f60303c5eaa60ed9c77a"],"summary_source_locator":"README.md: Model Description, pretrained model list, Application, Installation and licence frontmatter","sections":[{"title":"How it works","body":"Geneformer represents single-cell transcriptomes as ranked genes and learns contextual gene and cell representations. Transformer encoder trained to recover masked genes from rank-value-encoded expression profiles. The documented inputs are single-cell gene expression converted to corpus-normalized gene ranks. The output consists of contextual gene and cell representations; task-specific outputs after the documented fine-tuning or perturbation workflow.","source_ids":["evidence-official-d3d9afa5c6172982ed01","evidence-official-f60303c5eaa60ed9c77a"],"source_locator":"README.md: Model Description, pretrained model list, Application, Installation and licence frontmatter"},{"title":"Versions and reproducibility","body":"Geneformer-V1-10M, V2-104M, V2-316M and V2-104M_CLcancer; the card states V2-316M is the repository default. V1: 2,048 gene tokens; V2: 4,096. Vocabularies also differ.","source_ids":["evidence-official-d3d9afa5c6172982ed01","evidence-official-f60303c5eaa60ed9c77a"],"source_locator":"README.md: Model Description, pretrained model list, Application, Installation and licence frontmatter"}],"facts":[{"label":"Model type","value":"Masked-gene transcriptomic transformer encoder","status":"source_checked","source_ids":["evidence-official-d3d9afa5c6172982ed01","evidence-official-f60303c5eaa60ed9c77a"],"source_locator":"README.md: Model Description, pretrained model list, Application, Installation and licence frontmatter"},{"label":"Architecture","value":"Transformer encoder trained to recover masked genes from rank-value-encoded expression profiles.","status":"source_checked","source_ids":["evidence-official-d3d9afa5c6172982ed01","evidence-official-f60303c5eaa60ed9c77a"],"source_locator":"README.md: Model Description, pretrained model list, Application, Installation and licence frontmatter"},{"label":"Inputs","value":"Single-cell gene expression converted to corpus-normalized gene ranks.","status":"source_checked","source_ids":["evidence-official-d3d9afa5c6172982ed01","evidence-official-f60303c5eaa60ed9c77a"],"source_locator":"README.md: Model Description, pretrained model list, Application, Installation and licence frontmatter"},{"label":"Outputs","value":"Contextual gene and cell representations; task-specific outputs after the documented fine-tuning or perturbation workflow.","status":"source_checked","source_ids":["evidence-official-d3d9afa5c6172982ed01","evidence-official-f60303c5eaa60ed9c77a"],"source_locator":"README.md: Model Description, pretrained model list, Application, Installation and licence frontmatter"},{"label":"Parameters","value":"V1: 10M; V2: 104M or 316M.","status":"source_checked","source_ids":["evidence-official-d3d9afa5c6172982ed01","evidence-official-f60303c5eaa60ed9c77a"],"source_locator":"README.md: Model Description, pretrained model list, Application, Installation and licence frontmatter"},{"label":"Known versions","value":"Geneformer-V1-10M, V2-104M, V2-316M and V2-104M_CLcancer; the card states V2-316M is the repository default.","status":"source_checked","source_ids":["evidence-official-d3d9afa5c6172982ed01","evidence-official-f60303c5eaa60ed9c77a"],"source_locator":"README.md: Model Description, pretrained model list, Application, Installation and licence frontmatter"},{"label":"Training data","value":"V1: approximately 30M human single-cell transcriptomes. V2: approximately 104M non-cancer human transcriptomes; the cancer continual-learning variant adds approximately 14M cancer cells.","status":"source_checked","source_ids":["evidence-official-d3d9afa5c6172982ed01","evidence-official-f60303c5eaa60ed9c77a"],"source_locator":"README.md: Model Description, pretrained model list, Application, Installation and licence frontmatter"},{"label":"Training cutoff","value":"Training dates reported: June 2021 for V1 and December 2024 for V2. These are training dates, not independently verified data-collection cutoffs.","status":"source_checked","source_ids":["evidence-official-d3d9afa5c6172982ed01","evidence-official-f60303c5eaa60ed9c77a"],"source_locator":"README.md: Model Description, pretrained model list, Application, Installation and licence frontmatter"},{"label":"Context limits","value":"V1: 2,048 gene tokens; V2: 4,096. Vocabularies also differ.","status":"source_checked","source_ids":["evidence-official-d3d9afa5c6172982ed01","evidence-official-f60303c5eaa60ed9c77a"],"source_locator":"README.md: Model Description, pretrained model list, Application, Installation and licence frontmatter"},{"label":"Weights licence","value":"Apache-2.0 is declared in the official model-card metadata.","status":"source_checked","source_ids":["evidence-official-d3d9afa5c6172982ed01","evidence-official-f60303c5eaa60ed9c77a"],"source_locator":"README.md: Model Description, pretrained model list, Application, Installation and licence frontmatter"},{"label":"Access","value":"Official downloadable model/card and usage examples: https://huggingface.co/ctheodoris/Geneformer","status":"source_checked","source_ids":["evidence-official-d3d9afa5c6172982ed01","evidence-official-f60303c5eaa60ed9c77a"],"source_locator":"README.md: Model Description, pretrained model list, Application, Installation and licence frontmatter"},{"label":"Code licence","value":"The official repository declares Apache-2.0 in its model card. No separate code-licence file appears in the complete inspected revision; this records the repository declaration rather than an independently reviewed licence grant for every bundled dependency.","status":"source_checked","source_ids":["evidence-official-d3d9afa5c6172982ed01","evidence-final-model-geneformer-tree"],"source_locator":"Pinned README license metadata; complete two-page recursive repository inventory at revision 1f7fbae4e469a5f4f1af8c111a529cfe1b3829f5"}],"strengths":[{"text":"Self-supervised pretraining uses unlabeled cells and supports downstream gene/cell classification and in-silico perturbation workflows.","source_ids":["evidence-official-d3d9afa5c6172982ed01","evidence-official-f60303c5eaa60ed9c77a"],"source_locator":"README.md: Model Description, pretrained model list, Application, Installation and licence frontmatter"}],"limitations":[{"text":"V1, V2 and cancer-tuned V2 use different corpora and vocabularies. The authors recommend task-specific hyperparameter tuning; there is no universally suitable fine-tuning configuration.","source_ids":["evidence-official-d3d9afa5c6172982ed01","evidence-official-f60303c5eaa60ed9c77a"],"source_locator":"README.md: Model Description, pretrained model list, Application, Installation and licence frontmatter"}],"diagram":{"title":"Geneformer workflow","steps":["Expression profile","Corpus-normalized gene ranks","Masked-gene transformer","Cell and gene representations"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-d3d9afa5c6172982ed01","evidence-official-f60303c5eaa60ed9c77a"],"source_locator":"README.md: Model Description, pretrained model list, Application, Installation and licence frontmatter"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied. Follow-up retrieved the complete original RNA-FM PDF and inspected the full pinned Geneformer repository inventory; unavailable labels were updated only where new evidence resolved the earlier retrieval gap."}}}}
{"id":"catalog-model-kraken2","kind":"model","name":"Kraken2","description":"Existing candidate catalogue entry; exact checkpoint and metadata require primary-source review.","status":"discovered","facets":{"areas":["microbes-communities"],"method_types":["baseline"]},"source_ids":["catalog-source-kraken2"],"links":[],"attributes":{"entity_level":"family","version":"current database pinned at run time","reported_name":"Kraken2","access":"Public classifier; database build/version must be pinned separately.","method_type":"baseline","historical_missing_metadata":{"checkpoint_revision":"not_yet_extracted","training_data":"not_yet_extracted","licence":"not_yet_extracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Kraken 2 assigns taxonomic labels to sequence reads by consulting a reference-derived minimizer database.","summary_source_ids":["evidence-official-40f46dc4b9fcc306773b","evidence-official-33fe0374c8ebbc2c7294","evidence-official-9ba2b6dc494c7b6df951","evidence-official-1f756fefa823dd402b0f"],"summary_source_locator":"docs/MANUAL.markdown: Introduction, Standard Kraken 2 Database, Classification and Confidence scoring","sections":[{"title":"How it works","body":"Kraken 2 breaks query sequences into k-mers and looks up selected minimizers in a compact hash table. Each stored minimizer is associated with a lowest-common-ancestor taxonomic label. The classifier combines that evidence to assign a taxon; the reference database and confidence settings are therefore part of the evaluated procedure.","source_ids":["evidence-official-40f46dc4b9fcc306773b","evidence-official-33fe0374c8ebbc2c7294","evidence-official-9ba2b6dc494c7b6df951","evidence-official-1f756fefa823dd402b0f"],"source_locator":"docs/MANUAL.markdown: Introduction, Standard Kraken 2 Database, Classification and Confidence scoring"},{"title":"Versions and reproducibility","body":"Kraken 2 is a rewrite of Kraken 1 and is not backwards compatible. Read/contig input, not a learned fixed token window.","source_ids":["evidence-official-40f46dc4b9fcc306773b","evidence-official-33fe0374c8ebbc2c7294","evidence-official-9ba2b6dc494c7b6df951","evidence-official-1f756fefa823dd402b0f"],"source_locator":"docs/MANUAL.markdown: Introduction, Standard Kraken 2 Database, Classification and Confidence scoring"}],"facts":[{"label":"Model type","value":"Minimizer-based taxonomic classifier","status":"source_checked","source_ids":["evidence-official-40f46dc4b9fcc306773b","evidence-official-33fe0374c8ebbc2c7294","evidence-official-9ba2b6dc494c7b6df951","evidence-official-1f756fefa823dd402b0f"],"source_locator":"docs/MANUAL.markdown: Introduction, Standard Kraken 2 Database, Classification and Confidence scoring"},{"label":"Architecture","value":"Minimizer-based sequence classification using a compact hash table and lowest-common-ancestor taxonomy assignments.","status":"source_checked","source_ids":["evidence-official-40f46dc4b9fcc306773b","evidence-official-33fe0374c8ebbc2c7294","evidence-official-9ba2b6dc494c7b6df951","evidence-official-1f756fefa823dd402b0f"],"source_locator":"docs/MANUAL.markdown: Introduction, Standard Kraken 2 Database, Classification and Confidence scoring"},{"label":"Inputs","value":"DNA reads or, in translated-search mode, sequences searched against an amino-acid database.","status":"source_checked","source_ids":["evidence-official-40f46dc4b9fcc306773b","evidence-official-33fe0374c8ebbc2c7294","evidence-official-9ba2b6dc494c7b6df951","evidence-official-1f756fefa823dd402b0f"],"source_locator":"docs/MANUAL.markdown: Introduction, Standard Kraken 2 Database, Classification and Confidence scoring"},{"label":"Outputs","value":"Per-read taxonomic assignments and aggregate classification reports.","status":"source_checked","source_ids":["evidence-official-40f46dc4b9fcc306773b","evidence-official-33fe0374c8ebbc2c7294","evidence-official-9ba2b6dc494c7b6df951","evidence-official-1f756fefa823dd402b0f"],"source_locator":"docs/MANUAL.markdown: Introduction, Standard Kraken 2 Database, Classification and Confidence scoring"},{"label":"Parameters","value":"Inapplicable as a neural parameter total; k-mer/minimizer length, confidence and database choices are algorithm settings.","status":"inapplicable","source_ids":["evidence-official-40f46dc4b9fcc306773b","evidence-official-33fe0374c8ebbc2c7294","evidence-official-9ba2b6dc494c7b6df951","evidence-official-1f756fefa823dd402b0f"],"source_locator":"docs/MANUAL.markdown: Introduction, Standard Kraken 2 Database, Classification and Confidence scoring"},{"label":"Known versions","value":"Kraken 2 is a rewrite of Kraken 1 and is not backwards compatible.","status":"source_checked","source_ids":["evidence-official-40f46dc4b9fcc306773b","evidence-official-33fe0374c8ebbc2c7294","evidence-official-9ba2b6dc494c7b6df951","evidence-official-1f756fefa823dd402b0f"],"source_locator":"docs/MANUAL.markdown: Introduction, Standard Kraken 2 Database, Classification and Confidence scoring"},{"label":"Training data","value":"Not neural pretraining: build a database from selected reference sequences and taxonomy.","status":"source_checked","source_ids":["evidence-official-40f46dc4b9fcc306773b","evidence-official-33fe0374c8ebbc2c7294","evidence-official-9ba2b6dc494c7b6df951","evidence-official-1f756fefa823dd402b0f"],"source_locator":"docs/MANUAL.markdown: Introduction, Standard Kraken 2 Database, Classification and Confidence scoring"},{"label":"Training cutoff","value":"Inapplicable to neural pretraining; reference-database and input-data dates must be recorded for each run.","status":"inapplicable","source_ids":["evidence-official-40f46dc4b9fcc306773b","evidence-official-33fe0374c8ebbc2c7294","evidence-official-9ba2b6dc494c7b6df951","evidence-official-1f756fefa823dd402b0f"],"source_locator":"docs/MANUAL.markdown: Introduction, Standard Kraken 2 Database, Classification and Confidence scoring"},{"label":"Context limits","value":"Read/contig input, not a learned fixed token window.","status":"source_checked","source_ids":["evidence-official-40f46dc4b9fcc306773b","evidence-official-33fe0374c8ebbc2c7294","evidence-official-9ba2b6dc494c7b6df951","evidence-official-1f756fefa823dd402b0f"],"source_locator":"docs/MANUAL.markdown: Introduction, Standard Kraken 2 Database, Classification and Confidence scoring"},{"label":"Weights licence","value":"Inapplicable to this classifier: database contents and their licences replace neural weights.","status":"inapplicable","source_ids":["evidence-official-40f46dc4b9fcc306773b","evidence-official-33fe0374c8ebbc2c7294","evidence-official-9ba2b6dc494c7b6df951","evidence-official-1f756fefa823dd402b0f"],"source_locator":"docs/MANUAL.markdown: Introduction, Standard Kraken 2 Database, Classification and Confidence scoring"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/DerrickWood/kraken2","status":"source_checked","source_ids":["evidence-official-40f46dc4b9fcc306773b","evidence-official-33fe0374c8ebbc2c7294","evidence-official-9ba2b6dc494c7b6df951","evidence-official-1f756fefa823dd402b0f"],"source_locator":"docs/MANUAL.markdown: Introduction, Standard Kraken 2 Database, Classification and Confidence scoring"},{"label":"Code licence","value":"MIT","status":"source_checked","source_ids":["evidence-official-bf1a5e03cd84873f4b04"],"source_locator":"LICENSE: licence text"}],"strengths":[{"text":"A reference-based procedural comparator with explicit database construction and confidence settings.","source_ids":["evidence-official-40f46dc4b9fcc306773b","evidence-official-33fe0374c8ebbc2c7294","evidence-official-9ba2b6dc494c7b6df951","evidence-official-1f756fefa823dd402b0f"],"source_locator":"docs/MANUAL.markdown: Introduction, Standard Kraken 2 Database, Classification and Confidence scoring"}],"limitations":[{"text":"Classification depends on the reference database, taxonomy version and minimizer configuration. Compact hashing can introduce false matches; software version alone does not identify a reproducible classifier.","source_ids":["evidence-official-40f46dc4b9fcc306773b","evidence-official-33fe0374c8ebbc2c7294","evidence-official-9ba2b6dc494c7b6df951","evidence-official-1f756fefa823dd402b0f"],"source_locator":"docs/MANUAL.markdown: Introduction, Standard Kraken 2 Database, Classification and Confidence scoring"}],"diagram":{"title":"Kraken2 workflow","steps":["Reference genomes and taxonomy","Minimizer database","Read minimizer lookup","Taxonomic assignment"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-40f46dc4b9fcc306773b","evidence-official-33fe0374c8ebbc2c7294","evidence-official-9ba2b6dc494c7b6df951","evidence-official-1f756fefa823dd402b0f"],"source_locator":"docs/MANUAL.markdown: Introduction, Standard Kraken 2 Database, Classification and Confidence scoring"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}}}
{"id":"catalog-model-metagene-1","kind":"model","name":"METAGENE-1","description":"Existing candidate catalogue entry; exact checkpoint and metadata require primary-source review.","status":"discovered","facets":{"areas":["microbes-communities"],"method_types":["foundation model"]},"source_ids":["catalog-source-metagene-1"],"links":[],"attributes":{"entity_level":"family","version":"6B","reported_name":"METAGENE-1","access":"Public Apache 2.0 checkpoint; 512-token context and large local memory requirement.","method_type":"foundation model","historical_missing_metadata":{"checkpoint_revision":"not_yet_extracted","training_data":"not_yet_extracted","licence":"not_yet_extracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"METAGENE-1 is an autoregressive DNA/RNA sequence model trained on wastewater metagenomic data.","summary_source_ids":["evidence-official-8e894fc6d180746a1808","evidence-official-f71e6c488025a4d82997","evidence-official-a6b46962aae5cfc348c6","evidence-official-b554679ea1503ce3d9f6"],"summary_source_locator":"README.md: Model Overview and Usage; config.json; config.json: model_type, hidden_size, num_hidden_layers, num_attention_heads, vocab_size and sequence-length fields; official pretraining train/config_hub/pretrain/genomicsllama.yml: train.max_seq_length","sections":[{"title":"How it works","body":"METAGENE-1 is an autoregressive DNA/RNA sequence model trained on wastewater metagenomic data. Llama-style autoregressive transformer with 32 layers, width 4,096, 32 attention heads and a 1,024-token vocabulary in the inspected configuration. The documented inputs are DNA or RNA nucleotide sequences. The output consists of sequence generation and representations for downstream metagenomic analyses.","source_ids":["evidence-official-8e894fc6d180746a1808","evidence-official-f71e6c488025a4d82997","evidence-official-a6b46962aae5cfc348c6","evidence-official-b554679ea1503ce3d9f6"],"source_locator":"README.md: Model Overview and Usage; config.json; config.json: model_type, hidden_size, num_hidden_layers, num_attention_heads, vocab_size and sequence-length fields; official pretraining train/config_hub/pretrain/genomicsllama.yml: train.max_seq_length"},{"title":"Versions and reproducibility","body":"METAGENE-1; exact model-card and configuration revision pinned in sources. Configuration fields differ: max_position_embeddings=512 and max_sequence_length=2,048. The effective supported window remains unresolved; neither value is silently promoted to a validated inference limit.","source_ids":["evidence-official-8e894fc6d180746a1808","evidence-official-f71e6c488025a4d82997","evidence-official-a6b46962aae5cfc348c6","evidence-official-b554679ea1503ce3d9f6"],"source_locator":"README.md: Model Overview and Usage; config.json; config.json: model_type, hidden_size, num_hidden_layers, num_attention_heads, vocab_size and sequence-length fields; official pretraining train/config_hub/pretrain/genomicsllama.yml: train.max_seq_length"}],"facts":[{"label":"Model type","value":"Autoregressive metagenomic transformer","status":"source_checked","source_ids":["evidence-official-8e894fc6d180746a1808","evidence-official-f71e6c488025a4d82997","evidence-official-a6b46962aae5cfc348c6","evidence-official-b554679ea1503ce3d9f6"],"source_locator":"README.md: Model Overview and Usage; config.json; config.json: model_type, hidden_size, num_hidden_layers, num_attention_heads, vocab_size and sequence-length fields; official pretraining train/config_hub/pretrain/genomicsllama.yml: train.max_seq_length"},{"label":"Architecture","value":"Llama-style autoregressive transformer with 32 layers, width 4,096, 32 attention heads and a 1,024-token vocabulary in the inspected configuration.","status":"source_checked","source_ids":["evidence-official-8e894fc6d180746a1808","evidence-official-f71e6c488025a4d82997","evidence-official-a6b46962aae5cfc348c6","evidence-official-b554679ea1503ce3d9f6"],"source_locator":"README.md: Model Overview and Usage; config.json; config.json: model_type, hidden_size, num_hidden_layers, num_attention_heads, vocab_size and sequence-length fields; official pretraining train/config_hub/pretrain/genomicsllama.yml: train.max_seq_length"},{"label":"Inputs","value":"DNA or RNA nucleotide sequences.","status":"source_checked","source_ids":["evidence-official-8e894fc6d180746a1808","evidence-official-f71e6c488025a4d82997","evidence-official-a6b46962aae5cfc348c6","evidence-official-b554679ea1503ce3d9f6"],"source_locator":"README.md: Model Overview and Usage; config.json; config.json: model_type, hidden_size, num_hidden_layers, num_attention_heads, vocab_size and sequence-length fields; official pretraining train/config_hub/pretrain/genomicsllama.yml: train.max_seq_length"},{"label":"Outputs","value":"Sequence generation and representations for downstream metagenomic analyses.","status":"source_checked","source_ids":["evidence-official-8e894fc6d180746a1808","evidence-official-f71e6c488025a4d82997","evidence-official-a6b46962aae5cfc348c6","evidence-official-b554679ea1503ce3d9f6"],"source_locator":"README.md: Model Overview and Usage; config.json; config.json: model_type, hidden_size, num_hidden_layers, num_attention_heads, vocab_size and sequence-length fields; official pretraining train/config_hub/pretrain/genomicsllama.yml: train.max_seq_length"},{"label":"Parameters","value":"7 billion.","status":"source_checked","source_ids":["evidence-official-8e894fc6d180746a1808","evidence-official-f71e6c488025a4d82997","evidence-official-a6b46962aae5cfc348c6","evidence-official-b554679ea1503ce3d9f6"],"source_locator":"README.md: Model Overview and Usage; config.json; config.json: model_type, hidden_size, num_hidden_layers, num_attention_heads, vocab_size and sequence-length fields; official pretraining train/config_hub/pretrain/genomicsllama.yml: train.max_seq_length"},{"label":"Known versions","value":"METAGENE-1; exact model-card and configuration revision pinned in sources.","status":"source_checked","source_ids":["evidence-official-8e894fc6d180746a1808","evidence-official-f71e6c488025a4d82997","evidence-official-a6b46962aae5cfc348c6","evidence-official-b554679ea1503ce3d9f6"],"source_locator":"README.md: Model Overview and Usage; config.json; config.json: model_type, hidden_size, num_hidden_layers, num_attention_heads, vocab_size and sequence-length fields; official pretraining train/config_hub/pretrain/genomicsllama.yml: train.max_seq_length"},{"label":"Training data","value":"More than 1.5 trillion base pairs sequenced from human wastewater samples, according to the model card.","status":"source_checked","source_ids":["evidence-official-8e894fc6d180746a1808","evidence-official-f71e6c488025a4d82997","evidence-official-a6b46962aae5cfc348c6","evidence-official-b554679ea1503ce3d9f6"],"source_locator":"README.md: Model Overview and Usage; config.json; config.json: model_type, hidden_size, num_hidden_layers, num_attention_heads, vocab_size and sequence-length fields; official pretraining train/config_hub/pretrain/genomicsllama.yml: train.max_seq_length"},{"label":"Training cutoff","value":"The official pretraining README states that the wastewater corpus is not yet publicly released; a latest sample-collection date is not provided there.","status":"unreported","source_ids":["evidence-official-8e894fc6d180746a1808","evidence-official-f71e6c488025a4d82997","evidence-official-a6b46962aae5cfc348c6","evidence-official-b554679ea1503ce3d9f6"],"source_locator":"README.md: Model Overview and Usage; config.json; config.json: model_type, hidden_size, num_hidden_layers, num_attention_heads, vocab_size and sequence-length fields; official pretraining train/config_hub/pretrain/genomicsllama.yml: train.max_seq_length"},{"label":"Context limits","value":"The released pretraining YAML sets max_seq_length=512. The model config lists max_position_embeddings=512 and max_sequence_length=2,048; these differing configuration fields do not establish a single validated inference maximum.","status":"source_checked","source_ids":["evidence-official-8e894fc6d180746a1808","evidence-official-f71e6c488025a4d82997","evidence-official-a6b46962aae5cfc348c6","evidence-official-b554679ea1503ce3d9f6"],"source_locator":"README.md: Model Overview and Usage; config.json; config.json: model_type, hidden_size, num_hidden_layers, num_attention_heads, vocab_size and sequence-length fields; official pretraining train/config_hub/pretrain/genomicsllama.yml: train.max_seq_length"},{"label":"Weights licence","value":"Apache-2.0 declared in the model card.","status":"source_checked","source_ids":["evidence-official-8e894fc6d180746a1808","evidence-official-f71e6c488025a4d82997","evidence-official-a6b46962aae5cfc348c6","evidence-official-b554679ea1503ce3d9f6"],"source_locator":"README.md: Model Overview and Usage; config.json; config.json: model_type, hidden_size, num_hidden_layers, num_attention_heads, vocab_size and sequence-length fields; official pretraining train/config_hub/pretrain/genomicsllama.yml: train.max_seq_length"},{"label":"Access","value":"Official downloadable model/card and usage examples: https://huggingface.co/metagene-ai/METAGENE-1","status":"source_checked","source_ids":["evidence-official-8e894fc6d180746a1808","evidence-official-f71e6c488025a4d82997","evidence-official-a6b46962aae5cfc348c6","evidence-official-b554679ea1503ce3d9f6"],"source_locator":"README.md: Model Overview and Usage; config.json; config.json: model_type, hidden_size, num_hidden_layers, num_attention_heads, vocab_size and sequence-length fields; official pretraining train/config_hub/pretrain/genomicsllama.yml: train.max_seq_length"},{"label":"Code licence","value":"Apache-2.0","status":"source_checked","source_ids":["evidence-official-c2dd029568d234cb0d16"],"source_locator":"train/LICENSE: licence text"}],"strengths":[{"text":"Its pretraining corpus includes diverse mixed-community DNA and RNA rather than a single reference genome.","source_ids":["evidence-official-8e894fc6d180746a1808","evidence-official-f71e6c488025a4d82997","evidence-official-a6b46962aae5cfc348c6","evidence-official-b554679ea1503ce3d9f6"],"source_locator":"README.md: Model Overview and Usage; config.json; config.json: model_type, hidden_size, num_hidden_layers, num_attention_heads, vocab_size and sequence-length fields; official pretraining train/config_hub/pretrain/genomicsllama.yml: train.max_seq_length"}],"limitations":[{"text":"The stated biosurveillance and pathogen-detection applications require their own downstream evaluation; the pretraining objective is sequence modeling.","source_ids":["evidence-official-8e894fc6d180746a1808","evidence-official-f71e6c488025a4d82997","evidence-official-a6b46962aae5cfc348c6","evidence-official-b554679ea1503ce3d9f6"],"source_locator":"README.md: Model Overview and Usage; config.json; config.json: model_type, hidden_size, num_hidden_layers, num_attention_heads, vocab_size and sequence-length fields; official pretraining train/config_hub/pretrain/genomicsllama.yml: train.max_seq_length"}],"diagram":{"title":"METAGENE-1 workflow","steps":["Nucleotide sequence","BPE tokens","Autoregressive transformer","Sequence or representations"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-8e894fc6d180746a1808","evidence-official-f71e6c488025a4d82997","evidence-official-a6b46962aae5cfc348c6","evidence-official-b554679ea1503ce3d9f6"],"source_locator":"README.md: Model Overview and Usage; config.json; config.json: model_type, hidden_size, num_hidden_layers, num_attention_heads, vocab_size and sequence-length fields; official pretraining train/config_hub/pretrain/genomicsllama.yml: train.max_seq_length"},"coverage":"limited","gaps":["Training cutoff: The official pretraining README states that the wastewater corpus is not yet publicly released; a latest sample-collection date is not provided there."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}}}
{"id":"catalog-model-metaphlan","kind":"model","name":"MetaPhlAn","description":"Existing candidate catalogue entry; exact checkpoint and metadata require primary-source review.","status":"discovered","facets":{"areas":["microbes-communities"],"method_types":["specialist"]},"source_ids":["catalog-source-metaphlan"],"links":[],"attributes":{"entity_level":"family","version":"current marker database pinned at run time","reported_name":"MetaPhlAn","access":"Public profiler; marker database version must be pinned separately.","method_type":"specialist","historical_missing_metadata":{"checkpoint_revision":"not_yet_extracted","training_data":"not_yet_extracted","licence":"not_yet_extracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"MetaPhlAn profiles microbial community composition from shotgun metagenomic reads using clade-specific marker genes.","summary_source_ids":["evidence-official-72fdaed158e4ff851fbe","evidence-official-37f825d1226e59cbbfbf"],"summary_source_locator":"README.md: description; MetaPhlAn 4 paper: Building the expanded SGB catalog and Discussion","sections":[{"title":"How it works","body":"MetaPhlAn profiles microbial community composition from shotgun metagenomic reads using clade-specific marker genes. Reference marker-gene profiling; MetaPhlAn 4 organizes reference and metagenome-assembled genomes into species-level genome bins. The documented inputs are shotgun metagenomic reads and a selected MetaPhlAn marker database. The output consists of taxonomic relative-abundance profiles; StrainPhlAn is a separate strain-level analysis.","source_ids":["evidence-official-72fdaed158e4ff851fbe","evidence-official-37f825d1226e59cbbfbf"],"source_locator":"README.md: description; MetaPhlAn 4 paper: Building the expanded SGB catalog and Discussion"},{"title":"Versions and reproducibility","body":"MetaPhlAn 4 paper and 4.2-linked current documentation; database version is a separate reproducibility requirement. Shotgun reads; no fixed neural token context.","source_ids":["evidence-official-72fdaed158e4ff851fbe","evidence-official-37f825d1226e59cbbfbf"],"source_locator":"README.md: description; MetaPhlAn 4 paper: Building the expanded SGB catalog and Discussion"}],"facts":[{"label":"Model type","value":"Marker-based taxonomic profiling","status":"source_checked","source_ids":["evidence-official-72fdaed158e4ff851fbe","evidence-official-37f825d1226e59cbbfbf"],"source_locator":"README.md: description; MetaPhlAn 4 paper: Building the expanded SGB catalog and Discussion"},{"label":"Architecture","value":"Reference marker-gene profiling; MetaPhlAn 4 organizes reference and metagenome-assembled genomes into species-level genome bins.","status":"source_checked","source_ids":["evidence-official-72fdaed158e4ff851fbe","evidence-official-37f825d1226e59cbbfbf"],"source_locator":"README.md: description; MetaPhlAn 4 paper: Building the expanded SGB catalog and Discussion"},{"label":"Inputs","value":"Shotgun metagenomic reads and a selected MetaPhlAn marker database.","status":"source_checked","source_ids":["evidence-official-72fdaed158e4ff851fbe","evidence-official-37f825d1226e59cbbfbf"],"source_locator":"README.md: description; MetaPhlAn 4 paper: Building the expanded SGB catalog and Discussion"},{"label":"Outputs","value":"Taxonomic relative-abundance profiles; StrainPhlAn is a separate strain-level analysis.","status":"source_checked","source_ids":["evidence-official-72fdaed158e4ff851fbe","evidence-official-37f825d1226e59cbbfbf"],"source_locator":"README.md: description; MetaPhlAn 4 paper: Building the expanded SGB catalog and Discussion"},{"label":"Parameters","value":"Inapplicable as a neural parameter count.","status":"inapplicable","source_ids":["evidence-official-72fdaed158e4ff851fbe","evidence-official-37f825d1226e59cbbfbf"],"source_locator":"README.md: description; MetaPhlAn 4 paper: Building the expanded SGB catalog and Discussion"},{"label":"Known versions","value":"MetaPhlAn 4 paper and 4.2-linked current documentation; database version is a separate reproducibility requirement.","status":"source_checked","source_ids":["evidence-official-72fdaed158e4ff851fbe","evidence-official-37f825d1226e59cbbfbf"],"source_locator":"README.md: description; MetaPhlAn 4 paper: Building the expanded SGB catalog and Discussion"},{"label":"Training data","value":"Reference-derived marker database rather than neural pretraining; record the exact database release.","status":"source_checked","source_ids":["evidence-official-72fdaed158e4ff851fbe","evidence-official-37f825d1226e59cbbfbf"],"source_locator":"README.md: description; MetaPhlAn 4 paper: Building the expanded SGB catalog and Discussion"},{"label":"Training cutoff","value":"Inapplicable to neural pretraining; reference-database and input-data dates must be recorded for each run.","status":"inapplicable","source_ids":["evidence-official-72fdaed158e4ff851fbe","evidence-official-37f825d1226e59cbbfbf"],"source_locator":"README.md: description; MetaPhlAn 4 paper: Building the expanded SGB catalog and Discussion"},{"label":"Context limits","value":"Shotgun reads; no fixed neural token context.","status":"source_checked","source_ids":["evidence-official-72fdaed158e4ff851fbe","evidence-official-37f825d1226e59cbbfbf"],"source_locator":"README.md: description; MetaPhlAn 4 paper: Building the expanded SGB catalog and Discussion"},{"label":"Weights licence","value":"Inapplicable to this procedural method; marker databases have their own provenance and terms.","status":"inapplicable","source_ids":["evidence-official-72fdaed158e4ff851fbe","evidence-official-37f825d1226e59cbbfbf"],"source_locator":"README.md: description; MetaPhlAn 4 paper: Building the expanded SGB catalog and Discussion"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/biobakery/MetaPhlAn","status":"source_checked","source_ids":["evidence-official-72fdaed158e4ff851fbe","evidence-official-37f825d1226e59cbbfbf"],"source_locator":"README.md: description; MetaPhlAn 4 paper: Building the expanded SGB catalog and Discussion"},{"label":"Code licence","value":"MIT","status":"source_checked","source_ids":["evidence-official-1a0775cac85be75449ef"],"source_locator":"license.txt: licence text"}],"strengths":[{"text":"Marker-based profiling can incorporate characterized and previously uncharacterized species groups.","source_ids":["evidence-official-72fdaed158e4ff851fbe","evidence-official-37f825d1226e59cbbfbf"],"source_locator":"README.md: description; MetaPhlAn 4 paper: Building the expanded SGB catalog and Discussion"}],"limitations":[{"text":"Coverage depends on the marker database and habitat. The MetaPhlAn 4 paper identifies remaining gaps for under-studied environmental communities; newer software/databases may have different scope.","source_ids":["evidence-official-72fdaed158e4ff851fbe","evidence-official-37f825d1226e59cbbfbf"],"source_locator":"README.md: description; MetaPhlAn 4 paper: Building the expanded SGB catalog and Discussion"}],"diagram":{"title":"MetaPhlAn workflow","steps":["Metagenomic reads","Marker-gene mapping","Species-group quantification","Relative-abundance profile"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-72fdaed158e4ff851fbe","evidence-official-37f825d1226e59cbbfbf"],"source_locator":"README.md: description; MetaPhlAn 4 paper: Building the expanded SGB catalog and Discussion"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}}}
{"id":"catalog-model-mimic","kind":"model","name":"MIMIC","description":"Existing candidate catalogue entry; exact checkpoint and metadata require primary-source review.","status":"discovered","facets":{"areas":["rna-transcriptomes","proteins-complexes"],"method_types":["foundation model"]},"source_ids":["catalog-source-mimic"],"links":[],"attributes":{"entity_level":"family","version":"1.0","reported_name":"MIMIC","access":"Public MIT code and 1.25B-parameter weights; large local memory requirement.","method_type":"foundation model","historical_missing_metadata":{"checkpoint_revision":"not_yet_extracted","training_data":"not_yet_extracted","licence":"not_yet_extracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"MIMIC represents DNA, RNA, protein and associated molecular measurements in a shared multimodal model.","summary_source_ids":["evidence-official-af8da643e933eae36a68","evidence-official-4565e66fdeb787b99afa","evidence-official-c14c0e6a0fdbe1479658","evidence-official-0ffe1031786a71400ffd"],"summary_source_locator":"MIMIC paper Section 3 and Discussion; official model-card Architecture and License; config.json num_input_tokens and num_target_tokens","sections":[{"title":"How it works","body":"MIMIC represents DNA, RNA, protein and associated molecular measurements in a shared multimodal model. Transformer encoder-decoder: 20 encoder layers and 12 decoder layers, width 1,536, rotary positional embeddings, mixed attention and five register tokens. The documented inputs are co-observed molecular modalities and optional text context, grouped into nucleic, protein and text tracks. The output consists of embeddings or generated modalities, such as residue solvent accessibility or splice-site classes.","source_ids":["evidence-official-af8da643e933eae36a68","evidence-official-4565e66fdeb787b99afa","evidence-official-c14c0e6a0fdbe1479658","evidence-official-0ffe1031786a71400ffd"],"source_locator":"MIMIC paper Section 3 and Discussion; official model-card Architecture and License; config.json num_input_tokens and num_target_tokens"},{"title":"Versions and reproducibility","body":"MIMIC 1.0; load_pretrained(version=\"1.0\") example. Training curriculum increases encoder context from 1k to 10k tokens; the released configuration specifies 10,000 input tokens and 1,000 target tokens.","source_ids":["evidence-official-af8da643e933eae36a68","evidence-official-4565e66fdeb787b99afa","evidence-official-c14c0e6a0fdbe1479658","evidence-official-0ffe1031786a71400ffd"],"source_locator":"MIMIC paper Section 3 and Discussion; official model-card Architecture and License; config.json num_input_tokens and num_target_tokens"}],"facts":[{"label":"Model type","value":"Multimodal biomolecular transformer encoder-decoder","status":"source_checked","source_ids":["evidence-official-af8da643e933eae36a68","evidence-official-4565e66fdeb787b99afa","evidence-official-c14c0e6a0fdbe1479658","evidence-official-0ffe1031786a71400ffd"],"source_locator":"MIMIC paper Section 3 and Discussion; official model-card Architecture and License; config.json num_input_tokens and num_target_tokens"},{"label":"Architecture","value":"Transformer encoder-decoder: 20 encoder layers and 12 decoder layers, width 1,536, rotary positional embeddings, mixed attention and five register tokens.","status":"source_checked","source_ids":["evidence-official-af8da643e933eae36a68","evidence-official-4565e66fdeb787b99afa","evidence-official-c14c0e6a0fdbe1479658","evidence-official-0ffe1031786a71400ffd"],"source_locator":"MIMIC paper Section 3 and Discussion; official model-card Architecture and License; config.json num_input_tokens and num_target_tokens"},{"label":"Inputs","value":"Co-observed molecular modalities and optional text context, grouped into nucleic, protein and text tracks.","status":"source_checked","source_ids":["evidence-official-af8da643e933eae36a68","evidence-official-4565e66fdeb787b99afa","evidence-official-c14c0e6a0fdbe1479658","evidence-official-0ffe1031786a71400ffd"],"source_locator":"MIMIC paper Section 3 and Discussion; official model-card Architecture and License; config.json num_input_tokens and num_target_tokens"},{"label":"Outputs","value":"Embeddings or generated modalities, such as residue solvent accessibility or splice-site classes.","status":"source_checked","source_ids":["evidence-official-af8da643e933eae36a68","evidence-official-4565e66fdeb787b99afa","evidence-official-c14c0e6a0fdbe1479658","evidence-official-0ffe1031786a71400ffd"],"source_locator":"MIMIC paper Section 3 and Discussion; official model-card Architecture and License; config.json num_input_tokens and num_target_tokens"},{"label":"Parameters","value":"Approximately 1.25 billion.","status":"source_checked","source_ids":["evidence-official-af8da643e933eae36a68","evidence-official-4565e66fdeb787b99afa","evidence-official-c14c0e6a0fdbe1479658","evidence-official-0ffe1031786a71400ffd"],"source_locator":"MIMIC paper Section 3 and Discussion; official model-card Architecture and License; config.json num_input_tokens and num_target_tokens"},{"label":"Known versions","value":"MIMIC 1.0; load_pretrained(version=\"1.0\") example.","status":"source_checked","source_ids":["evidence-official-af8da643e933eae36a68","evidence-official-4565e66fdeb787b99afa","evidence-official-c14c0e6a0fdbe1479658","evidence-official-0ffe1031786a71400ffd"],"source_locator":"MIMIC paper Section 3 and Discussion; official model-card Architecture and License; config.json num_input_tokens and num_target_tokens"},{"label":"Training data","value":"LORE aligns approximately 15.5M proteins and 13M RNA transcripts from over 6,000 organisms with associated molecular measurements and over 4B text tokens. Missing modalities are retained as partially observed examples.","status":"source_checked","source_ids":["evidence-official-af8da643e933eae36a68","evidence-official-4565e66fdeb787b99afa","evidence-official-c14c0e6a0fdbe1479658","evidence-official-0ffe1031786a71400ffd"],"source_locator":"MIMIC paper Section 3 and Discussion; official model-card Architecture and License; config.json num_input_tokens and num_target_tokens"},{"label":"Training cutoff","value":"LORE protein/transcript linking uses UniProt release 2024_04. This component version does not establish a common latest-data date for all molecular and text tracks.","status":"source_checked","source_ids":["evidence-official-af8da643e933eae36a68","evidence-official-4565e66fdeb787b99afa","evidence-official-c14c0e6a0fdbe1479658","evidence-official-0ffe1031786a71400ffd"],"source_locator":"MIMIC paper Section 3 and Discussion; official model-card Architecture and License; config.json num_input_tokens and num_target_tokens"},{"label":"Context limits","value":"Training curriculum increases encoder context from 1k to 10k tokens; the released configuration specifies 10,000 input tokens and 1,000 target tokens.","status":"source_checked","source_ids":["evidence-official-af8da643e933eae36a68","evidence-official-4565e66fdeb787b99afa","evidence-official-c14c0e6a0fdbe1479658","evidence-official-0ffe1031786a71400ffd"],"source_locator":"MIMIC paper Section 3 and Discussion; official model-card Architecture and License; config.json num_input_tokens and num_target_tokens"},{"label":"Weights licence","value":"MIT; model card explicitly covers both model and source code.","status":"source_checked","source_ids":["evidence-official-af8da643e933eae36a68","evidence-official-4565e66fdeb787b99afa","evidence-official-c14c0e6a0fdbe1479658","evidence-official-0ffe1031786a71400ffd"],"source_locator":"MIMIC paper Section 3 and Discussion; official model-card Architecture and License; config.json num_input_tokens and num_target_tokens"},{"label":"Access","value":"Official downloadable model/card and usage examples: https://huggingface.co/polymathic-ai/MIMIC","status":"source_checked","source_ids":["evidence-official-af8da643e933eae36a68","evidence-official-4565e66fdeb787b99afa","evidence-official-c14c0e6a0fdbe1479658","evidence-official-0ffe1031786a71400ffd"],"source_locator":"MIMIC paper Section 3 and Discussion; official model-card Architecture and License; config.json num_input_tokens and num_target_tokens"},{"label":"Code licence","value":"MIT","status":"source_checked","source_ids":["evidence-official-34faacf1ad53d6aa6bef"],"source_locator":"LICENSE: licence text"}],"strengths":[{"text":"The interface accepts partially observed modality sets and supports both embedding and conditional generation.","source_ids":["evidence-official-af8da643e933eae36a68","evidence-official-4565e66fdeb787b99afa","evidence-official-c14c0e6a0fdbe1479658","evidence-official-0ffe1031786a71400ffd"],"source_locator":"MIMIC paper Section 3 and Discussion; official model-card Architecture and License; config.json num_input_tokens and num_target_tokens"}],"limitations":[{"text":"The paper identifies uneven modality coverage, missing biological measurements and limited encoder/decoder windows. Predictions for a missing modality remain model inferences rather than experimental observations.","source_ids":["evidence-official-af8da643e933eae36a68","evidence-official-4565e66fdeb787b99afa","evidence-official-c14c0e6a0fdbe1479658","evidence-official-0ffe1031786a71400ffd"],"source_locator":"MIMIC paper Section 3 and Discussion; official model-card Architecture and License; config.json num_input_tokens and num_target_tokens"}],"diagram":{"title":"MIMIC workflow","steps":["Observed modalities","Shared encoder","Track-aware decoder","Embeddings or generated tracks"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-af8da643e933eae36a68","evidence-official-4565e66fdeb787b99afa","evidence-official-c14c0e6a0fdbe1479658","evidence-official-0ffe1031786a71400ffd"],"source_locator":"MIMIC paper Section 3 and Discussion; official model-card Architecture and License; config.json num_input_tokens and num_target_tokens"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}}}
{"id":"catalog-model-mrna-fm","kind":"model","name":"mRNA-FM","description":"Existing candidate catalogue entry; exact checkpoint and metadata require primary-source review.","status":"discovered","facets":{"areas":["rna-transcriptomes"],"method_types":["foundation model"]},"source_ids":["catalog-source-mrna-fm"],"links":[{"relation":"variant_of","target_id":"discovery-model-rna-fm"}],"attributes":{"entity_level":"family","version":"codon-tokenised","reported_name":"mRNA-FM","access":"Public checkpoint trained on coding sequences (CDS); input must be codon aligned. UTR-only sequences are outside its training modality.","method_type":"foundation model","historical_missing_metadata":{"checkpoint_revision":"not_yet_extracted","training_data":"not_yet_extracted","licence":"not_yet_extracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"mRNA-FM encodes coding RNA with codon-level tokens to produce representations for downstream analysis.","summary_source_ids":["evidence-official-fd8e332abdf04a75195b","evidence-official-e17e3864c464d981afc7"],"summary_source_locator":"README.md: Foundation Models table and mRNA-FM quick-start example","sections":[{"title":"How it works","body":"mRNA-FM encodes coding RNA with codon-level tokens to produce representations for downstream analysis. 12-layer transformer encoder with hidden width 1,280 and codon-level input tokens. The documented inputs are coding RNA in the correct reading frame, with sequence length divisible by three. The output consists of contextual token embeddings for a specified downstream RNA task.","source_ids":["evidence-official-fd8e332abdf04a75195b","evidence-official-e17e3864c464d981afc7"],"source_locator":"README.md: Foundation Models table and mRNA-FM quick-start example"},{"title":"Versions and reproducibility","body":"mrna_fm_t12; distinct from base-tokenized rna_fm_t12. The applicable input limits require configuration-specific checking.","source_ids":["evidence-official-fd8e332abdf04a75195b","evidence-official-e17e3864c464d981afc7"],"source_locator":"README.md: Foundation Models table and mRNA-FM quick-start example"}],"facts":[{"label":"Model type","value":"Codon-token RNA transformer encoder","status":"source_checked","source_ids":["evidence-official-fd8e332abdf04a75195b","evidence-official-e17e3864c464d981afc7"],"source_locator":"README.md: Foundation Models table and mRNA-FM quick-start example"},{"label":"Architecture","value":"12-layer transformer encoder with hidden width 1,280 and codon-level input tokens.","status":"source_checked","source_ids":["evidence-official-fd8e332abdf04a75195b","evidence-official-e17e3864c464d981afc7"],"source_locator":"README.md: Foundation Models table and mRNA-FM quick-start example"},{"label":"Inputs","value":"Coding RNA in the correct reading frame, with sequence length divisible by three.","status":"source_checked","source_ids":["evidence-official-fd8e332abdf04a75195b","evidence-official-e17e3864c464d981afc7"],"source_locator":"README.md: Foundation Models table and mRNA-FM quick-start example"},{"label":"Outputs","value":"Contextual token embeddings for a specified downstream RNA task.","status":"source_checked","source_ids":["evidence-official-fd8e332abdf04a75195b","evidence-official-e17e3864c464d981afc7"],"source_locator":"README.md: Foundation Models table and mRNA-FM quick-start example"},{"label":"Parameters","value":"239M, as printed in the official Foundation Models table.","status":"source_checked","source_ids":["evidence-official-fd8e332abdf04a75195b","evidence-official-e17e3864c464d981afc7"],"source_locator":"README.md: Foundation Models table and mRNA-FM quick-start example"},{"label":"Known versions","value":"mrna_fm_t12; distinct from base-tokenized rna_fm_t12.","status":"source_checked","source_ids":["evidence-official-fd8e332abdf04a75195b","evidence-official-e17e3864c464d981afc7"],"source_locator":"README.md: Foundation Models table and mRNA-FM quick-start example"},{"label":"Training data","value":"45M messenger RNA sequences, as printed in the official Foundation Models table.","status":"source_checked","source_ids":["evidence-official-fd8e332abdf04a75195b","evidence-official-e17e3864c464d981afc7"],"source_locator":"README.md: Foundation Models table and mRNA-FM quick-start example"},{"label":"Training cutoff","value":"The official model table reports 45M coding RNAs but does not identify a latest-data date for that separate mRNA-FM corpus.","status":"unreported","source_ids":["evidence-official-fd8e332abdf04a75195b","evidence-official-e17e3864c464d981afc7"],"source_locator":"README.md: Foundation Models table and mRNA-FM quick-start example"},{"label":"Context limits","value":"The inspected mRNA-FM usage example and loader do not specify a validated maximum codon-sequence length. The retrieved original RNA-FM paper describes the nucleotide-based non-coding RNA model; its limit cannot establish the separate mRNA-FM checkpoint limit.","status":"unreported","source_ids":["evidence-official-fd8e332abdf04a75195b","evidence-official-e17e3864c464d981afc7","evidence-final-model-rna-fm-paper"],"source_locator":"Official README: Foundation Models and mRNA-FM quick start; loader; original RNA-FM paper Methods, pp.22–23"},{"label":"Weights licence","value":"Separate checkpoint-distribution terms are not stated in the inspected release documentation and licence material. The source-code licence alone is not recorded as an explicit weight grant.","status":"unreported","source_ids":["evidence-official-fd8e332abdf04a75195b","evidence-official-e17e3864c464d981afc7","evidence-official-3fde3df73e79e455bd86"],"source_locator":"README.md: Foundation Models table and mRNA-FM quick-start example; LICENSE: licence text"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/ml4bio/RNA-FM","status":"source_checked","source_ids":["evidence-official-fd8e332abdf04a75195b","evidence-official-e17e3864c464d981afc7"],"source_locator":"README.md: Foundation Models table and mRNA-FM quick-start example"},{"label":"Code licence","value":"MIT","status":"source_checked","source_ids":["evidence-official-3fde3df73e79e455bd86"],"source_locator":"LICENSE: licence text"}],"strengths":[{"text":"The repository exposes embedding extraction and examples for downstream RNA analyses.","source_ids":["evidence-official-fd8e332abdf04a75195b","evidence-official-e17e3864c464d981afc7"],"source_locator":"README.md: Foundation Models table and mRNA-FM quick-start example"}],"limitations":[{"text":"Base-level RNA-FM and codon-level mRNA-FM are not interchangeable. The task head and tokenization must be specified in each evaluation.","source_ids":["evidence-official-fd8e332abdf04a75195b","evidence-official-e17e3864c464d981afc7"],"source_locator":"README.md: Foundation Models table and mRNA-FM quick-start example"}],"diagram":{"title":"mRNA-FM workflow","steps":["Coding RNA","Codon tokenizer","12-layer transformer encoder","Contextual codon representations"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-fd8e332abdf04a75195b","evidence-official-e17e3864c464d981afc7"],"source_locator":"README.md: Foundation Models table and mRNA-FM quick-start example"},"coverage":"limited","gaps":["Training cutoff: The official model table reports 45M coding RNAs but does not identify a latest-data date for that separate mRNA-FM corpus.","Context limits: The inspected mRNA-FM usage example and loader do not specify a validated maximum codon-sequence length. The retrieved original RNA-FM paper describes the nucleotide-based non-coding RNA model; its limit cannot establish the separate mRNA-FM checkpoint limit.","Weights licence: Separate checkpoint-distribution terms are not stated in the inspected release documentation and licence material. The source-code licence alone is not recorded as an explicit weight grant."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied. Follow-up retrieved the complete original RNA-FM PDF and inspected the full pinned Geneformer repository inventory; unavailable labels were updated only where new evidence resolved the earlier retrieval gap."}}}}
{"id":"catalog-model-nt-v2","kind":"model","name":"Nucleotide Transformer v2","description":"Existing candidate catalogue entry; exact checkpoint and metadata require primary-source review.","status":"discovered","facets":{"areas":["dna-genomes"],"method_types":["foundation model"]},"source_ids":["catalog-source-nt-v2"],"links":[],"attributes":{"entity_level":"family","version":"50M multi-species","reported_name":"Nucleotide Transformer v2","access":"Public checkpoint.","method_type":"foundation model","historical_missing_metadata":{"checkpoint_revision":"not_yet_extracted","training_data":"not_yet_extracted","licence":"not_yet_extracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Nucleotide Transformer v2 represents DNA using an encoder pretrained on multiple species.","summary_source_ids":["evidence-official-4abb9affb1a9e5438c91","evidence-official-6b79ddfdd330693bf4fb","evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-aadbeb0f10ec55d34f5f"],"summary_source_locator":"README.md: Model Summary, Training data and licence metadata; config.json; Nucleotide Transformer paper Methods: Architecture and Training (v2); source conflict with 50M card retained","sections":[{"title":"How it works","body":"Nucleotide Transformer v2 represents DNA using an encoder pretrained on multiple species. Encoder-only transformer with 6-mer tokenization, rotary position embeddings and SwiGLU feed-forward layers. The documented inputs are DNA sequences with tokenization determined by 6-mers and individual ambiguous/remainder bases. The output consists of contextual DNA embeddings and masked-token probabilities; downstream tasks need adaptation.","source_ids":["evidence-official-4abb9affb1a9e5438c91","evidence-official-6b79ddfdd330693bf4fb","evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-aadbeb0f10ec55d34f5f"],"source_locator":"README.md: Model Summary, Training data and licence metadata; config.json; Nucleotide Transformer paper Methods: Architecture and Training (v2); source conflict with 50M card retained"},{"title":"Versions and reproducibility","body":"nucleotide-transformer-v2-50m-multi-species is the linked checkpoint; distinguish it from other family scales. Source conflict retained: the model card describes 1,000-token pretraining, while the official NT-v2 documentation describes 2,048-token capacity. Token and base counts must be stated separately.","source_ids":["evidence-official-4abb9affb1a9e5438c91","evidence-official-6b79ddfdd330693bf4fb","evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-aadbeb0f10ec55d34f5f"],"source_locator":"README.md: Model Summary, Training data and licence metadata; config.json; Nucleotide Transformer paper Methods: Architecture and Training (v2); source conflict with 50M card retained"}],"facts":[{"label":"Model type","value":"DNA transformer encoder","status":"source_checked","source_ids":["evidence-official-4abb9affb1a9e5438c91","evidence-official-6b79ddfdd330693bf4fb","evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-aadbeb0f10ec55d34f5f"],"source_locator":"README.md: Model Summary, Training data and licence metadata; config.json; Nucleotide Transformer paper Methods: Architecture and Training (v2); source conflict with 50M card retained"},{"label":"Architecture","value":"Encoder-only transformer with 6-mer tokenization, rotary position embeddings and SwiGLU feed-forward layers.","status":"source_checked","source_ids":["evidence-official-4abb9affb1a9e5438c91","evidence-official-6b79ddfdd330693bf4fb","evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-aadbeb0f10ec55d34f5f"],"source_locator":"README.md: Model Summary, Training data and licence metadata; config.json; Nucleotide Transformer paper Methods: Architecture and Training (v2); source conflict with 50M card retained"},{"label":"Inputs","value":"DNA sequences with tokenization determined by 6-mers and individual ambiguous/remainder bases.","status":"source_checked","source_ids":["evidence-official-4abb9affb1a9e5438c91","evidence-official-6b79ddfdd330693bf4fb","evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-aadbeb0f10ec55d34f5f"],"source_locator":"README.md: Model Summary, Training data and licence metadata; config.json; Nucleotide Transformer paper Methods: Architecture and Training (v2); source conflict with 50M card retained"},{"label":"Outputs","value":"Contextual DNA embeddings and masked-token probabilities; downstream tasks need adaptation.","status":"source_checked","source_ids":["evidence-official-4abb9affb1a9e5438c91","evidence-official-6b79ddfdd330693bf4fb","evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-aadbeb0f10ec55d34f5f"],"source_locator":"README.md: Model Summary, Training data and licence metadata; config.json; Nucleotide Transformer paper Methods: Architecture and Training (v2); source conflict with 50M card retained"},{"label":"Parameters","value":"The linked checkpoint is 50M; v2 family also includes 100M, 250M and 500M.","status":"source_checked","source_ids":["evidence-official-4abb9affb1a9e5438c91","evidence-official-6b79ddfdd330693bf4fb","evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-aadbeb0f10ec55d34f5f"],"source_locator":"README.md: Model Summary, Training data and licence metadata; config.json; Nucleotide Transformer paper Methods: Architecture and Training (v2); source conflict with 50M card retained"},{"label":"Known versions","value":"nucleotide-transformer-v2-50m-multi-species is the linked checkpoint; distinguish it from other family scales.","status":"source_checked","source_ids":["evidence-official-4abb9affb1a9e5438c91","evidence-official-6b79ddfdd330693bf4fb","evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-aadbeb0f10ec55d34f5f"],"source_locator":"README.md: Model Summary, Training data and licence metadata; config.json; Nucleotide Transformer paper Methods: Architecture and Training (v2); source conflict with 50M card retained"},{"label":"Training data","value":"The linked 50M card reports 850 reference genomes, excluding plants and viruses; 174B source nucleotides and 300B training tokens. Source corpus size is distinct from repeated training exposure.","status":"source_checked","source_ids":["evidence-official-4abb9affb1a9e5438c91","evidence-official-6b79ddfdd330693bf4fb","evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-aadbeb0f10ec55d34f5f"],"source_locator":"README.md: Model Summary, Training data and licence metadata; config.json; Nucleotide Transformer paper Methods: Architecture and Training (v2); source conflict with 50M card retained"},{"label":"Training cutoff","value":"The paper specifies human-reference, 1000 Genomes and multispecies training collections by variant. A single latest-deposition date for all sequences is not supplied in the inspected pretraining-data section.","status":"unreported","source_ids":["evidence-official-4abb9affb1a9e5438c91","evidence-official-6b79ddfdd330693bf4fb","evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-aadbeb0f10ec55d34f5f"],"source_locator":"README.md: Model Summary, Training data and licence metadata; config.json; Nucleotide Transformer paper Methods: Architecture and Training (v2); source conflict with 50M card retained"},{"label":"Context limits","value":"The inspected 50M model card describes 1,000-token pretraining, whereas the NT-v2 paper and documentation describe 2,048-token capacity. Preserve this source discrepancy and the selected checkpoint; tokens and bases differ.","status":"source_checked","source_ids":["evidence-official-4abb9affb1a9e5438c91","evidence-official-6b79ddfdd330693bf4fb","evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-aadbeb0f10ec55d34f5f"],"source_locator":"README.md: Model Summary, Training data and licence metadata; config.json; Nucleotide Transformer paper Methods: Architecture and Training (v2); source conflict with 50M card retained"},{"label":"Weights licence","value":"CC-BY-NC-SA-4.0 as declared in the official model card.","status":"source_checked","source_ids":["evidence-official-4abb9affb1a9e5438c91","evidence-official-6b79ddfdd330693bf4fb","evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-aadbeb0f10ec55d34f5f"],"source_locator":"README.md: Model Summary, Training data and licence metadata; config.json; Nucleotide Transformer paper Methods: Architecture and Training (v2); source conflict with 50M card retained"},{"label":"Access","value":"Official downloadable model/card and usage examples: https://huggingface.co/InstaDeepAI/nucleotide-transformer-v2-50m-multi-species","status":"source_checked","source_ids":["evidence-official-4abb9affb1a9e5438c91","evidence-official-6b79ddfdd330693bf4fb","evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-aadbeb0f10ec55d34f5f"],"source_locator":"README.md: Model Summary, Training data and licence metadata; config.json; Nucleotide Transformer paper Methods: Architecture and Training (v2); source conflict with 50M card retained"},{"label":"Code licence","value":"CC-BY-NC-SA-4.0","status":"source_checked","source_ids":["evidence-official-7e4b193e47ba209860a1"],"source_locator":"LICENSE.md: licence text"}],"strengths":[{"text":"The v2 architecture extends the token context relative to v1 while providing several parameter scales.","source_ids":["evidence-official-4abb9affb1a9e5438c91","evidence-official-6b79ddfdd330693bf4fb","evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-aadbeb0f10ec55d34f5f"],"source_locator":"README.md: Model Summary, Training data and licence metadata; config.json; Nucleotide Transformer paper Methods: Architecture and Training (v2); source conflict with 50M card retained"}],"limitations":[{"text":"Token length is not identical to base-pair length: ambiguous bases consume individual tokens. The linked 50M card does not make every v2 checkpoint identical.","source_ids":["evidence-official-4abb9affb1a9e5438c91","evidence-official-6b79ddfdd330693bf4fb","evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-aadbeb0f10ec55d34f5f"],"source_locator":"README.md: Model Summary, Training data and licence metadata; config.json; Nucleotide Transformer paper Methods: Architecture and Training (v2); source conflict with 50M card retained"}],"diagram":{"title":"Nucleotide Transformer v2 workflow","steps":["DNA sequence","6-mer tokenizer","NT-v2 encoder","Embeddings or task adaptation"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-4abb9affb1a9e5438c91","evidence-official-6b79ddfdd330693bf4fb","evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-aadbeb0f10ec55d34f5f"],"source_locator":"README.md: Model Summary, Training data and licence metadata; config.json; Nucleotide Transformer paper Methods: Architecture and Training (v2); source conflict with 50M card retained"},"coverage":"limited","gaps":["Training cutoff: The paper specifies human-reference, 1000 Genomes and multispecies training collections by variant. A single latest-deposition date for all sequences is not supplied in the inspected pretraining-data section."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}}}
{"id":"catalog-model-pangolin","kind":"model","name":"Pangolin","description":"Existing candidate catalogue entry; exact checkpoint and metadata require primary-source review.","status":"discovered","facets":{"areas":["dna-genomes"],"method_types":["specialist"]},"source_ids":["catalog-source-pangolin"],"links":[{"relation":"family","target_id":"discovery-model-pangolin"}],"attributes":{"entity_level":"family","version":"published checkpoints","reported_name":"Pangolin","access":"Public specialist code and models under GPL-3.0.","method_type":"specialist","historical_missing_metadata":{"checkpoint_revision":"not_yet_extracted","training_data":"not_yet_extracted","licence":"not_yet_extracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Pangolin predicts splice-site strength and changes caused by genetic variants.","summary_source_ids":["evidence-official-eaf3efab8850672f44e4","evidence-official-de1ee43bdd6f04de9850"],"summary_source_locator":"Paper: Deep neural network architecture and Generating training and test sets; README.md: Usage","sections":[{"title":"How it works","body":"Pangolin predicts splice-site strength and changes caused by genetic variants. Dilated convolutional network with 16 residual blocks and skip connections; separate probability and usage outputs for heart, liver, brain and testis. The documented inputs are VCF or CSV variants, reference FASTA and matching gene annotations; custom sequence inference is also available. The output consists of predicted increases/decreases in splice-site strength and their positions.","source_ids":["evidence-official-eaf3efab8850672f44e4","evidence-official-de1ee43bdd6f04de9850"],"source_locator":"Paper: Deep neural network architecture and Generating training and test sets; README.md: Usage"},{"title":"Versions and reproducibility","body":"Pangolin implementation; gene-annotation database and selected weights must be recorded with a run. 5,000 bases upstream and downstream each output position; minimum 10,001-base input for one prediction, with 15,000-base training blocks producing 5,000 central outputs.","source_ids":["evidence-official-eaf3efab8850672f44e4","evidence-official-de1ee43bdd6f04de9850"],"source_locator":"Paper: Deep neural network architecture and Generating training and test sets; README.md: Usage"}],"facts":[{"label":"Model type","value":"Dilated convolutional splicing predictor","status":"source_checked","source_ids":["evidence-official-eaf3efab8850672f44e4","evidence-official-de1ee43bdd6f04de9850"],"source_locator":"Paper: Deep neural network architecture and Generating training and test sets; README.md: Usage"},{"label":"Architecture","value":"Dilated convolutional network with 16 residual blocks and skip connections; separate probability and usage outputs for heart, liver, brain and testis.","status":"source_checked","source_ids":["evidence-official-eaf3efab8850672f44e4","evidence-official-de1ee43bdd6f04de9850"],"source_locator":"Paper: Deep neural network architecture and Generating training and test sets; README.md: Usage"},{"label":"Inputs","value":"VCF or CSV variants, reference FASTA and matching gene annotations; custom sequence inference is also available.","status":"source_checked","source_ids":["evidence-official-eaf3efab8850672f44e4","evidence-official-de1ee43bdd6f04de9850"],"source_locator":"Paper: Deep neural network architecture and Generating training and test sets; README.md: Usage"},{"label":"Outputs","value":"Predicted increases/decreases in splice-site strength and their positions.","status":"source_checked","source_ids":["evidence-official-eaf3efab8850672f44e4","evidence-official-de1ee43bdd6f04de9850"],"source_locator":"Paper: Deep neural network architecture and Generating training and test sets; README.md: Usage"},{"label":"Parameters","value":"The reviewed architecture section specifies the dilated residual network, but does not give a complete parameter total for the released ensemble.","status":"unreported","source_ids":["evidence-official-eaf3efab8850672f44e4","evidence-official-de1ee43bdd6f04de9850"],"source_locator":"Paper: Deep neural network architecture and Generating training and test sets; README.md: Usage"},{"label":"Known versions","value":"Pangolin implementation; gene-annotation database and selected weights must be recorded with a run.","status":"source_checked","source_ids":["evidence-official-eaf3efab8850672f44e4","evidence-official-de1ee43bdd6f04de9850"],"source_locator":"Paper: Deep neural network architecture and Generating training and test sets; README.md: Usage"},{"label":"Training data","value":"Human, rhesus macaque, mouse and rat sequence/splicing data. Human test chromosomes 1, 3, 5, 7 and 9 are held out, with homologous training genes filtered using Ensembl BioMart.","status":"source_checked","source_ids":["evidence-official-eaf3efab8850672f44e4","evidence-official-de1ee43bdd6f04de9850"],"source_locator":"Paper: Deep neural network architecture and Generating training and test sets; README.md: Usage"},{"label":"Training cutoff","value":"Training annotations are GENCODE 34 (human), Ensembl 100 (rhesus), GENCODE M25 (mouse) and Ensembl 101 (rat). These component releases do not establish one latest RNA-seq collection date.","status":"source_checked","source_ids":["evidence-official-eaf3efab8850672f44e4","evidence-official-de1ee43bdd6f04de9850"],"source_locator":"Paper: Deep neural network architecture and Generating training and test sets; README.md: Usage"},{"label":"Context limits","value":"5,000 bases upstream and downstream each output position; minimum 10,001-base input for one prediction, with 15,000-base training blocks producing 5,000 central outputs.","status":"source_checked","source_ids":["evidence-official-eaf3efab8850672f44e4","evidence-official-de1ee43bdd6f04de9850"],"source_locator":"Paper: Deep neural network architecture and Generating training and test sets; README.md: Usage"},{"label":"Weights licence","value":"Separate checkpoint-distribution terms are not stated in the inspected release documentation and licence material. The source-code licence alone is not recorded as an explicit weight grant.","status":"unreported","source_ids":["evidence-official-eaf3efab8850672f44e4","evidence-official-de1ee43bdd6f04de9850","evidence-official-dd29c6cbb629171059a6"],"source_locator":"Paper: Deep neural network architecture and Generating training and test sets; README.md: Usage; LICENSE: licence text"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/tkzeng/Pangolin","status":"source_checked","source_ids":["evidence-official-eaf3efab8850672f44e4","evidence-official-de1ee43bdd6f04de9850"],"source_locator":"Paper: Deep neural network architecture and Generating training and test sets; README.md: Usage"},{"label":"Code licence","value":"GPL-3.0; inspect the pinned licence and any file-specific terms.","status":"source_checked","source_ids":["evidence-official-dd29c6cbb629171059a6"],"source_locator":"LICENSE: licence text"}],"strengths":[{"text":"Supports custom sequences and annotation-aware variant scoring with configurable search distance.","source_ids":["evidence-official-eaf3efab8850672f44e4","evidence-official-de1ee43bdd6f04de9850"],"source_locator":"Paper: Deep neural network architecture and Generating training and test sets; README.md: Usage"}],"limitations":[{"text":"Only substitutions and simple insertions/deletions are supported. The documented tool skips variants outside annotated genes, near chromosome ends, inconsistent with the reference or beyond supported deletion lengths.","source_ids":["evidence-official-eaf3efab8850672f44e4","evidence-official-de1ee43bdd6f04de9850"],"source_locator":"Paper: Deep neural network architecture and Generating training and test sets; README.md: Usage"}],"diagram":{"title":"Pangolin workflow","steps":["Variant and reference genome","Construct sequence inputs","Splice-strength prediction","Reference/alternate comparison"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-eaf3efab8850672f44e4","evidence-official-de1ee43bdd6f04de9850"],"source_locator":"Paper: Deep neural network architecture and Generating training and test sets; README.md: Usage"},"coverage":"limited","gaps":["Parameters: The reviewed architecture section specifies the dilated residual network, but does not give a complete parameter total for the released ensemble.","Weights licence: Separate checkpoint-distribution terms are not stated in the inspected release documentation and licence material. The source-code licence alone is not recorded as an explicit weight grant."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}}}
{"id":"catalog-model-prokbert","kind":"model","name":"ProkBERT","description":"Existing candidate catalogue entry; exact checkpoint and metadata require primary-source review.","status":"discovered","facets":{"areas":["microbes-communities"],"method_types":["foundation model"]},"source_ids":["catalog-source-prokbert"],"links":[],"attributes":{"entity_level":"family","version":"mini","reported_name":"ProkBERT","access":"Public model family and mini checkpoint.","method_type":"foundation model","historical_missing_metadata":{"checkpoint_revision":"not_yet_extracted","training_data":"not_yet_extracted","licence":"not_yet_extracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"ProkBERT is a family of microbial DNA encoders used for sequence representation, promoter prediction and phage identification.","summary_source_ids":["evidence-official-2efbfaba5a1c09f4f7fa","evidence-official-965600234c25fbd104c8","evidence-official-ee6e270af125ddf16eb3","evidence-official-5e268f31f0347fc5564c"],"summary_source_locator":"ProkBERT paper Section 2.1.2 and Table 1; official ProkBERT-mini card: Model Description and licence metadata","sections":[{"title":"How it works","body":"ProkBERT is a family of microbial DNA encoders used for sequence representation, promoter prediction and phage identification. MegatronBERT-based encoder using local-context-aware k-mer tokenization, learned relative key/value positions and masked-language pretraining. The documented inputs are microbial DNA segments tokenized with the selected variant. The output consists of sequence representations or predictions from separately fine-tuned promoter/phage heads.","source_ids":["evidence-official-2efbfaba5a1c09f4f7fa","evidence-official-965600234c25fbd104c8","evidence-official-ee6e270af125ddf16eb3","evidence-official-5e268f31f0347fc5564c"],"source_locator":"ProkBERT paper Section 2.1.2 and Table 1; official ProkBERT-mini card: Model Description and licence metadata"},{"title":"Versions and reproducibility","body":"ProkBERT-mini (6-mer, shift 1), mini-c (single base), mini-long (6-mer, shift 2), plus promoter/phage fine-tunes. The paper reports maximum sequence lengths of 1,024bp for mini and 2,048bp for mini-long; tokenization stride distinguishes the variants.","source_ids":["evidence-official-2efbfaba5a1c09f4f7fa","evidence-official-965600234c25fbd104c8","evidence-official-ee6e270af125ddf16eb3","evidence-official-5e268f31f0347fc5564c"],"source_locator":"ProkBERT paper Section 2.1.2 and Table 1; official ProkBERT-mini card: Model Description and licence metadata"}],"facts":[{"label":"Model type","value":"Microbial DNA BERT encoder","status":"source_checked","source_ids":["evidence-official-2efbfaba5a1c09f4f7fa","evidence-official-965600234c25fbd104c8","evidence-official-ee6e270af125ddf16eb3","evidence-official-5e268f31f0347fc5564c"],"source_locator":"ProkBERT paper Section 2.1.2 and Table 1; official ProkBERT-mini card: Model Description and licence metadata"},{"label":"Architecture","value":"MegatronBERT-based encoder using local-context-aware k-mer tokenization, learned relative key/value positions and masked-language pretraining.","status":"source_checked","source_ids":["evidence-official-2efbfaba5a1c09f4f7fa","evidence-official-965600234c25fbd104c8","evidence-official-ee6e270af125ddf16eb3","evidence-official-5e268f31f0347fc5564c"],"source_locator":"ProkBERT paper Section 2.1.2 and Table 1; official ProkBERT-mini card: Model Description and licence metadata"},{"label":"Inputs","value":"Microbial DNA segments tokenized with the selected variant.","status":"source_checked","source_ids":["evidence-official-2efbfaba5a1c09f4f7fa","evidence-official-965600234c25fbd104c8","evidence-official-ee6e270af125ddf16eb3","evidence-official-5e268f31f0347fc5564c"],"source_locator":"ProkBERT paper Section 2.1.2 and Table 1; official ProkBERT-mini card: Model Description and licence metadata"},{"label":"Outputs","value":"Sequence representations or predictions from separately fine-tuned promoter/phage heads.","status":"source_checked","source_ids":["evidence-official-2efbfaba5a1c09f4f7fa","evidence-official-965600234c25fbd104c8","evidence-official-ee6e270af125ddf16eb3","evidence-official-5e268f31f0347fc5564c"],"source_locator":"ProkBERT paper Section 2.1.2 and Table 1; official ProkBERT-mini card: Model Description and licence metadata"},{"label":"Parameters","value":"20.6M for the inspected ProkBERT-mini checkpoint; do not assign that count automatically to all variants.","status":"source_checked","source_ids":["evidence-official-2efbfaba5a1c09f4f7fa","evidence-official-965600234c25fbd104c8","evidence-official-ee6e270af125ddf16eb3","evidence-official-5e268f31f0347fc5564c"],"source_locator":"ProkBERT paper Section 2.1.2 and Table 1; official ProkBERT-mini card: Model Description and licence metadata"},{"label":"Known versions","value":"ProkBERT-mini (6-mer, shift 1), mini-c (single base), mini-long (6-mer, shift 2), plus promoter/phage fine-tunes.","status":"source_checked","source_ids":["evidence-official-2efbfaba5a1c09f4f7fa","evidence-official-965600234c25fbd104c8","evidence-official-ee6e270af125ddf16eb3","evidence-official-5e268f31f0347fc5564c"],"source_locator":"ProkBERT paper Section 2.1.2 and Table 1; official ProkBERT-mini card: Model Description and licence metadata"},{"label":"Training data","value":"NCBI RefSeq genomes covering bacteria, viruses, archaea and fungi; README reports 976,878 contigs from 17,178 assemblies and 3,882 genera.","status":"source_checked","source_ids":["evidence-official-2efbfaba5a1c09f4f7fa","evidence-official-965600234c25fbd104c8","evidence-official-ee6e270af125ddf16eb3","evidence-official-5e268f31f0347fc5564c"],"source_locator":"ProkBERT paper Section 2.1.2 and Table 1; official ProkBERT-mini card: Model Description and licence metadata"},{"label":"Training cutoff","value":"NCBI RefSeq training-genome retrieval date: 6 January 2023, as reported in Methods.","status":"source_checked","source_ids":["evidence-official-2efbfaba5a1c09f4f7fa","evidence-official-965600234c25fbd104c8","evidence-official-ee6e270af125ddf16eb3","evidence-official-5e268f31f0347fc5564c"],"source_locator":"ProkBERT paper Section 2.1.2 and Table 1; official ProkBERT-mini card: Model Description and licence metadata"},{"label":"Context limits","value":"The paper reports maximum sequence lengths of 1,024bp for mini and 2,048bp for mini-long; tokenization stride distinguishes the variants.","status":"source_checked","source_ids":["evidence-official-2efbfaba5a1c09f4f7fa","evidence-official-965600234c25fbd104c8","evidence-official-ee6e270af125ddf16eb3","evidence-official-5e268f31f0347fc5564c"],"source_locator":"ProkBERT paper Section 2.1.2 and Table 1; official ProkBERT-mini card: Model Description and licence metadata"},{"label":"Weights licence","value":"The official neuralbioinfo/prokbert-mini card declares CC-BY-NC-4.0. The code repository is MIT, so code and model-weight rights differ.","status":"source_checked","source_ids":["evidence-official-2efbfaba5a1c09f4f7fa","evidence-official-965600234c25fbd104c8","evidence-official-ee6e270af125ddf16eb3","evidence-official-5e268f31f0347fc5564c"],"source_locator":"ProkBERT paper Section 2.1.2 and Table 1; official ProkBERT-mini card: Model Description and licence metadata"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/nbrg-ppcu/prokbert","status":"source_checked","source_ids":["evidence-official-2efbfaba5a1c09f4f7fa","evidence-official-965600234c25fbd104c8","evidence-official-ee6e270af125ddf16eb3","evidence-official-5e268f31f0347fc5564c"],"source_locator":"ProkBERT paper Section 2.1.2 and Table 1; official ProkBERT-mini card: Model Description and licence metadata"},{"label":"Code licence","value":"MIT","status":"source_checked","source_ids":["evidence-official-03106594cde7719aa359"],"source_locator":"LICENSE: licence text"}],"strengths":[{"text":"The released family includes character and shifted k-mer variants plus task-specific fine-tuned models.","source_ids":["evidence-official-2efbfaba5a1c09f4f7fa","evidence-official-965600234c25fbd104c8","evidence-official-ee6e270af125ddf16eb3","evidence-official-5e268f31f0347fc5564c"],"source_locator":"ProkBERT paper Section 2.1.2 and Table 1; official ProkBERT-mini card: Model Description and licence metadata"}],"limitations":[{"text":"The authors explicitly identify restricted context as a limitation in phage analysis. A promoter or phage classifier is a distinct fine-tuned configuration.","source_ids":["evidence-official-2efbfaba5a1c09f4f7fa","evidence-official-965600234c25fbd104c8","evidence-official-ee6e270af125ddf16eb3","evidence-official-5e268f31f0347fc5564c"],"source_locator":"ProkBERT paper Section 2.1.2 and Table 1; official ProkBERT-mini card: Model Description and licence metadata"}],"diagram":{"title":"ProkBERT workflow","steps":["DNA segment","Variant-specific LCA tokens","BERT encoder","Representation or task head"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-2efbfaba5a1c09f4f7fa","evidence-official-965600234c25fbd104c8","evidence-official-ee6e270af125ddf16eb3","evidence-official-5e268f31f0347fc5564c"],"source_locator":"ProkBERT paper Section 2.1.2 and Table 1; official ProkBERT-mini card: Model Description and licence metadata"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}}}
{"id":"catalog-model-proteinmpnn","kind":"model","name":"ProteinMPNN","description":"Existing candidate catalogue entry; exact checkpoint and metadata require primary-source review.","status":"discovered","facets":{"areas":["proteins-complexes"],"method_types":["specialist"]},"source_ids":["catalog-source-proteinmpnn"],"links":[{"relation":"variant_of","target_id":"discovery-model-proteinmpnn"}],"attributes":{"entity_level":"family","version":"v_48_020","reported_name":"ProteinMPNN","access":"Public code and checkpoints; requires a suitable protein structure.","method_type":"specialist","historical_missing_metadata":{"checkpoint_revision":"not_yet_extracted","training_data":"not_yet_extracted","licence":"not_yet_extracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"ProteinMPNN designs amino-acid sequences for a supplied protein backbone.","summary_source_ids":["evidence-official-ea568a52ea3ab476db6a","evidence-official-7b6fdf915d9c6950ad0d","evidence-official-019aff235ae2c3cf29d6","evidence-official-92b0ae57a020cbc40fe7","evidence-official-0b8003691827eb09b0a4"],"summary_source_locator":"ProteinMPNN paper: main-text model development; training/README.md: multi-chain training set, list.csv and cluster manifests; protein_mpnn_utils.py: ProteinMPNN; Supplementary Materials: Methods for training multi-chain models / Training data","sections":[{"title":"How it works","body":"ProteinMPNN converts a supplied backbone into a graph whose edges encode interatomic distances. Message-passing layers update node and edge features, and an autoregressive decoder samples amino acids while conditioning on the backbone and previously assigned residues. Fixed residues, tied positions and chain choices change the design task and must accompany its result.","source_ids":["evidence-official-ea568a52ea3ab476db6a","evidence-official-7b6fdf915d9c6950ad0d","evidence-official-019aff235ae2c3cf29d6","evidence-official-92b0ae57a020cbc40fe7","evidence-official-0b8003691827eb09b0a4"],"source_locator":"ProteinMPNN paper: main-text model development; training/README.md: multi-chain training set, list.csv and cluster manifests; protein_mpnn_utils.py: ProteinMPNN; Supplementary Materials: Methods for training multi-chain models / Training data"},{"title":"Versions and reproducibility","body":"v_48_002, v_48_010, v_48_020 and v_48_030; distinct soluble and C-alpha-only weights. Structure-size and memory dependent. README --max_length is an implementation guard, not a validated scientific context limit.","source_ids":["evidence-official-ea568a52ea3ab476db6a","evidence-official-7b6fdf915d9c6950ad0d","evidence-official-019aff235ae2c3cf29d6","evidence-official-92b0ae57a020cbc40fe7","evidence-official-0b8003691827eb09b0a4"],"source_locator":"ProteinMPNN paper: main-text model development; training/README.md: multi-chain training set, list.csv and cluster manifests; protein_mpnn_utils.py: ProteinMPNN; Supplementary Materials: Methods for training multi-chain models / Training data"}],"facts":[{"label":"Model type","value":"Structure-conditioned message-passing sequence design model","status":"source_checked","source_ids":["evidence-official-ea568a52ea3ab476db6a","evidence-official-7b6fdf915d9c6950ad0d","evidence-official-019aff235ae2c3cf29d6","evidence-official-92b0ae57a020cbc40fe7","evidence-official-0b8003691827eb09b0a4"],"source_locator":"ProteinMPNN paper: main-text model development; training/README.md: multi-chain training set, list.csv and cluster manifests; protein_mpnn_utils.py: ProteinMPNN; Supplementary Materials: Methods for training multi-chain models / Training data"},{"label":"Architecture","value":"Message-passing encoder-decoder with structural interatomic-distance features and edge updates; sequences are sampled with the configured autoregressive decoding procedure.","status":"source_checked","source_ids":["evidence-official-ea568a52ea3ab476db6a","evidence-official-7b6fdf915d9c6950ad0d","evidence-official-019aff235ae2c3cf29d6","evidence-official-92b0ae57a020cbc40fe7","evidence-official-0b8003691827eb09b0a4"],"source_locator":"ProteinMPNN paper: main-text model development; training/README.md: multi-chain training set, list.csv and cluster manifests; protein_mpnn_utils.py: ProteinMPNN; Supplementary Materials: Methods for training multi-chain models / Training data"},{"label":"Inputs","value":"Protein backbone coordinates, with optional fixed residues, chain choices, tied positions and amino-acid constraints.","status":"source_checked","source_ids":["evidence-official-ea568a52ea3ab476db6a","evidence-official-7b6fdf915d9c6950ad0d","evidence-official-019aff235ae2c3cf29d6","evidence-official-92b0ae57a020cbc40fe7","evidence-official-0b8003691827eb09b0a4"],"source_locator":"ProteinMPNN paper: main-text model development; training/README.md: multi-chain training set, list.csv and cluster manifests; protein_mpnn_utils.py: ProteinMPNN; Supplementary Materials: Methods for training multi-chain models / Training data"},{"label":"Outputs","value":"Designed sequences, sequence scores and conditional amino-acid probabilities.","status":"source_checked","source_ids":["evidence-official-ea568a52ea3ab476db6a","evidence-official-7b6fdf915d9c6950ad0d","evidence-official-019aff235ae2c3cf29d6","evidence-official-92b0ae57a020cbc40fe7","evidence-official-0b8003691827eb09b0a4"],"source_locator":"ProteinMPNN paper: main-text model development; training/README.md: multi-chain training set, list.csv and cluster manifests; protein_mpnn_utils.py: ProteinMPNN; Supplementary Materials: Methods for training multi-chain models / Training data"},{"label":"Parameters","value":"The inspected model implementation is configured through encoder/decoder depth and feature width. The paper and training README do not state an exact total for every released checkpoint.","status":"unreported","source_ids":["evidence-official-ea568a52ea3ab476db6a","evidence-official-7b6fdf915d9c6950ad0d","evidence-official-019aff235ae2c3cf29d6","evidence-official-92b0ae57a020cbc40fe7","evidence-official-0b8003691827eb09b0a4"],"source_locator":"ProteinMPNN paper: main-text model development; training/README.md: multi-chain training set, list.csv and cluster manifests; protein_mpnn_utils.py: ProteinMPNN; Supplementary Materials: Methods for training multi-chain models / Training data"},{"label":"Known versions","value":"v_48_002, v_48_010, v_48_020 and v_48_030; distinct soluble and C-alpha-only weights.","status":"source_checked","source_ids":["evidence-official-ea568a52ea3ab476db6a","evidence-official-7b6fdf915d9c6950ad0d","evidence-official-019aff235ae2c3cf29d6","evidence-official-92b0ae57a020cbc40fe7","evidence-official-0b8003691827eb09b0a4"],"source_locator":"ProteinMPNN paper: main-text model development; training/README.md: multi-chain training set, list.csv and cluster manifests; protein_mpnn_utils.py: ProteinMPNN; Supplementary Materials: Methods for training multi-chain models / Training data"},{"label":"Training data","value":"Released multi-chain training set of PDB biological units, with chain metadata and validation/test cluster manifests. The documented set is dated 2 August 2021.","status":"source_checked","source_ids":["evidence-official-ea568a52ea3ab476db6a","evidence-official-7b6fdf915d9c6950ad0d","evidence-official-019aff235ae2c3cf29d6","evidence-official-92b0ae57a020cbc40fe7","evidence-official-0b8003691827eb09b0a4"],"source_locator":"ProteinMPNN paper: main-text model development; training/README.md: multi-chain training set, list.csv and cluster manifests; protein_mpnn_utils.py: ProteinMPNN; Supplementary Materials: Methods for training multi-chain models / Training data"},{"label":"Training cutoff","value":"The released PDB training-set snapshot is dated 2021-08-02; preserve its chain-level deposition metadata and cluster split for a run.","status":"source_checked","source_ids":["evidence-official-ea568a52ea3ab476db6a","evidence-official-7b6fdf915d9c6950ad0d","evidence-official-019aff235ae2c3cf29d6","evidence-official-92b0ae57a020cbc40fe7","evidence-official-0b8003691827eb09b0a4"],"source_locator":"ProteinMPNN paper: main-text model development; training/README.md: multi-chain training set, list.csv and cluster manifests; protein_mpnn_utils.py: ProteinMPNN; Supplementary Materials: Methods for training multi-chain models / Training data"},{"label":"Context limits","value":"Structure-size and memory dependent. README --max_length is an implementation guard, not a validated scientific context limit.","status":"source_checked","source_ids":["evidence-official-ea568a52ea3ab476db6a","evidence-official-7b6fdf915d9c6950ad0d","evidence-official-019aff235ae2c3cf29d6","evidence-official-92b0ae57a020cbc40fe7","evidence-official-0b8003691827eb09b0a4"],"source_locator":"ProteinMPNN paper: main-text model development; training/README.md: multi-chain training set, list.csv and cluster manifests; protein_mpnn_utils.py: ProteinMPNN; Supplementary Materials: Methods for training multi-chain models / Training data"},{"label":"Weights licence","value":"Separate checkpoint-distribution terms are not stated in the inspected release documentation and licence material. The source-code licence alone is not recorded as an explicit weight grant.","status":"unreported","source_ids":["evidence-official-ea568a52ea3ab476db6a","evidence-official-7b6fdf915d9c6950ad0d","evidence-official-019aff235ae2c3cf29d6","evidence-official-92b0ae57a020cbc40fe7","evidence-official-0b8003691827eb09b0a4","evidence-official-eebe7c91156963e6ddc0"],"source_locator":"ProteinMPNN paper: main-text model development; training/README.md: multi-chain training set, list.csv and cluster manifests; protein_mpnn_utils.py: ProteinMPNN; Supplementary Materials: Methods for training multi-chain models / Training data; LICENSE: licence text"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/dauparas/ProteinMPNN","status":"source_checked","source_ids":["evidence-official-ea568a52ea3ab476db6a","evidence-official-7b6fdf915d9c6950ad0d","evidence-official-019aff235ae2c3cf29d6","evidence-official-92b0ae57a020cbc40fe7","evidence-official-0b8003691827eb09b0a4"],"source_locator":"ProteinMPNN paper: main-text model development; training/README.md: multi-chain training set, list.csv and cluster manifests; protein_mpnn_utils.py: ProteinMPNN; Supplementary Materials: Methods for training multi-chain models / Training data"},{"label":"Code licence","value":"MIT","status":"source_checked","source_ids":["evidence-official-eebe7c91156963e6ddc0"],"source_locator":"LICENSE: licence text"}],"strengths":[{"text":"The interface makes design constraints explicit and includes full-backbone, C-alpha-only and soluble-protein weight sets.","source_ids":["evidence-official-ea568a52ea3ab476db6a","evidence-official-7b6fdf915d9c6950ad0d","evidence-official-019aff235ae2c3cf29d6","evidence-official-92b0ae57a020cbc40fe7","evidence-official-0b8003691827eb09b0a4"],"source_locator":"ProteinMPNN paper: main-text model development; training/README.md: multi-chain training set, list.csv and cluster manifests; protein_mpnn_utils.py: ProteinMPNN; Supplementary Materials: Methods for training multi-chain models / Training data"}],"limitations":[{"text":"The requested backbone and constraints are part of the evaluated problem. A command-line maximum-length guard is not evidence that designs at that size have been validated.","source_ids":["evidence-official-ea568a52ea3ab476db6a","evidence-official-7b6fdf915d9c6950ad0d","evidence-official-019aff235ae2c3cf29d6","evidence-official-92b0ae57a020cbc40fe7","evidence-official-0b8003691827eb09b0a4"],"source_locator":"ProteinMPNN paper: main-text model development; training/README.md: multi-chain training set, list.csv and cluster manifests; protein_mpnn_utils.py: ProteinMPNN; Supplementary Materials: Methods for training multi-chain models / Training data"}],"diagram":{"title":"ProteinMPNN workflow","steps":["Protein backbone","Structural graph features","Message-passing model","Constrained sequence sampling"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-ea568a52ea3ab476db6a","evidence-official-7b6fdf915d9c6950ad0d","evidence-official-019aff235ae2c3cf29d6","evidence-official-92b0ae57a020cbc40fe7","evidence-official-0b8003691827eb09b0a4"],"source_locator":"ProteinMPNN paper: main-text model development; training/README.md: multi-chain training set, list.csv and cluster manifests; protein_mpnn_utils.py: ProteinMPNN; Supplementary Materials: Methods for training multi-chain models / Training data"},"coverage":"limited","gaps":["Parameters: The inspected model implementation is configured through encoder/decoder depth and feature width. The paper and training README do not state an exact total for every released checkpoint.","Weights licence: Separate checkpoint-distribution terms are not stated in the inspected release documentation and licence material. The source-code licence alone is not recorded as an explicit weight grant."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}}}
{"id":"catalog-model-rhofold","kind":"model","name":"RhoFold+","description":"Existing candidate catalogue entry; exact checkpoint and metadata require primary-source review.","status":"discovered","facets":{"areas":["rna-transcriptomes"],"method_types":["specialist"]},"source_ids":["catalog-source-rhofold"],"links":[],"attributes":{"entity_level":"family","version":"pretrained","reported_name":"RhoFold+","access":"Public code and checkpoint instructions.","method_type":"specialist","historical_missing_metadata":{"checkpoint_revision":"not_yet_extracted","training_data":"not_yet_extracted","licence":"not_yet_extracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"RhoFold+ predicts RNA three-dimensional structures from RNA sequence with alignment information.","summary_source_ids":["evidence-official-955922130ff85a74fa18","evidence-official-8b4e1bd487c21cda9862","evidence-official-3e1890c3652bc5c150f7"],"summary_source_locator":"Paper: Automated end-to-end platform, Large-scale pretraining dataset, Efficient development of a self-distillation dataset, Feature processing with Rhoformer and Data availability; README.md: Usage; official cuhkaih/rhofold card licence metadata","sections":[{"title":"How it works","body":"RhoFold+ predicts RNA three-dimensional structures from RNA sequence with alignment information. RNA-FM embeddings and MSA representations enter the Rhoformer transformer stack; a geometry-aware invariant-point-attention structure module predicts frames and torsion angles with recycling. The documented inputs are RNA FASTA (A, U, G, C), optionally a supplied MSA; an MSA is otherwise generated by the full workflow. The output consists of three-dimensional PDB models, predicted distograms, secondary structure and per-residue confidence.","source_ids":["evidence-official-955922130ff85a74fa18","evidence-official-8b4e1bd487c21cda9862","evidence-official-3e1890c3652bc5c150f7"],"source_locator":"Paper: Automated end-to-end platform, Large-scale pretraining dataset, Efficient development of a self-distillation dataset, Feature processing with Rhoformer and Data availability; README.md: Usage; official cuhkaih/rhofold card licence metadata"},{"title":"Versions and reproducibility","body":"RhoFold+; the README links a pretrained checkpoint and the 2024 Nature Methods paper. The paper limits MSA depth to256 sequences during training and default inference. This is alignment depth, not an RNA-length maximum; the latter remains unextracted.","source_ids":["evidence-official-955922130ff85a74fa18","evidence-official-8b4e1bd487c21cda9862","evidence-official-3e1890c3652bc5c150f7"],"source_locator":"Paper: Automated end-to-end platform, Large-scale pretraining dataset, Efficient development of a self-distillation dataset, Feature processing with Rhoformer and Data availability; README.md: Usage; official cuhkaih/rhofold card licence metadata"}],"facts":[{"label":"Model type","value":"RNA structure predictor with language-model and MSA inputs","status":"source_checked","source_ids":["evidence-official-955922130ff85a74fa18","evidence-official-8b4e1bd487c21cda9862","evidence-official-3e1890c3652bc5c150f7"],"source_locator":"Paper: Automated end-to-end platform, Large-scale pretraining dataset, Efficient development of a self-distillation dataset, Feature processing with Rhoformer and Data availability; README.md: Usage; official cuhkaih/rhofold card licence metadata"},{"label":"Architecture","value":"RNA-FM embeddings and MSA representations enter the Rhoformer transformer stack; a geometry-aware invariant-point-attention structure module predicts frames and torsion angles with recycling.","status":"source_checked","source_ids":["evidence-official-955922130ff85a74fa18","evidence-official-8b4e1bd487c21cda9862","evidence-official-3e1890c3652bc5c150f7"],"source_locator":"Paper: Automated end-to-end platform, Large-scale pretraining dataset, Efficient development of a self-distillation dataset, Feature processing with Rhoformer and Data availability; README.md: Usage; official cuhkaih/rhofold card licence metadata"},{"label":"Inputs","value":"RNA FASTA (A, U, G, C), optionally a supplied MSA; an MSA is otherwise generated by the full workflow.","status":"source_checked","source_ids":["evidence-official-955922130ff85a74fa18","evidence-official-8b4e1bd487c21cda9862","evidence-official-3e1890c3652bc5c150f7"],"source_locator":"Paper: Automated end-to-end platform, Large-scale pretraining dataset, Efficient development of a self-distillation dataset, Feature processing with Rhoformer and Data availability; README.md: Usage; official cuhkaih/rhofold card licence metadata"},{"label":"Outputs","value":"Three-dimensional PDB models, predicted distograms, secondary structure and per-residue confidence.","status":"source_checked","source_ids":["evidence-official-955922130ff85a74fa18","evidence-official-8b4e1bd487c21cda9862","evidence-official-3e1890c3652bc5c150f7"],"source_locator":"Paper: Automated end-to-end platform, Large-scale pretraining dataset, Efficient development of a self-distillation dataset, Feature processing with Rhoformer and Data availability; README.md: Usage; official cuhkaih/rhofold card licence metadata"},{"label":"Parameters","value":"The inspected paper describes RNA-FM, Rhoformer and structure modules without stating a total for the complete selected predictor in those architecture sections.","status":"unreported","source_ids":["evidence-official-955922130ff85a74fa18","evidence-official-8b4e1bd487c21cda9862","evidence-official-3e1890c3652bc5c150f7"],"source_locator":"Paper: Automated end-to-end platform, Large-scale pretraining dataset, Efficient development of a self-distillation dataset, Feature processing with Rhoformer and Data availability; README.md: Usage; official cuhkaih/rhofold card licence metadata"},{"label":"Known versions","value":"RhoFold+; the README links a pretrained checkpoint and the 2024 Nature Methods paper.","status":"source_checked","source_ids":["evidence-official-955922130ff85a74fa18","evidence-official-8b4e1bd487c21cda9862","evidence-official-3e1890c3652bc5c150f7"],"source_locator":"Paper: Automated end-to-end platform, Large-scale pretraining dataset, Efficient development of a self-distillation dataset, Feature processing with Rhoformer and Data availability; README.md: Usage; official cuhkaih/rhofold card licence metadata"},{"label":"Training data","value":"RNA-FM pretraining uses RNAcentral100. Structure training uses PDB RNA chains selected through BGSU representative sets, with additional self-distillation from RNAStralign/bpRNA-derived sequences.","status":"source_checked","source_ids":["evidence-official-955922130ff85a74fa18","evidence-official-8b4e1bd487c21cda9862","evidence-official-3e1890c3652bc5c150f7"],"source_locator":"Paper: Automated end-to-end platform, Large-scale pretraining dataset, Efficient development of a self-distillation dataset, Feature processing with Rhoformer and Data availability; README.md: Usage; official cuhkaih/rhofold card licence metadata"},{"label":"Training cutoff","value":"Structural-data selection uses the BGSU representative set dated 2022-04-13. This does not establish a single cutoff for every RNA-FM or MSA resource.","status":"source_checked","source_ids":["evidence-official-955922130ff85a74fa18","evidence-official-8b4e1bd487c21cda9862","evidence-official-3e1890c3652bc5c150f7"],"source_locator":"Paper: Automated end-to-end platform, Large-scale pretraining dataset, Efficient development of a self-distillation dataset, Feature processing with Rhoformer and Data availability; README.md: Usage; official cuhkaih/rhofold card licence metadata"},{"label":"Context limits","value":"MSA depth is capped at 256 during documented training and default inference. The reviewed paper does not establish a single RNA-length maximum; MSA depth is a different quantity.","status":"unreported","source_ids":["evidence-official-955922130ff85a74fa18","evidence-official-8b4e1bd487c21cda9862","evidence-official-3e1890c3652bc5c150f7"],"source_locator":"Paper: Automated end-to-end platform, Large-scale pretraining dataset, Efficient development of a self-distillation dataset, Feature processing with Rhoformer and Data availability; README.md: Usage; official cuhkaih/rhofold card licence metadata"},{"label":"Weights licence","value":"Apache-2.0 declared in the official cuhkaih/rhofold model-card metadata; this is distinct from access conditions for training data.","status":"source_checked","source_ids":["evidence-official-955922130ff85a74fa18","evidence-official-8b4e1bd487c21cda9862","evidence-official-3e1890c3652bc5c150f7"],"source_locator":"Paper: Automated end-to-end platform, Large-scale pretraining dataset, Efficient development of a self-distillation dataset, Feature processing with Rhoformer and Data availability; README.md: Usage; official cuhkaih/rhofold card licence metadata"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/ml4bio/RhoFold","status":"source_checked","source_ids":["evidence-official-955922130ff85a74fa18","evidence-official-8b4e1bd487c21cda9862","evidence-official-3e1890c3652bc5c150f7"],"source_locator":"Paper: Automated end-to-end platform, Large-scale pretraining dataset, Efficient development of a self-distillation dataset, Feature processing with Rhoformer and Data availability; README.md: Usage; official cuhkaih/rhofold card licence metadata"},{"label":"Code licence","value":"Apache-2.0","status":"source_checked","source_ids":["evidence-official-4e0280294c1215534077"],"source_locator":"LICENSE: licence text"}],"strengths":[{"text":"Produces secondary and tertiary outputs and supports provided or automatically generated alignments.","source_ids":["evidence-official-955922130ff85a74fa18","evidence-official-8b4e1bd487c21cda9862","evidence-official-3e1890c3652bc5c150f7"],"source_locator":"Paper: Automated end-to-end platform, Large-scale pretraining dataset, Efficient development of a self-distillation dataset, Feature processing with Rhoformer and Data availability; README.md: Usage; official cuhkaih/rhofold card licence metadata"}],"limitations":[{"text":"The README labels sequence-only inference as a lower-accuracy testing mode. Constructing the full MSA databases needs substantial local storage; macOS is not supported by its documented setup.","source_ids":["evidence-official-955922130ff85a74fa18","evidence-official-8b4e1bd487c21cda9862","evidence-official-3e1890c3652bc5c150f7"],"source_locator":"Paper: Automated end-to-end platform, Large-scale pretraining dataset, Efficient development of a self-distillation dataset, Feature processing with Rhoformer and Data availability; README.md: Usage; official cuhkaih/rhofold card licence metadata"}],"diagram":{"title":"RhoFold+ workflow","steps":["RNA sequence and MSA","RhoFold+ prediction","Structure and confidence","Optional relaxation"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-955922130ff85a74fa18","evidence-official-8b4e1bd487c21cda9862","evidence-official-3e1890c3652bc5c150f7"],"source_locator":"Paper: Automated end-to-end platform, Large-scale pretraining dataset, Efficient development of a self-distillation dataset, Feature processing with Rhoformer and Data availability; README.md: Usage; official cuhkaih/rhofold card licence metadata"},"coverage":"limited","gaps":["Parameters: The inspected paper describes RNA-FM, Rhoformer and structure modules without stating a total for the complete selected predictor in those architecture sections.","Context limits: MSA depth is capped at 256 during documented training and default inference. The reviewed paper does not establish a single RNA-length maximum; MSA depth is a different quantity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}}}
{"id":"catalog-model-rna-fm","kind":"model","name":"RNA-FM","description":"Existing candidate catalogue entry; exact checkpoint and metadata require primary-source review.","status":"discovered","facets":{"areas":["rna-transcriptomes"],"method_types":["foundation model"]},"source_ids":["catalog-source-rna-fm"],"links":[{"relation":"family","target_id":"discovery-model-rna-fm"}],"attributes":{"entity_level":"family","version":"ncRNA","reported_name":"RNA-FM","access":"Public code and checkpoint instructions.","method_type":"foundation model","historical_missing_metadata":{"checkpoint_revision":"not_yet_extracted","training_data":"not_yet_extracted","licence":"not_yet_extracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"RNA-FM learns contextual representations of RNA nucleotides for downstream RNA analyses.","summary_source_ids":["evidence-official-fd8e332abdf04a75195b","evidence-official-e17e3864c464d981afc7"],"summary_source_locator":"README.md: Quick Start, embedding examples and RNA foundation model comparison table","sections":[{"title":"How it works","body":"RNA-FM converts an RNA sequence into one token per nucleotide. Twelve transformer encoder blocks use self-attention to produce a 640-dimensional representation at each position. During pretraining, the model learns to recover masked nucleotides from their surrounding sequence. The resulting representations can be supplied to a separately specified downstream model; they are not, by themselves, a structure or functional prediction.","source_ids":["evidence-final-model-rna-fm-paper"],"source_locator":"arXiv:2204.00300v5, Methods: ncRNA data collection and preprocessing and RNA foundation model training details (p.22)"},{"title":"Versions and reproducibility","body":"The nucleotide-based rna_fm_t12 and codon-based mrna_fm_t12 interfaces are distinct. The original RNA-FM paper sets a training input-length limit of 1,024 and describes a usable limit of 1,022 nucleotides. That limit should not be assigned to mRNA-FM without checking its separate configuration.","source_ids":["evidence-official-fd8e332abdf04a75195b","evidence-official-e17e3864c464d981afc7","evidence-final-model-rna-fm-paper"],"source_locator":"Official README: Quick Start and RNA-FM/mRNA-FM examples; arXiv:2204.00300v5, Methods: training input length and SARS-CoV-2 genome embedding extraction (pp.22–23)"}],"facts":[{"label":"Model type","value":"Masked-token RNA transformer encoder","status":"source_checked","source_ids":["evidence-official-fd8e332abdf04a75195b","evidence-official-e17e3864c464d981afc7"],"source_locator":"README.md: Quick Start, embedding examples and RNA foundation model comparison table"},{"label":"Architecture","value":"12-layer masked-token transformer encoder with hidden width 640 and 20 attention heads; nucleotide tokens produce contextual representations.","status":"source_checked","source_ids":["evidence-final-model-rna-fm-paper"],"source_locator":"arXiv:2204.00300v5, Methods: ncRNA data collection and preprocessing and RNA foundation model training details (p.22)"},{"label":"Inputs","value":"RNA sequences tokenized at nucleotide resolution.","status":"source_checked","source_ids":["evidence-official-fd8e332abdf04a75195b","evidence-official-e17e3864c464d981afc7"],"source_locator":"README.md: Quick Start, embedding examples and RNA foundation model comparison table"},{"label":"Outputs","value":"Contextual token embeddings for a specified downstream RNA task.","status":"source_checked","source_ids":["evidence-official-fd8e332abdf04a75195b","evidence-official-e17e3864c464d981afc7"],"source_locator":"README.md: Quick Start, embedding examples and RNA foundation model comparison table"},{"label":"Parameters","value":"99M, as printed in the official Foundation Models table.","status":"source_checked","source_ids":["evidence-official-fd8e332abdf04a75195b","evidence-official-e17e3864c464d981afc7"],"source_locator":"README.md: Quick Start, embedding examples and RNA foundation model comparison table"},{"label":"Known versions","value":"rna_fm_t12 and mrna_fm_t12 are separate pretrained interfaces.","status":"source_checked","source_ids":["evidence-official-fd8e332abdf04a75195b","evidence-official-e17e3864c464d981afc7"],"source_locator":"README.md: Quick Start, embedding examples and RNA foundation model comparison table"},{"label":"Training data","value":"23.7 million non-coding RNA sequences collected from RNAcentral. The authors replace T with U and remove identical sequences using CD-HIT-EST at 100% identity, naming the resulting corpus RNAcentral100.","status":"source_checked","source_ids":["evidence-final-model-rna-fm-paper"],"source_locator":"arXiv:2204.00300v5, Methods: ncRNA data collection and preprocessing and RNA foundation model training details (p.22)"},{"label":"Training cutoff","value":"The inspected Methods and official README do not establish an exact dated RNAcentral release. RNAcentral100 is the authors’ processed-corpus label, not a verified release number.","status":"unreported","source_ids":["evidence-final-model-rna-fm-paper","evidence-official-fd8e332abdf04a75195b"],"source_locator":"arXiv:2204.00300v5, Methods: ncRNA data collection and preprocessing (p.22); official README: Foundation Models table"},{"label":"Context limits","value":"The original paper sets a training input-length limit of 1,024 and describes a usable input limit of 1,022 nucleotides. These are the original RNA-FM settings, not a validated limit for later codon-based mRNA-FM checkpoints.","status":"source_checked","source_ids":["evidence-final-model-rna-fm-paper"],"source_locator":"arXiv:2204.00300v5, Methods: RNA foundation model training details (pp.22–23) and RNA-FM application input-limit statement (p.23)"},{"label":"Weights licence","value":"Separate checkpoint-distribution terms are not stated in the inspected release documentation and licence material. The source-code licence alone is not recorded as an explicit weight grant.","status":"unreported","source_ids":["evidence-official-fd8e332abdf04a75195b","evidence-official-e17e3864c464d981afc7","evidence-official-3fde3df73e79e455bd86"],"source_locator":"README.md: Quick Start, embedding examples and RNA foundation model comparison table; LICENSE: licence text"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/ml4bio/RNA-FM","status":"source_checked","source_ids":["evidence-official-fd8e332abdf04a75195b","evidence-official-e17e3864c464d981afc7"],"source_locator":"README.md: Quick Start, embedding examples and RNA foundation model comparison table"},{"label":"Code licence","value":"MIT","status":"source_checked","source_ids":["evidence-official-3fde3df73e79e455bd86"],"source_locator":"LICENSE: licence text"}],"strengths":[{"text":"The repository exposes embedding extraction and examples for downstream RNA analyses.","source_ids":["evidence-official-fd8e332abdf04a75195b","evidence-official-e17e3864c464d981afc7"],"source_locator":"README.md: Quick Start, embedding examples and RNA foundation model comparison table"}],"limitations":[{"text":"Base-level RNA-FM and codon-level mRNA-FM are not interchangeable. The task head and tokenization must be specified in each evaluation.","source_ids":["evidence-official-fd8e332abdf04a75195b","evidence-official-e17e3864c464d981afc7"],"source_locator":"README.md: Quick Start, embedding examples and RNA foundation model comparison table"}],"diagram":{"title":"RNA-FM workflow","steps":["RNA sequence","Nucleotide tokenizer","12-layer transformer encoder","Contextual nucleotide representations"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-fd8e332abdf04a75195b","evidence-official-e17e3864c464d981afc7"],"source_locator":"README.md: Quick Start, embedding examples and RNA foundation model comparison table"},"coverage":"limited","gaps":["Weights licence: Separate checkpoint-distribution terms are not stated in the inspected release documentation and licence material. The source-code licence alone is not recorded as an explicit weight grant.","Training cutoff: An exact dated RNAcentral release is not established by the inspected original Methods or official README."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied. Follow-up retrieved the complete original RNA-FM PDF and inspected the full pinned Geneformer repository inventory; unavailable labels were updated only where new evidence resolved the earlier retrieval gap. Follow-up audit reconciles the narrative with the verified RNA-FM input limit and clarifies the original paper’s RNAcentral100 preprocessing definition; mRNA-FM remains separate."}}}}
{"id":"catalog-model-scfoundation","kind":"model","name":"scFoundation","description":"Existing candidate catalogue entry; exact checkpoint and metadata require primary-source review.","status":"discovered","facets":{"areas":["cells-tissues"],"method_types":["foundation model"]},"source_ids":["catalog-source-scfoundation"],"links":[],"attributes":{"entity_level":"family","version":"100M","reported_name":"scFoundation","access":"Public code; model weights have separate terms that must be checked.","method_type":"foundation model","historical_missing_metadata":{"checkpoint_revision":"not_yet_extracted","training_data":"not_yet_extracted","licence":"not_yet_extracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"scFoundation produces contextual cell and gene representations from gene-expression measurements.","summary_source_ids":["evidence-official-bfbc8babf3630ddb3ed7","evidence-official-7972ce2bdd6a5e40d8ad","evidence-official-9863509ad10b824dc96b"],"summary_source_locator":"June 15 2023 scFoundation manuscript: Results pre-training framework; Methods Data collection, Embedding, Encoder, Decoder and Read-depth-aware pre-training; official model README","sections":[{"title":"How it works","body":"scFoundation embeds gene identity and continuous expression together with source and target read-depth indicators. Its encoder processes only nonzero, unmasked genes. Those contextual embeddings are combined with zero and mask embeddings before a Performer decoder predicts expression across the full gene vocabulary. Pooled encoder outputs represent cells; decoder outputs provide gene-level context.","source_ids":["evidence-official-bfbc8babf3630ddb3ed7","evidence-official-7972ce2bdd6a5e40d8ad","evidence-official-9863509ad10b824dc96b"],"source_locator":"June 15 2023 scFoundation manuscript: Results pre-training framework; Methods Data collection, Embedding, Encoder, Decoder and Read-depth-aware pre-training; official model README"},{"title":"Versions and reproducibility","body":"scFoundation / xTrimoscFoundation-alpha; the repository exposes separately configured embedding, enhancement and downstream prediction workflows. Fixed input gene vocabulary of 19,264 genes; not a nucleotide token context.","source_ids":["evidence-official-bfbc8babf3630ddb3ed7","evidence-official-7972ce2bdd6a5e40d8ad","evidence-official-9863509ad10b824dc96b"],"source_locator":"June 15 2023 scFoundation manuscript: Results pre-training framework; Methods Data collection, Embedding, Encoder, Decoder and Read-depth-aware pre-training; official model README"}],"facts":[{"label":"Model type","value":"Transcriptomic representation model","status":"source_checked","source_ids":["evidence-official-bfbc8babf3630ddb3ed7","evidence-official-7972ce2bdd6a5e40d8ad","evidence-official-9863509ad10b824dc96b"],"source_locator":"June 15 2023 scFoundation manuscript: Results pre-training framework; Methods Data collection, Embedding, Encoder, Decoder and Read-depth-aware pre-training; official model README"},{"label":"Architecture","value":"An asymmetric transformer encoder-decoder: learned continuous-expression embeddings enter a transformer encoder for nonzero, unmasked genes, then a Performer decoder reconstructs the full gene set.","status":"source_checked","source_ids":["evidence-official-bfbc8babf3630ddb3ed7","evidence-official-7972ce2bdd6a5e40d8ad","evidence-official-9863509ad10b824dc96b"],"source_locator":"June 15 2023 scFoundation manuscript: Results pre-training framework; Methods Data collection, Embedding, Encoder, Decoder and Read-depth-aware pre-training; official model README"},{"label":"Inputs","value":"Single-cell or bulk expression aligned to the documented 19,264-gene vocabulary.","status":"source_checked","source_ids":["evidence-official-bfbc8babf3630ddb3ed7","evidence-official-7972ce2bdd6a5e40d8ad","evidence-official-9863509ad10b824dc96b"],"source_locator":"June 15 2023 scFoundation manuscript: Results pre-training framework; Methods Data collection, Embedding, Encoder, Decoder and Read-depth-aware pre-training; official model README"},{"label":"Outputs","value":"Cell embeddings, contextual gene embeddings and outputs of separately configured downstream models.","status":"source_checked","source_ids":["evidence-official-bfbc8babf3630ddb3ed7","evidence-official-7972ce2bdd6a5e40d8ad","evidence-official-9863509ad10b824dc96b"],"source_locator":"June 15 2023 scFoundation manuscript: Results pre-training framework; Methods Data collection, Embedding, Encoder, Decoder and Read-depth-aware pre-training; official model README"},{"label":"Parameters","value":"100 million.","status":"source_checked","source_ids":["evidence-official-bfbc8babf3630ddb3ed7","evidence-official-7972ce2bdd6a5e40d8ad","evidence-official-9863509ad10b824dc96b"],"source_locator":"June 15 2023 scFoundation manuscript: Results pre-training framework; Methods Data collection, Embedding, Encoder, Decoder and Read-depth-aware pre-training; official model README"},{"label":"Known versions","value":"scFoundation / xTrimoscFoundation-alpha; the repository exposes separately configured embedding, enhancement and downstream prediction workflows.","status":"source_checked","source_ids":["evidence-official-bfbc8babf3630ddb3ed7","evidence-official-7972ce2bdd6a5e40d8ad","evidence-official-9863509ad10b824dc96b"],"source_locator":"June 15 2023 scFoundation manuscript: Results pre-training framework; Methods Data collection, Embedding, Encoder, Decoder and Read-depth-aware pre-training; official model README"},{"label":"Training data","value":"The June 2023 manuscript describes more than 50M human cells collected from GEO, Single Cell Portal, HCA and EMBL-EBI, aligned to 19,264 genes after quality control.","status":"source_checked","source_ids":["evidence-official-bfbc8babf3630ddb3ed7","evidence-official-7972ce2bdd6a5e40d8ad","evidence-official-9863509ad10b824dc96b"],"source_locator":"June 15 2023 scFoundation manuscript: Results pre-training framework; Methods Data collection, Embedding, Encoder, Decoder and Read-depth-aware pre-training; official model README"},{"label":"Training cutoff","value":"The June 2023 manuscript lists GEO, Single Cell Portal, HCA and EMBL-EBI as collection sources; its data-collection section does not give a shared last-included-study date.","status":"unreported","source_ids":["evidence-official-bfbc8babf3630ddb3ed7","evidence-official-7972ce2bdd6a5e40d8ad","evidence-official-9863509ad10b824dc96b"],"source_locator":"June 15 2023 scFoundation manuscript: Results pre-training framework; Methods Data collection, Embedding, Encoder, Decoder and Read-depth-aware pre-training; official model README"},{"label":"Context limits","value":"Fixed input gene vocabulary of 19,264 genes; not a nucleotide token context.","status":"source_checked","source_ids":["evidence-official-bfbc8babf3630ddb3ed7","evidence-official-7972ce2bdd6a5e40d8ad","evidence-official-9863509ad10b824dc96b"],"source_locator":"June 15 2023 scFoundation manuscript: Results pre-training framework; Methods Data collection, Embedding, Encoder, Decoder and Read-depth-aware pre-training; official model README"},{"label":"Weights licence","value":"Separate Model License; the Apache source-code notice explicitly excludes model-weight rights.","status":"source_checked","source_ids":["evidence-official-bfbc8babf3630ddb3ed7","evidence-official-7972ce2bdd6a5e40d8ad","evidence-official-9863509ad10b824dc96b"],"source_locator":"June 15 2023 scFoundation manuscript: Results pre-training framework; Methods Data collection, Embedding, Encoder, Decoder and Read-depth-aware pre-training; official model README"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/biomap-research/scFoundation","status":"source_checked","source_ids":["evidence-official-bfbc8babf3630ddb3ed7","evidence-official-7972ce2bdd6a5e40d8ad","evidence-official-9863509ad10b824dc96b"],"source_locator":"June 15 2023 scFoundation manuscript: Results pre-training framework; Methods Data collection, Embedding, Encoder, Decoder and Read-depth-aware pre-training; official model README"},{"label":"Code licence","value":"Apache-2.0","status":"source_checked","source_ids":["evidence-official-6189052c702a948da02d"],"source_locator":"LICENSE: licence text"}],"strengths":[{"text":"The released workflow includes embedding extraction, read-depth enhancement and integration with downstream predictors.","source_ids":["evidence-official-bfbc8babf3630ddb3ed7","evidence-official-7972ce2bdd6a5e40d8ad","evidence-official-9863509ad10b824dc96b"],"source_locator":"June 15 2023 scFoundation manuscript: Results pre-training framework; Methods Data collection, Embedding, Encoder, Decoder and Read-depth-aware pre-training; official model README"}],"limitations":[{"text":"Gene identifiers must be aligned to the supplied gene index. Source-code licensing does not grant the separate model-weight rights.","source_ids":["evidence-official-bfbc8babf3630ddb3ed7","evidence-official-7972ce2bdd6a5e40d8ad","evidence-official-9863509ad10b824dc96b"],"source_locator":"June 15 2023 scFoundation manuscript: Results pre-training framework; Methods Data collection, Embedding, Encoder, Decoder and Read-depth-aware pre-training; official model README"}],"diagram":{"title":"scFoundation workflow","steps":["Expression and read-depth indicators","Gene and continuous-value embeddings","Sparse-input transformer encoder","Full-gene Performer decoder","Cell/gene embeddings or expression"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-bfbc8babf3630ddb3ed7","evidence-official-7972ce2bdd6a5e40d8ad","evidence-official-9863509ad10b824dc96b"],"source_locator":"June 15 2023 scFoundation manuscript: Results pre-training framework; Methods Data collection, Embedding, Encoder, Decoder and Read-depth-aware pre-training; official model README"},"coverage":"limited","gaps":["Training cutoff: The June 2023 manuscript lists GEO, Single Cell Portal, HCA and EMBL-EBI as collection sources; its data-collection section does not give a shared last-included-study date."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}}}
{"id":"catalog-model-scgpt","kind":"model","name":"scGPT","description":"Existing candidate catalogue entry; exact checkpoint and metadata require primary-source review.","status":"discovered","facets":{"areas":["cells-tissues"],"method_types":["foundation model"]},"source_ids":["catalog-source-scgpt"],"links":[{"relation":"variant_of","target_id":"discovery-model-scgpt"}],"attributes":{"entity_level":"family","version":"whole-human","reported_name":"scGPT","access":"Public code and downloadable checkpoints; use the unfine-tuned whole-human model for a new task.","method_type":"foundation model","historical_missing_metadata":{"checkpoint_revision":"not_yet_extracted","training_data":"not_yet_extracted","licence":"not_yet_extracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"scGPT learns representations of single-cell molecular measurements and supports task-specific adaptation.","summary_source_ids":["evidence-official-0e4dac85aa0e0ff1d4b2","evidence-official-d8e2d4f2e752370b6565","evidence-official-efa6ec9d886a21534a20"],"summary_source_locator":"May 2023 scGPT preprint Sections 4.1–4.3 and 4.8; current scgpt/model/model.py and README pretrained model table; README.md: Pretrained scGPT Model Zoo, whole-human row","sections":[{"title":"How it works","body":"scGPT combines each gene identity with its expression-value encoding before transformer attention. The implementation supports several expression encoders and cell-pooling choices. Task heads then predict expression or cell labels; optional masking and batch objectives depend on the training configuration.","source_ids":["evidence-official-0e4dac85aa0e0ff1d4b2","evidence-official-d8e2d4f2e752370b6565","evidence-official-efa6ec9d886a21534a20"],"source_locator":"May 2023 scGPT preprint Sections 4.1–4.3 and 4.8; current scgpt/model/model.py and README pretrained model table; README.md: Pretrained scGPT Model Zoo, whole-human row"},{"title":"Versions and reproducibility","body":"The May 2023 preprint reports an early 10M-cell model. The current whole-human checkpoint table reports 33M normal human cells; these sources describe different releases. The applicable input limits require configuration-specific checking.","source_ids":["evidence-official-0e4dac85aa0e0ff1d4b2","evidence-official-d8e2d4f2e752370b6565","evidence-official-efa6ec9d886a21534a20"],"source_locator":"May 2023 scGPT preprint Sections 4.1–4.3 and 4.8; current scgpt/model/model.py and README pretrained model table; README.md: Pretrained scGPT Model Zoo, whole-human row"}],"facts":[{"label":"Model type","value":"Generative single-cell transformer","status":"source_checked","source_ids":["evidence-official-0e4dac85aa0e0ff1d4b2","evidence-official-d8e2d4f2e752370b6565","evidence-official-efa6ec9d886a21534a20"],"source_locator":"May 2023 scGPT preprint Sections 4.1–4.3 and 4.8; current scgpt/model/model.py and README pretrained model table; README.md: Pretrained scGPT Model Zoo, whole-human row"},{"label":"Architecture","value":"Transformer backbone combining learned gene-token embeddings with expression-value encodings and optional batch encodings. Separate expression, cell-classification and optional masked-value or batch-discriminator heads support configured tasks.","status":"source_checked","source_ids":["evidence-official-0e4dac85aa0e0ff1d4b2","evidence-official-d8e2d4f2e752370b6565","evidence-official-efa6ec9d886a21534a20"],"source_locator":"May 2023 scGPT preprint Sections 4.1–4.3 and 4.8; current scgpt/model/model.py and README pretrained model table; README.md: Pretrained scGPT Model Zoo, whole-human row"},{"label":"Inputs","value":"Gene-expression measurements with the checkpoint-matched gene vocabulary.","status":"source_checked","source_ids":["evidence-official-0e4dac85aa0e0ff1d4b2","evidence-official-d8e2d4f2e752370b6565","evidence-official-efa6ec9d886a21534a20"],"source_locator":"May 2023 scGPT preprint Sections 4.1–4.3 and 4.8; current scgpt/model/model.py and README pretrained model table; README.md: Pretrained scGPT Model Zoo, whole-human row"},{"label":"Outputs","value":"Cell/gene representations and task-specific predictions after the relevant workflow.","status":"source_checked","source_ids":["evidence-official-0e4dac85aa0e0ff1d4b2","evidence-official-d8e2d4f2e752370b6565","evidence-official-efa6ec9d886a21534a20"],"source_locator":"May 2023 scGPT preprint Sections 4.1–4.3 and 4.8; current scgpt/model/model.py and README pretrained model table; README.md: Pretrained scGPT Model Zoo, whole-human row"},{"label":"Parameters","value":"The May 2023 model has 12 transformer blocks, width 512 and eight heads. The inspected current model-zoo table does not state the exact parameter total of its separate 33M-cell checkpoint.","status":"unreported","source_ids":["evidence-official-0e4dac85aa0e0ff1d4b2","evidence-official-d8e2d4f2e752370b6565","evidence-official-efa6ec9d886a21534a20"],"source_locator":"May 2023 scGPT preprint Sections 4.1–4.3 and 4.8; current scgpt/model/model.py and README pretrained model table; README.md: Pretrained scGPT Model Zoo, whole-human row"},{"label":"Known versions","value":"The May 2023 preprint reports an early 10M-cell model. The current whole-human checkpoint table reports 33M normal human cells; these sources describe different releases.","status":"source_checked","source_ids":["evidence-official-0e4dac85aa0e0ff1d4b2","evidence-official-d8e2d4f2e752370b6565","evidence-official-efa6ec9d886a21534a20"],"source_locator":"May 2023 scGPT preprint Sections 4.1–4.3 and 4.8; current scgpt/model/model.py and README pretrained model table; README.md: Pretrained scGPT Model Zoo, whole-human row"},{"label":"Training data","value":"The current whole-human model-zoo checkpoint uses 33M normal human cells, alongside separately released organ-specific and pan-cancer models. The earlier May 2023 preprint describes 10M training cells; its corpus is not the current checkpoint corpus.","status":"source_checked","source_ids":["evidence-official-0e4dac85aa0e0ff1d4b2","evidence-official-d8e2d4f2e752370b6565","evidence-official-efa6ec9d886a21534a20"],"source_locator":"May 2023 scGPT preprint Sections 4.1–4.3 and 4.8; current scgpt/model/model.py and README pretrained model table; README.md: Pretrained scGPT Model Zoo, whole-human row"},{"label":"Training cutoff","value":"The reviewed early manuscript and current whole-human model-zoo entry describe different corpora; neither supplies a shared latest-study date for the current checkpoint.","status":"unreported","source_ids":["evidence-official-0e4dac85aa0e0ff1d4b2","evidence-official-d8e2d4f2e752370b6565","evidence-official-efa6ec9d886a21534a20"],"source_locator":"May 2023 scGPT preprint Sections 4.1–4.3 and 4.8; current scgpt/model/model.py and README pretrained model table; README.md: Pretrained scGPT Model Zoo, whole-human row"},{"label":"Context limits","value":"The implementation accepts variable gene sets matched to its vocabulary. The reviewed model-zoo entry does not specify one validated maximum gene sequence for the current whole-human checkpoint.","status":"unreported","source_ids":["evidence-official-0e4dac85aa0e0ff1d4b2","evidence-official-d8e2d4f2e752370b6565","evidence-official-efa6ec9d886a21534a20"],"source_locator":"May 2023 scGPT preprint Sections 4.1–4.3 and 4.8; current scgpt/model/model.py and README pretrained model table; README.md: Pretrained scGPT Model Zoo, whole-human row"},{"label":"Weights licence","value":"Separate checkpoint-distribution terms are not stated in the inspected release documentation and licence material. The source-code licence alone is not recorded as an explicit weight grant.","status":"unreported","source_ids":["evidence-official-0e4dac85aa0e0ff1d4b2","evidence-official-d8e2d4f2e752370b6565","evidence-official-efa6ec9d886a21534a20","evidence-official-43484d4de29aacd65ed7"],"source_locator":"May 2023 scGPT preprint Sections 4.1–4.3 and 4.8; current scgpt/model/model.py and README pretrained model table; README.md: Pretrained scGPT Model Zoo, whole-human row; LICENSE: licence text"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/bowang-lab/scGPT","status":"source_checked","source_ids":["evidence-official-0e4dac85aa0e0ff1d4b2","evidence-official-d8e2d4f2e752370b6565","evidence-official-efa6ec9d886a21534a20"],"source_locator":"May 2023 scGPT preprint Sections 4.1–4.3 and 4.8; current scgpt/model/model.py and README pretrained model table; README.md: Pretrained scGPT Model Zoo, whole-human row"},{"label":"Code licence","value":"MIT","status":"source_checked","source_ids":["evidence-official-43484d4de29aacd65ed7"],"source_locator":"LICENSE: licence text"}],"strengths":[{"text":"Releases include whole-human and organ-specific checkpoints, with tutorials for reference mapping and other downstream tasks.","source_ids":["evidence-official-0e4dac85aa0e0ff1d4b2","evidence-official-d8e2d4f2e752370b6565","evidence-official-efa6ec9d886a21534a20"],"source_locator":"May 2023 scGPT preprint Sections 4.1–4.3 and 4.8; current scgpt/model/model.py and README pretrained model table; README.md: Pretrained scGPT Model Zoo, whole-human row"}],"limitations":[{"text":"Checkpoint choice and vocabulary must match the biological context. A whole-human pretrained encoder and a fine-tuned annotation or perturbation model are distinct evaluated configurations.","source_ids":["evidence-official-0e4dac85aa0e0ff1d4b2","evidence-official-d8e2d4f2e752370b6565","evidence-official-efa6ec9d886a21534a20"],"source_locator":"May 2023 scGPT preprint Sections 4.1–4.3 and 4.8; current scgpt/model/model.py and README pretrained model table; README.md: Pretrained scGPT Model Zoo, whole-human row"}],"diagram":{"title":"scGPT workflow","steps":["Gene expression and vocabulary","scGPT encoder","Cell and gene representations","Task-specific adaptation"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-0e4dac85aa0e0ff1d4b2","evidence-official-d8e2d4f2e752370b6565","evidence-official-efa6ec9d886a21534a20"],"source_locator":"May 2023 scGPT preprint Sections 4.1–4.3 and 4.8; current scgpt/model/model.py and README pretrained model table; README.md: Pretrained scGPT Model Zoo, whole-human row"},"coverage":"limited","gaps":["Parameters: The May 2023 model has 12 transformer blocks, width 512 and eight heads. The inspected current model-zoo table does not state the exact parameter total of its separate 33M-cell checkpoint.","Training cutoff: The reviewed early manuscript and current whole-human model-zoo entry describe different corpora; neither supplies a shared latest-study date for the current checkpoint.","Context limits: The implementation accepts variable gene sets matched to its vocabulary. The reviewed model-zoo entry does not specify one validated maximum gene sequence for the current whole-human checkpoint.","Weights licence: Separate checkpoint-distribution terms are not stated in the inspected release documentation and licence material. The source-code licence alone is not recorded as an explicit weight grant."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}}}
{"id":"catalog-model-scvi","kind":"model","name":"scVI","description":"Existing candidate catalogue entry; exact checkpoint and metadata require primary-source review.","status":"discovered","facets":{"areas":["cells-tissues"],"method_types":["baseline"]},"source_ids":["catalog-source-scvi"],"links":[],"attributes":{"entity_level":"family","version":"scvi-tools","reported_name":"scVI","access":"Public software; train a task-specific model on the permitted split.","method_type":"baseline","historical_missing_metadata":{"checkpoint_revision":"not_yet_extracted","training_data":"not_yet_extracted","licence":"not_yet_extracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"scVI models single-cell RNA counts with a probabilistic latent-variable model that accounts for observed covariates.","summary_source_ids":["evidence-official-550f6719c2f22f4d57b3","evidence-official-0cbab80c2d1769287a1f"],"summary_source_locator":"docs/user_guide/models/scvi.md: Preliminaries, Generative process, Inference and Limitations","sections":[{"title":"How it works","body":"scVI models single-cell RNA counts with a probabilistic latent-variable model that accounts for observed covariates. Variational autoencoder with a count likelihood and neural encoder/decoder; likelihood and batch/dispersion settings are configurable. The documented inputs are cell-by-gene count matrix, optionally with batch, donor or other covariates. The output consists of low-dimensional cell representations, normalized expression and probabilistic downstream quantities.","source_ids":["evidence-official-550f6719c2f22f4d57b3","evidence-official-0cbab80c2d1769287a1f"],"source_locator":"docs/user_guide/models/scvi.md: Preliminaries, Generative process, Inference and Limitations"},{"title":"Versions and reproducibility","body":"scVI model within scvi-tools; package version, likelihood, covariates and checkpoint are evaluation-specific. Gene-feature matrix rather than a fixed sequence-token window.","source_ids":["evidence-official-550f6719c2f22f4d57b3","evidence-official-0cbab80c2d1769287a1f"],"source_locator":"docs/user_guide/models/scvi.md: Preliminaries, Generative process, Inference and Limitations"}],"facts":[{"label":"Model type","value":"Variational autoencoder for count data","status":"source_checked","source_ids":["evidence-official-550f6719c2f22f4d57b3","evidence-official-0cbab80c2d1769287a1f"],"source_locator":"docs/user_guide/models/scvi.md: Preliminaries, Generative process, Inference and Limitations"},{"label":"Architecture","value":"Variational autoencoder with a count likelihood and neural encoder/decoder; likelihood and batch/dispersion settings are configurable.","status":"source_checked","source_ids":["evidence-official-550f6719c2f22f4d57b3","evidence-official-0cbab80c2d1769287a1f"],"source_locator":"docs/user_guide/models/scvi.md: Preliminaries, Generative process, Inference and Limitations"},{"label":"Inputs","value":"Cell-by-gene count matrix, optionally with batch, donor or other covariates.","status":"source_checked","source_ids":["evidence-official-550f6719c2f22f4d57b3","evidence-official-0cbab80c2d1769287a1f"],"source_locator":"docs/user_guide/models/scvi.md: Preliminaries, Generative process, Inference and Limitations"},{"label":"Outputs","value":"Low-dimensional cell representations, normalized expression and probabilistic downstream quantities.","status":"source_checked","source_ids":["evidence-official-550f6719c2f22f4d57b3","evidence-official-0cbab80c2d1769287a1f"],"source_locator":"docs/user_guide/models/scvi.md: Preliminaries, Generative process, Inference and Limitations"},{"label":"Parameters","value":"Configuration-dependent, including gene count and encoder/decoder dimensions.","status":"source_checked","source_ids":["evidence-official-550f6719c2f22f4d57b3","evidence-official-0cbab80c2d1769287a1f"],"source_locator":"docs/user_guide/models/scvi.md: Preliminaries, Generative process, Inference and Limitations"},{"label":"Known versions","value":"scVI model within scvi-tools; package version, likelihood, covariates and checkpoint are evaluation-specific.","status":"source_checked","source_ids":["evidence-official-550f6719c2f22f4d57b3","evidence-official-0cbab80c2d1769287a1f"],"source_locator":"docs/user_guide/models/scvi.md: Preliminaries, Generative process, Inference and Limitations"},{"label":"Training data","value":"Fitted to the user-selected count matrix or a specified pretrained reference; scVI is not one universal checkpoint.","status":"source_checked","source_ids":["evidence-official-550f6719c2f22f4d57b3","evidence-official-0cbab80c2d1769287a1f"],"source_locator":"docs/user_guide/models/scvi.md: Preliminaries, Generative process, Inference and Limitations"},{"label":"Training cutoff","value":"Inapplicable as one universal pretraining date: scVI is fitted to the supplied dataset, whose collection date and train/test split belong to the evaluation.","status":"inapplicable","source_ids":["evidence-official-550f6719c2f22f4d57b3","evidence-official-0cbab80c2d1769287a1f"],"source_locator":"docs/user_guide/models/scvi.md: Preliminaries, Generative process, Inference and Limitations"},{"label":"Context limits","value":"Gene-feature matrix rather than a fixed sequence-token window.","status":"source_checked","source_ids":["evidence-official-550f6719c2f22f4d57b3","evidence-official-0cbab80c2d1769287a1f"],"source_locator":"docs/user_guide/models/scvi.md: Preliminaries, Generative process, Inference and Limitations"},{"label":"Weights licence","value":"No universal weights release applies to a model fitted on each dataset; any reused checkpoint requires its own licence.","status":"source_checked","source_ids":["evidence-official-550f6719c2f22f4d57b3","evidence-official-0cbab80c2d1769287a1f"],"source_locator":"docs/user_guide/models/scvi.md: Preliminaries, Generative process, Inference and Limitations"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/scverse/scvi-tools","status":"source_checked","source_ids":["evidence-official-550f6719c2f22f4d57b3","evidence-official-0cbab80c2d1769287a1f"],"source_locator":"docs/user_guide/models/scvi.md: Preliminaries, Generative process, Inference and Limitations"},{"label":"Code licence","value":"BSD-3-Clause","status":"source_checked","source_ids":["evidence-official-6851724e3bcb7e9d2781"],"source_locator":"LICENSE: licence text"}],"strengths":[{"text":"Models count observations directly and supports batch-conditioned expression estimates and reference-to-query transfer.","source_ids":["evidence-official-550f6719c2f22f4d57b3","evidence-official-0cbab80c2d1769287a1f"],"source_locator":"docs/user_guide/models/scvi.md: Preliminaries, Generative process, Inference and Limitations"}],"limitations":[{"text":"The documentation notes that the latent space is less interpretable than a linear method and efficient inference generally benefits from a GPU. Covariates and likelihood must be reported.","source_ids":["evidence-official-550f6719c2f22f4d57b3","evidence-official-0cbab80c2d1769287a1f"],"source_locator":"docs/user_guide/models/scvi.md: Preliminaries, Generative process, Inference and Limitations"}],"diagram":{"title":"scVI workflow","steps":["RNA counts and covariates","Variational encoder","Latent cell state","Count decoder and estimates"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-550f6719c2f22f4d57b3","evidence-official-0cbab80c2d1769287a1f"],"source_locator":"docs/user_guide/models/scvi.md: Preliminaries, Generative process, Inference and Limitations"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}}}
{"id":"catalog-model-spliceai","kind":"model","name":"SpliceAI","description":"Existing candidate catalogue entry; exact checkpoint and metadata require primary-source review.","status":"discovered","facets":{"areas":["dna-genomes"],"method_types":["specialist"]},"source_ids":["catalog-source-spliceai"],"links":[{"relation":"variant_of","target_id":"discovery-model-spliceai"}],"attributes":{"entity_level":"family","version":"1.3.1","reported_name":"SpliceAI","access":"Public archived code under PolyForm Strict; model weights are CC BY-NC 4.0 for non-commercial use.","method_type":"specialist","historical_missing_metadata":{"checkpoint_revision":"not_yet_extracted","training_data":"not_yet_extracted","licence":"not_yet_extracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"SpliceAI annotates sequence variants with predicted splice acceptor and donor changes.","summary_source_ids":["evidence-official-864a2c5ea6e61aa310f9","evidence-official-ded281404bd1a0f3fdb7"],"summary_source_locator":"Jaganathan et al., Cell 2019, STAR Methods: SpliceAI architecture and Model training and testing (pp. e2–e3); README.md: usage and License","sections":[{"title":"How it works","body":"SpliceAI reads one-hot-encoded DNA through dilated convolutional residual blocks. Skip connections combine features at different depths, and a softmax layer assigns acceptor, donor or neither probabilities to the central positions. The 10kb version requires 5kb of sequence on each side of a scored position; variant scoring compares the reference and alternate predictions.","source_ids":["evidence-official-864a2c5ea6e61aa310f9","evidence-official-ded281404bd1a0f3fdb7"],"source_locator":"Jaganathan et al., Cell 2019, STAR Methods: SpliceAI architecture and Model training and testing (pp. e2–e3); README.md: usage and License"},{"title":"Versions and reproducibility","body":"The paper studies 80nt, 400nt, 2kb and 10kb receptive spans. Variant scoring averages five independently trained models; these are not five different assay results. SpliceAI-10k uses 5,000 flanking bases on each side. An input of length l + 10,000 produces predictions for l central positions; receptive span is distinct from maximum input length.","source_ids":["evidence-official-864a2c5ea6e61aa310f9","evidence-official-ded281404bd1a0f3fdb7"],"source_locator":"Jaganathan et al., Cell 2019, STAR Methods: SpliceAI architecture and Model training and testing (pp. e2–e3); README.md: usage and License"}],"facts":[{"label":"Model type","value":"Dilated convolutional splicing predictor","status":"source_checked","source_ids":["evidence-official-864a2c5ea6e61aa310f9","evidence-official-ded281404bd1a0f3fdb7"],"source_locator":"Jaganathan et al., Cell 2019, STAR Methods: SpliceAI architecture and Model training and testing (pp. e2–e3); README.md: usage and License"},{"label":"Architecture","value":"Residual one-dimensional convolutional network with dilated kernels and skip connections; a softmax head predicts acceptor, donor and neither at each central position.","status":"source_checked","source_ids":["evidence-official-864a2c5ea6e61aa310f9","evidence-official-ded281404bd1a0f3fdb7"],"source_locator":"Jaganathan et al., Cell 2019, STAR Methods: SpliceAI architecture and Model training and testing (pp. e2–e3); README.md: usage and License"},{"label":"Inputs","value":"VCF variants, reference FASTA and matching gene annotation, or custom one-hot-encoded sequence.","status":"source_checked","source_ids":["evidence-official-864a2c5ea6e61aa310f9","evidence-official-ded281404bd1a0f3fdb7"],"source_locator":"Jaganathan et al., Cell 2019, STAR Methods: SpliceAI architecture and Model training and testing (pp. e2–e3); README.md: usage and License"},{"label":"Outputs","value":"Acceptor/donor gain/loss scores and positions in VCF INFO annotations.","status":"source_checked","source_ids":["evidence-official-864a2c5ea6e61aa310f9","evidence-official-ded281404bd1a0f3fdb7"],"source_locator":"Jaganathan et al., Cell 2019, STAR Methods: SpliceAI architecture and Model training and testing (pp. e2–e3); README.md: usage and License"},{"label":"Parameters","value":"The original STAR Methods specifies residual blocks, dilation and receptive spans, but does not state a complete parameter count for each released five-model scoring ensemble.","status":"unreported","source_ids":["evidence-official-864a2c5ea6e61aa310f9","evidence-official-ded281404bd1a0f3fdb7"],"source_locator":"Jaganathan et al., Cell 2019, STAR Methods: SpliceAI architecture and Model training and testing (pp. e2–e3); README.md: usage and License"},{"label":"Known versions","value":"The paper studies 80nt, 400nt, 2kb and 10kb receptive spans. Variant scoring averages five independently trained models; these are not five different assay results.","status":"source_checked","source_ids":["evidence-official-864a2c5ea6e61aa310f9","evidence-official-ded281404bd1a0f3fdb7"],"source_locator":"Jaganathan et al., Cell 2019, STAR Methods: SpliceAI architecture and Model training and testing (pp. e2–e3); README.md: usage and License"},{"label":"Training data","value":"Human GRCh37 sequence and GENCODE V24lift37 principal protein-coding transcripts, split by chromosome with non-paralogous held-out test genes. The paper distinguishes GENCODE-only training from GTEx-junction-augmented models used for variant analyses.","status":"source_checked","source_ids":["evidence-official-864a2c5ea6e61aa310f9","evidence-official-ded281404bd1a0f3fdb7"],"source_locator":"Jaganathan et al., Cell 2019, STAR Methods: SpliceAI architecture and Model training and testing (pp. e2–e3); README.md: usage and License"},{"label":"Training cutoff","value":"GENCODE V24lift37 on GRCh37 defines the documented transcript annotations. GTEx-augmented training is separately described; the paper does not give one common latest-data date for both variants.","status":"source_checked","source_ids":["evidence-official-864a2c5ea6e61aa310f9","evidence-official-ded281404bd1a0f3fdb7"],"source_locator":"Jaganathan et al., Cell 2019, STAR Methods: SpliceAI architecture and Model training and testing (pp. e2–e3); README.md: usage and License"},{"label":"Context limits","value":"SpliceAI-10k uses 5,000 flanking bases on each side. An input of length l + 10,000 produces predictions for l central positions; receptive span is distinct from maximum input length.","status":"source_checked","source_ids":["evidence-official-864a2c5ea6e61aa310f9","evidence-official-ded281404bd1a0f3fdb7"],"source_locator":"Jaganathan et al., Cell 2019, STAR Methods: SpliceAI architecture and Model training and testing (pp. e2–e3); README.md: usage and License"},{"label":"Weights licence","value":"CC-BY-NC-4.0 for trained models; commercial use requires a separate licence. Code is PolyForm Strict 1.0.0.","status":"source_checked","source_ids":["evidence-official-864a2c5ea6e61aa310f9","evidence-official-ded281404bd1a0f3fdb7"],"source_locator":"Jaganathan et al., Cell 2019, STAR Methods: SpliceAI architecture and Model training and testing (pp. e2–e3); README.md: usage and License"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/Illumina/SpliceAI","status":"source_checked","source_ids":["evidence-official-864a2c5ea6e61aa310f9","evidence-official-ded281404bd1a0f3fdb7"],"source_locator":"Jaganathan et al., Cell 2019, STAR Methods: SpliceAI architecture and Model training and testing (pp. e2–e3); README.md: usage and License"},{"label":"Code licence","value":"PolyForm Strict 1.0.0 for code; trained weights have separate terms.","status":"source_checked","source_ids":["evidence-official-c465f6d4fcc04f9afe4b"],"source_locator":"LICENSE: licence text"}],"strengths":[{"text":"Provides direct sequence inference and an annotation workflow with explicit genome and distance settings.","source_ids":["evidence-official-864a2c5ea6e61aa310f9","evidence-official-ded281404bd1a0f3fdb7"],"source_locator":"Jaganathan et al., Cell 2019, STAR Methods: SpliceAI architecture and Model training and testing (pp. e2–e3); README.md: usage and License"}],"limitations":[{"text":"The command-line pipeline skips unsupported variants and variants outside its gene annotations. Code, model weights and downloadable precomputed scores have distinct licensing provisions.","source_ids":["evidence-official-864a2c5ea6e61aa310f9","evidence-official-ded281404bd1a0f3fdb7"],"source_locator":"Jaganathan et al., Cell 2019, STAR Methods: SpliceAI architecture and Model training and testing (pp. e2–e3); README.md: usage and License"}],"diagram":{"title":"SpliceAI workflow","steps":["Variant plus sequence context","Reference and alternate predictions","Splice-site differences","Gain/loss annotations"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-864a2c5ea6e61aa310f9","evidence-official-ded281404bd1a0f3fdb7"],"source_locator":"Jaganathan et al., Cell 2019, STAR Methods: SpliceAI architecture and Model training and testing (pp. e2–e3); README.md: usage and License"},"coverage":"limited","gaps":["Parameters: The original STAR Methods specifies residual blocks, dilation and receptive spans, but does not state a complete parameter count for each released five-model scoring ensemble."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}}}
{"id":"catalog-model-vina","kind":"model","name":"AutoDock Vina","description":"Existing candidate catalogue entry; exact checkpoint and metadata require primary-source review.","status":"discovered","facets":{"areas":["molecular-interactions"],"method_types":["baseline"]},"source_ids":["catalog-source-vina"],"links":[],"attributes":{"entity_level":"family","version":"1.2.7","reported_name":"AutoDock Vina","access":"Public docking software; receptor and ligand preparation required.","method_type":"baseline","historical_missing_metadata":{"checkpoint_revision":"not_yet_extracted","training_data":"not_yet_extracted","licence":"not_yet_extracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"AutoDock Vina searches for ligand conformations and poses in a molecular docking problem.","summary_source_ids":["evidence-official-23e72f7e93a0a633ab2b"],"summary_source_locator":"README.md: introduction, feature list, license and Citations","sections":[{"title":"How it works","body":"AutoDock Vina searches for ligand conformations and poses in a molecular docking problem. Scoring functions coupled to gradient-based conformational optimization and search. The documented inputs are prepared receptor and ligand structures with the configured search space and scoring function. The output consists of candidate docking poses and corresponding docking scores.","source_ids":["evidence-official-23e72f7e93a0a633ab2b"],"source_locator":"README.md: introduction, feature list, license and Citations"},{"title":"Versions and reproducibility","body":"AutoDock Vina; README cites the 1.2.0 feature expansion separately from the original 2010 method. Molecular geometry and search-box constraints rather than a sequence context window.","source_ids":["evidence-official-23e72f7e93a0a633ab2b"],"source_locator":"README.md: introduction, feature list, license and Citations"}],"facts":[{"label":"Model type","value":"Classical molecular docking software","status":"source_checked","source_ids":["evidence-official-23e72f7e93a0a633ab2b"],"source_locator":"README.md: introduction, feature list, license and Citations"},{"label":"Architecture","value":"Scoring functions coupled to gradient-based conformational optimization and search.","status":"source_checked","source_ids":["evidence-official-23e72f7e93a0a633ab2b"],"source_locator":"README.md: introduction, feature list, license and Citations"},{"label":"Inputs","value":"Prepared receptor and ligand structures with the configured search space and scoring function.","status":"source_checked","source_ids":["evidence-official-23e72f7e93a0a633ab2b"],"source_locator":"README.md: introduction, feature list, license and Citations"},{"label":"Outputs","value":"Candidate docking poses and corresponding docking scores.","status":"source_checked","source_ids":["evidence-official-23e72f7e93a0a633ab2b"],"source_locator":"README.md: introduction, feature list, license and Citations"},{"label":"Parameters","value":"Inapplicable as a neural parameter count; scoring/search parameters are separately configured.","status":"inapplicable","source_ids":["evidence-official-23e72f7e93a0a633ab2b"],"source_locator":"README.md: introduction, feature list, license and Citations"},{"label":"Known versions","value":"AutoDock Vina; README cites the 1.2.0 feature expansion separately from the original 2010 method.","status":"source_checked","source_ids":["evidence-official-23e72f7e93a0a633ab2b"],"source_locator":"README.md: introduction, feature list, license and Citations"},{"label":"Training data","value":"Not a pretrained neural model. The selected scoring function and parametrization define the procedural baseline.","status":"source_checked","source_ids":["evidence-official-23e72f7e93a0a633ab2b"],"source_locator":"README.md: introduction, feature list, license and Citations"},{"label":"Training cutoff","value":"Inapplicable to neural pretraining; reference-database and input-data dates must be recorded for each run.","status":"inapplicable","source_ids":["evidence-official-23e72f7e93a0a633ab2b"],"source_locator":"README.md: introduction, feature list, license and Citations"},{"label":"Context limits","value":"Molecular geometry and search-box constraints rather than a sequence context window.","status":"source_checked","source_ids":["evidence-official-23e72f7e93a0a633ab2b"],"source_locator":"README.md: introduction, feature list, license and Citations"},{"label":"Weights licence","value":"Inapplicable: no neural model-weight checkpoint.","status":"inapplicable","source_ids":["evidence-official-23e72f7e93a0a633ab2b"],"source_locator":"README.md: introduction, feature list, license and Citations"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/ccsb-scripps/AutoDock-Vina","status":"source_checked","source_ids":["evidence-official-23e72f7e93a0a633ab2b"],"source_locator":"README.md: introduction, feature list, license and Citations"},{"label":"Code licence","value":"Apache-2.0","status":"source_checked","source_ids":["evidence-official-640665f30e62eed9319b"],"source_locator":"LICENSE: licence text"}],"strengths":[{"text":"Supports Vina and AutoDock4 scoring, multiple-ligand/batch workflows, macrocycles and Python bindings.","source_ids":["evidence-official-23e72f7e93a0a633ab2b"],"source_locator":"README.md: introduction, feature list, license and Citations"}],"limitations":[{"text":"The chosen scoring function, molecular preparation and search configuration form part of the method. A docking score is a computational quantity rather than an experimental affinity measurement.","source_ids":["evidence-official-23e72f7e93a0a633ab2b"],"source_locator":"README.md: introduction, feature list, license and Citations"}],"diagram":{"title":"AutoDock Vina workflow","steps":["Prepared receptor and ligand","Conformational search","Docking score","Ranked candidate poses"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-23e72f7e93a0a633ab2b"],"source_locator":"README.md: introduction, feature list, license and Citations"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}}}
{"id":"catalog-source-alphafold-3-server","kind":"source","name":"AlphaFold 3 Server official resource","description":"","status":"discovered","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://alphafoldserver.com/output-terms","version":null,"missing_metadata":{"version":"not_pinned_in_legacy_catalogue"},"retrieved_at":"2026-09-16T10:50:02Z"}}
{"id":"catalog-source-alphagenome","kind":"source","name":"AlphaGenome official resource","description":"","status":"discovered","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/google-deepmind/alphagenome_research","version":null,"missing_metadata":{"version":"not_pinned_in_legacy_catalogue"},"retrieved_at":"2026-09-16T10:50:02Z"}}
{"id":"catalog-source-boltz-2","kind":"source","name":"Boltz-2 official resource","description":"","status":"discovered","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/jwohlwend/boltz","version":null,"missing_metadata":{"version":"not_pinned_in_legacy_catalogue"},"retrieved_at":"2026-09-16T10:50:02Z"}}
{"id":"catalog-source-chai-1","kind":"source","name":"Chai-1 official resource","description":"","status":"discovered","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/chaidiscovery/chai-lab","version":null,"missing_metadata":{"version":"not_pinned_in_legacy_catalogue"},"retrieved_at":"2026-09-16T10:50:02Z"}}
{"id":"catalog-source-diffdock-l","kind":"source","name":"DiffDock-L official resource","description":"","status":"discovered","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/gcorso/DiffDock","version":null,"missing_metadata":{"version":"not_pinned_in_legacy_catalogue"},"retrieved_at":"2026-09-16T10:50:02Z"}}
{"id":"catalog-source-dnabert-2","kind":"source","name":"DNABERT-2 official resource","description":"","status":"discovered","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://huggingface.co/zhihan1996/DNABERT-2-117M","version":null,"missing_metadata":{"version":"not_pinned_in_legacy_catalogue"},"retrieved_at":"2026-09-16T10:50:02Z"}}
{"id":"catalog-source-esm-2","kind":"source","name":"ESM-2 official resource","description":"","status":"discovered","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/facebookresearch/esm","version":null,"missing_metadata":{"version":"not_pinned_in_legacy_catalogue"},"retrieved_at":"2026-09-16T10:50:02Z"}}
{"id":"catalog-source-esmfold","kind":"source","name":"ESMFold official resource","description":"","status":"discovered","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/facebookresearch/esm","version":null,"missing_metadata":{"version":"not_pinned_in_legacy_catalogue"},"retrieved_at":"2026-09-16T10:50:02Z"}}
{"id":"catalog-source-evo-2","kind":"source","name":"Evo 2 official resource","description":"","status":"discovered","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/ArcInstitute/evo2","version":null,"missing_metadata":{"version":"not_pinned_in_legacy_catalogue"},"retrieved_at":"2026-09-16T10:50:02Z"}}
{"id":"catalog-source-gears","kind":"source","name":"GEARS official resource","description":"","status":"discovered","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/snap-stanford/GEARS","version":null,"missing_metadata":{"version":"not_pinned_in_legacy_catalogue"},"retrieved_at":"2026-09-16T10:50:02Z"}}
{"id":"catalog-source-geneformer","kind":"source","name":"Geneformer official resource","description":"","status":"discovered","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://huggingface.co/ctheodoris/Geneformer","version":null,"missing_metadata":{"version":"not_pinned_in_legacy_catalogue"},"retrieved_at":"2026-09-16T10:50:02Z"}}
{"id":"catalog-source-kraken2","kind":"source","name":"Kraken2 official resource","description":"","status":"discovered","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/DerrickWood/kraken2","version":null,"missing_metadata":{"version":"not_pinned_in_legacy_catalogue"},"retrieved_at":"2026-09-16T10:50:02Z"}}
{"id":"catalog-source-metagene-1","kind":"source","name":"METAGENE-1 official resource","description":"","status":"discovered","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://huggingface.co/metagene-ai/METAGENE-1","version":null,"missing_metadata":{"version":"not_pinned_in_legacy_catalogue"},"retrieved_at":"2026-09-16T10:50:02Z"}}
{"id":"catalog-source-metaphlan","kind":"source","name":"MetaPhlAn official resource","description":"","status":"discovered","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/biobakery/MetaPhlAn","version":null,"missing_metadata":{"version":"not_pinned_in_legacy_catalogue"},"retrieved_at":"2026-09-16T10:50:02Z"}}
{"id":"catalog-source-mimic","kind":"source","name":"MIMIC official resource","description":"","status":"discovered","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://huggingface.co/polymathic-ai/MIMIC","version":null,"missing_metadata":{"version":"not_pinned_in_legacy_catalogue"},"retrieved_at":"2026-09-16T10:50:02Z"}}
{"id":"catalog-source-mrna-fm","kind":"source","name":"mRNA-FM official resource","description":"","status":"discovered","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/ml4bio/RNA-FM","version":null,"missing_metadata":{"version":"not_pinned_in_legacy_catalogue"},"retrieved_at":"2026-09-16T10:50:02Z"}}
{"id":"catalog-source-nt-v2","kind":"source","name":"Nucleotide Transformer v2 official resource","description":"","status":"discovered","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://huggingface.co/InstaDeepAI/nucleotide-transformer-v2-50m-multi-species","version":null,"missing_metadata":{"version":"not_pinned_in_legacy_catalogue"},"retrieved_at":"2026-09-16T10:50:02Z"}}
{"id":"catalog-source-pangolin","kind":"source","name":"Pangolin official resource","description":"","status":"discovered","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/tkzeng/Pangolin","version":null,"missing_metadata":{"version":"not_pinned_in_legacy_catalogue"},"retrieved_at":"2026-09-16T10:50:02Z"}}
{"id":"catalog-source-prokbert","kind":"source","name":"ProkBERT official resource","description":"","status":"discovered","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/nbrg-ppcu/prokbert","version":null,"missing_metadata":{"version":"not_pinned_in_legacy_catalogue"},"retrieved_at":"2026-09-16T10:50:02Z"}}
{"id":"catalog-source-proteinmpnn","kind":"source","name":"ProteinMPNN official resource","description":"","status":"discovered","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/dauparas/ProteinMPNN","version":null,"missing_metadata":{"version":"not_pinned_in_legacy_catalogue"},"retrieved_at":"2026-09-16T10:50:02Z"}}
{"id":"catalog-source-rhofold","kind":"source","name":"RhoFold+ official resource","description":"","status":"discovered","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/ml4bio/RhoFold","version":null,"missing_metadata":{"version":"not_pinned_in_legacy_catalogue"},"retrieved_at":"2026-09-16T10:50:02Z"}}
{"id":"catalog-source-rna-fm","kind":"source","name":"RNA-FM official resource","description":"","status":"discovered","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/ml4bio/RNA-FM","version":null,"missing_metadata":{"version":"not_pinned_in_legacy_catalogue"},"retrieved_at":"2026-09-16T10:50:02Z"}}
{"id":"catalog-source-scfoundation","kind":"source","name":"scFoundation official resource","description":"","status":"discovered","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/biomap-research/scFoundation","version":null,"missing_metadata":{"version":"not_pinned_in_legacy_catalogue"},"retrieved_at":"2026-09-16T10:50:02Z"}}
{"id":"catalog-source-scgpt","kind":"source","name":"scGPT official resource","description":"","status":"discovered","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/bowang-lab/scGPT","version":null,"missing_metadata":{"version":"not_pinned_in_legacy_catalogue"},"retrieved_at":"2026-09-16T10:50:02Z"}}
{"id":"catalog-source-scvi","kind":"source","name":"scVI official resource","description":"","status":"discovered","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/scverse/scvi-tools","version":null,"missing_metadata":{"version":"not_pinned_in_legacy_catalogue"},"retrieved_at":"2026-09-16T10:50:02Z"}}
{"id":"catalog-source-spliceai","kind":"source","name":"SpliceAI official resource","description":"","status":"discovered","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/Illumina/SpliceAI","version":null,"missing_metadata":{"version":"not_pinned_in_legacy_catalogue"},"retrieved_at":"2026-09-16T10:50:02Z"}}
{"id":"catalog-source-vina","kind":"source","name":"AutoDock Vina official resource","description":"","status":"discovered","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/ccsb-scripps/AutoDock-Vina","version":null,"missing_metadata":{"version":"not_pinned_in_legacy_catalogue"},"retrieved_at":"2026-09-16T10:50:02Z"}}
{"id":"catalog-task-cell-batch-integration","kind":"benchmark","name":"Batch integration","description":"","status":"discovered","facets":{"areas":["cells-tissues"]},"source_ids":["catalog-source-scgpt","catalog-source-scvi"],"links":[],"attributes":{"entity_level":"task","version":null,"task":"Batch integration","scope_note":"Test whether cell identity is retained across donors and batches.","historical_missing_metadata":{"protocol_version":"not_yet_extracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Batch integration asks whether measurements from different experiments can be combined while preserving biological differences.","summary_source_ids":["src-discovery-theislab-scib"],"summary_source_locator":"README: Metrics, Biological Conservation, Batch Correction and Integration Tools","sections":[{"title":"Choosing an evaluation","body":"Assess two objectives separately: removal of technical batch effects and retention of cell identities and biological variation. scIB supplies metrics for both. A model that mixes every cell together can score well on mixing while destroying useful biology, so one objective cannot substitute for the other. Use a concrete scIB or Open Problems protocol for datasets, preprocessing and scoring; those resources are not interchangeable.","source_ids":["src-discovery-theislab-scib"],"source_locator":"README: Metrics, Biological Conservation, Batch Correction and Integration Tools"},{"title":"Task scope","body":"This is a task guide, not a single versioned benchmark protocol. The connected resources provide examples or concrete procedures. A candidate method or proposed control is not evidence that an evaluation has been completed.","source_ids":["src-discovery-theislab-scib"],"source_locator":"README: Metrics, Biological Conservation, Batch Correction and Integration Tools"}],"facts":[{"label":"Entity type","value":"Task guide; concrete protocol identities remain separate.","status":"source_checked","source_ids":["src-discovery-theislab-scib"],"source_locator":"README: Metrics, Biological Conservation, Batch Correction and Integration Tools"},{"label":"Datasets","value":"No single dataset is fixed by this guide. Select a linked protocol and its versioned data release.","status":"inapplicable","source_ids":["src-discovery-theislab-scib"],"source_locator":"README: Metrics, Biological Conservation, Batch Correction and Integration Tools"},{"label":"Organisms","value":"No shared organism population is defined at this guide level. Record it for each selected dataset.","status":"inapplicable","source_ids":["src-discovery-theislab-scib"],"source_locator":"README: Metrics, Biological Conservation, Batch Correction and Integration Tools"},{"label":"Assays","value":"No single measurement assay is fixed by this guide; the endpoint and assay belong to the selected protocol.","status":"inapplicable","source_ids":["src-discovery-theislab-scib"],"source_locator":"README: Metrics, Biological Conservation, Batch Correction and Integration Tools"},{"label":"Splits","value":"No executable split is attached to this task identity. Use the selected protocol’s split manifest.","status":"inapplicable","source_ids":["src-discovery-theislab-scib"],"source_locator":"README: Metrics, Biological Conservation, Batch Correction and Integration Tools"},{"label":"Allowed inputs","value":"Single-cell measurements with batch labels and biological annotations; the chosen method may return a corrected matrix, embedding or graph.","status":"source_checked","source_ids":["src-discovery-theislab-scib"],"source_locator":"README: Metrics, Biological Conservation, Batch Correction and Integration Tools"},{"label":"Adaptation","value":"No common fitting regime is imposed here. Keep pretrained, frozen, probed, fine-tuned and conventional methods distinct where applicable.","status":"inapplicable","source_ids":["src-discovery-theislab-scib"],"source_locator":"README: Metrics, Biological Conservation, Batch Correction and Integration Tools"},{"label":"Metrics","value":"scIB includes cell-type silhouette, ARI/NMI and trajectory conservation for biology; batch silhouette, iLISI, kBET and graph connectivity for batch effects.","status":"source_checked","source_ids":["src-discovery-theislab-scib"],"source_locator":"README: Metrics, Biological Conservation, Batch Correction and Integration Tools"},{"label":"Baselines","value":"Documented integration comparators include Harmony, MNN, scVI and Seurat. Versions and preprocessing are part of the configuration.","status":"source_checked","source_ids":["src-discovery-theislab-scib"],"source_locator":"README: Metrics, Biological Conservation, Batch Correction and Integration Tools"}],"strengths":[{"text":"A two-part assessment can expose overcorrection that a batch-mixing score alone misses.","source_ids":["src-discovery-theislab-scib"],"source_locator":"README: Metrics, Biological Conservation, Batch Correction and Integration Tools"}],"limitations":[{"text":"Metric choice depends on available labels and output representation. This guide supplies neither a shared cohort nor a universal combined score.","source_ids":["src-discovery-theislab-scib"],"source_locator":"README: Metrics, Biological Conservation, Batch Correction and Integration Tools"}],"diagram":{"title":"Conceptual evaluation workflow","steps":["Select annotated datasets and protocol","Apply the integration method","Measure retained biological variation","Measure residual batch effects"],"caption":"Conceptual task guide. Dataset preparation, parameters and scoring must come from a separately identified protocol.","source_ids":["src-discovery-theislab-scib"],"source_locator":"README: Metrics, Biological Conservation, Batch Correction and Integration Tools"},"coverage":"reviewed","gaps":["A concrete evaluation still needs a protocol version, dataset release, eligible/scored counts and documented uncertainty. This guide does not manufacture those run-specific values."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Reviewed named primary documentation and existing protocol records. Editorial task scope is distinguished from source-reported procedures. No model execution or new numerical result; a reviewed guide is not a complete runnable protocol."}}}}
{"id":"catalog-task-cell-perturbation","kind":"benchmark","name":"Perturbation response","description":"","status":"discovered","facets":{"areas":["cells-tissues"]},"source_ids":["catalog-source-geneformer","catalog-source-scfoundation","catalog-source-gears"],"links":[],"attributes":{"entity_level":"task","version":null,"task":"Perturbation response","scope_note":"Predict expression changes after unseen perturbations.","historical_missing_metadata":{"protocol_version":"not_yet_extracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Perturbation-response prediction asks whether a model can predict molecular measurements under a changed cellular condition.","summary_source_ids":["src-discovery-altoslabs-perturbench"],"summary_source_locator":"README: Downloading and Preparing Datasets, Data Splitting, Evaluator Class and Automated Evaluation","sections":[{"title":"Choosing an evaluation","body":"A concrete protocol defines which conditions are observed during fitting and which are held out. Compare predicted and measured responses with matched covariates and a declared aggregation. PerturBench separates average expression, changes from controls and distributional comparisons; these answer different questions. Report the dataset-specific split rather than assuming that a random cell split tests unseen perturbations.","source_ids":["src-discovery-altoslabs-perturbench"],"source_locator":"README: Downloading and Preparing Datasets, Data Splitting, Evaluator Class and Automated Evaluation"},{"title":"Task scope","body":"This is a task guide, not a single versioned benchmark protocol. The connected resources provide examples or concrete procedures. A candidate method or proposed control is not evidence that an evaluation has been completed.","source_ids":["src-discovery-altoslabs-perturbench"],"source_locator":"README: Downloading and Preparing Datasets, Data Splitting, Evaluator Class and Automated Evaluation"}],"facts":[{"label":"Entity type","value":"Task guide; concrete protocol identities remain separate.","status":"source_checked","source_ids":["src-discovery-altoslabs-perturbench"],"source_locator":"README: Downloading and Preparing Datasets, Data Splitting, Evaluator Class and Automated Evaluation"},{"label":"Datasets","value":"No single dataset is fixed by this guide. Select a linked protocol and its versioned data release.","status":"inapplicable","source_ids":["src-discovery-altoslabs-perturbench"],"source_locator":"README: Downloading and Preparing Datasets, Data Splitting, Evaluator Class and Automated Evaluation"},{"label":"Organisms","value":"No shared organism population is defined at this guide level. Record it for each selected dataset.","status":"inapplicable","source_ids":["src-discovery-altoslabs-perturbench"],"source_locator":"README: Downloading and Preparing Datasets, Data Splitting, Evaluator Class and Automated Evaluation"},{"label":"Assays","value":"No single measurement assay is fixed by this guide; the endpoint and assay belong to the selected protocol.","status":"inapplicable","source_ids":["src-discovery-altoslabs-perturbench"],"source_locator":"README: Downloading and Preparing Datasets, Data Splitting, Evaluator Class and Automated Evaluation"},{"label":"Splits","value":"No executable split is attached to this task identity. Use the selected protocol’s split manifest.","status":"inapplicable","source_ids":["src-discovery-altoslabs-perturbench"],"source_locator":"README: Downloading and Preparing Datasets, Data Splitting, Evaluator Class and Automated Evaluation"},{"label":"Allowed inputs","value":"Control-cell measurements, requested condition and permitted covariates; predictions and observed responses aligned to the same features.","status":"source_checked","source_ids":["src-discovery-altoslabs-perturbench"],"source_locator":"README: Downloading and Preparing Datasets, Data Splitting, Evaluator Class and Automated Evaluation"},{"label":"Adaptation","value":"No common fitting regime is imposed here. Keep pretrained, frozen, probed, fine-tuned and conventional methods distinct where applicable.","status":"inapplicable","source_ids":["src-discovery-altoslabs-perturbench"],"source_locator":"README: Downloading and Preparing Datasets, Data Splitting, Evaluator Class and Automated Evaluation"},{"label":"Metrics","value":"PerturBench supports expression-error, correlation, change-from-control and distributional metrics; aggregation and ranking settings are explicit configuration choices.","status":"source_checked","source_ids":["src-discovery-altoslabs-perturbench"],"source_locator":"README: Downloading and Preparing Datasets, Data Splitting, Evaluator Class and Automated Evaluation"},{"label":"Baselines","value":"Choose the baselines supplied by the selected protocol and record their access to controls and training conditions. No baseline result is asserted by this guide.","status":"source_checked","source_ids":["src-discovery-altoslabs-perturbench"],"source_locator":"README: Downloading and Preparing Datasets, Data Splitting, Evaluator Class and Automated Evaluation"}],"strengths":[{"text":"Matched controls and held-out conditions allow an evaluation to separate reconstruction from generalisation to a new condition.","source_ids":["src-discovery-altoslabs-perturbench"],"source_locator":"README: Downloading and Preparing Datasets, Data Splitting, Evaluator Class and Automated Evaluation"}],"limitations":[{"text":"A good score on average expression need not establish correct changes or cell-to-cell variation. Dose, time and cell context must match.","source_ids":["src-discovery-altoslabs-perturbench"],"source_locator":"README: Downloading and Preparing Datasets, Data Splitting, Evaluator Class and Automated Evaluation"}],"diagram":{"title":"Conceptual evaluation workflow","steps":["Select data and held-out conditions","Specify allowed controls and covariates","Predict molecular responses","Score declared aggregates and distributions"],"caption":"Conceptual task guide. Dataset preparation, parameters and scoring must come from a separately identified protocol.","source_ids":["src-discovery-altoslabs-perturbench"],"source_locator":"README: Downloading and Preparing Datasets, Data Splitting, Evaluator Class and Automated Evaluation"},"coverage":"reviewed","gaps":["A concrete evaluation still needs a protocol version, dataset release, eligible/scored counts and documented uncertainty. This guide does not manufacture those run-specific values."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Reviewed named primary documentation and existing protocol records. Editorial task scope is distinguished from source-reported procedures. No model execution or new numerical result; a reviewed guide is not a complete runnable protocol."}}}}
{"id":"catalog-task-cell-reference-mapping","kind":"benchmark","name":"Donor-held-out reference mapping","description":"","status":"discovered","facets":{"areas":["cells-tissues"]},"source_ids":["catalog-source-geneformer","catalog-source-scgpt","catalog-source-scfoundation","catalog-source-scvi"],"links":[],"attributes":{"entity_level":"task","version":null,"task":"Donor-held-out reference mapping","scope_note":"Map unseen donors to a labelled cell-type reference.","historical_missing_metadata":{"protocol_version":"not_yet_extracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Donor-held-out reference mapping asks whether a reference learned from some donors remains useful for cells from a new donor.","summary_source_ids":["evidence-official-0cbab80c2d1769287a1f","src-discovery-theislab-scib"],"summary_source_locator":"scVI model guide: generative model and latent representation; scIB README: biological-conservation metrics. Donor holdout is this guide’s evaluation requirement, not an asserted scVI default.","sections":[{"title":"Choosing an evaluation","body":"Keep donor identity visible throughout the evaluation. The reference and any fitted annotation rules belong to the training data; a query donor supplies the held-out assessment. Cell representations and transferred labels are different outputs and require different scoring. This is a proposed task definition, not a claim that every reference-mapping paper uses donor holdout.","source_ids":["evidence-official-0cbab80c2d1769287a1f","src-discovery-theislab-scib"],"source_locator":"scVI model guide: generative model and latent representation; scIB README: biological-conservation metrics. Donor holdout is this guide’s evaluation requirement, not an asserted scVI default."},{"title":"Task scope","body":"This is a task guide, not a single versioned benchmark protocol. The connected resources provide examples or concrete procedures. A candidate method or proposed control is not evidence that an evaluation has been completed.","source_ids":["evidence-official-0cbab80c2d1769287a1f","src-discovery-theislab-scib"],"source_locator":"scVI model guide: generative model and latent representation; scIB README: biological-conservation metrics. Donor holdout is this guide’s evaluation requirement, not an asserted scVI default."}],"facts":[{"label":"Entity type","value":"Task guide; concrete protocol identities remain separate.","status":"source_checked","source_ids":["evidence-official-0cbab80c2d1769287a1f","src-discovery-theislab-scib"],"source_locator":"scVI model guide: generative model and latent representation; scIB README: biological-conservation metrics. Donor holdout is this guide’s evaluation requirement, not an asserted scVI default."},{"label":"Datasets","value":"No single dataset is fixed by this guide. Select a linked protocol and its versioned data release.","status":"inapplicable","source_ids":["evidence-official-0cbab80c2d1769287a1f","src-discovery-theislab-scib"],"source_locator":"scVI model guide: generative model and latent representation; scIB README: biological-conservation metrics. Donor holdout is this guide’s evaluation requirement, not an asserted scVI default."},{"label":"Organisms","value":"No shared organism population is defined at this guide level. Record it for each selected dataset.","status":"inapplicable","source_ids":["evidence-official-0cbab80c2d1769287a1f","src-discovery-theislab-scib"],"source_locator":"scVI model guide: generative model and latent representation; scIB README: biological-conservation metrics. Donor holdout is this guide’s evaluation requirement, not an asserted scVI default."},{"label":"Assays","value":"No single measurement assay is fixed by this guide; the endpoint and assay belong to the selected protocol.","status":"inapplicable","source_ids":["evidence-official-0cbab80c2d1769287a1f","src-discovery-theislab-scib"],"source_locator":"scVI model guide: generative model and latent representation; scIB README: biological-conservation metrics. Donor holdout is this guide’s evaluation requirement, not an asserted scVI default."},{"label":"Splits","value":"This proposed guide requires donor separation; no particular donor assignment or cohort is fixed here.","status":"inapplicable","source_ids":["evidence-official-0cbab80c2d1769287a1f","src-discovery-theislab-scib"],"source_locator":"scVI model guide: generative model and latent representation; scIB README: biological-conservation metrics. Donor holdout is this guide’s evaluation requirement, not an asserted scVI default."},{"label":"Allowed inputs","value":"Reference and query molecular measurements, compatible feature identities, donor metadata and labels reserved for assessment.","status":"source_checked","source_ids":["evidence-official-0cbab80c2d1769287a1f","src-discovery-theislab-scib"],"source_locator":"scVI model guide: generative model and latent representation; scIB README: biological-conservation metrics. Donor holdout is this guide’s evaluation requirement, not an asserted scVI default."},{"label":"Adaptation","value":"No common fitting regime is imposed here. Keep pretrained, frozen, probed, fine-tuned and conventional methods distinct where applicable.","status":"inapplicable","source_ids":["evidence-official-0cbab80c2d1769287a1f","src-discovery-theislab-scib"],"source_locator":"scVI model guide: generative model and latent representation; scIB README: biological-conservation metrics. Donor holdout is this guide’s evaluation requirement, not an asserted scVI default."},{"label":"Metrics","value":"Choose label-transfer metrics for annotations and biological-conservation metrics for representations. The exact metric, averaging and unknown-cell handling belong to the concrete protocol.","status":"source_checked","source_ids":["evidence-official-0cbab80c2d1769287a1f","src-discovery-theislab-scib"],"source_locator":"scVI model guide: generative model and latent representation; scIB README: biological-conservation metrics. Donor holdout is this guide’s evaluation requirement, not an asserted scVI default."},{"label":"Baselines","value":"Specify a reference-only comparator with the same available labels and features. No measured baseline is assigned here.","status":"source_checked","source_ids":["evidence-official-0cbab80c2d1769287a1f","src-discovery-theislab-scib"],"source_locator":"scVI model guide: generative model and latent representation; scIB README: biological-conservation metrics. Donor holdout is this guide’s evaluation requirement, not an asserted scVI default."}],"strengths":[{"text":"Separating donors makes the intended generalisation question explicit and avoids confusing held-out cells with a new biological replicate.","source_ids":["evidence-official-0cbab80c2d1769287a1f","src-discovery-theislab-scib"],"source_locator":"scVI model guide: generative model and latent representation; scIB README: biological-conservation metrics. Donor holdout is this guide’s evaluation requirement, not an asserted scVI default."}],"limitations":[{"text":"A donor split alone does not control every difference in assay, tissue or cell composition. This guide does not define a particular cohort.","source_ids":["evidence-official-0cbab80c2d1769287a1f","src-discovery-theislab-scib"],"source_locator":"scVI model guide: generative model and latent representation; scIB README: biological-conservation metrics. Donor holdout is this guide’s evaluation requirement, not an asserted scVI default."}],"diagram":{"title":"Conceptual evaluation workflow","steps":["Define reference and query donors","Fit reference using permitted data","Map held-out query cells","Score labels and biological structure"],"caption":"Conceptual task guide. Dataset preparation, parameters and scoring must come from a separately identified protocol.","source_ids":["evidence-official-0cbab80c2d1769287a1f","src-discovery-theislab-scib"],"source_locator":"scVI model guide: generative model and latent representation; scIB README: biological-conservation metrics. Donor holdout is this guide’s evaluation requirement, not an asserted scVI default."},"coverage":"reviewed","gaps":["A concrete evaluation still needs a protocol version, dataset release, eligible/scored counts and documented uncertainty. This guide does not manufacture those run-specific values."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Reviewed named primary documentation and existing protocol records. Editorial task scope is distinguished from source-reported procedures. No model execution or new numerical result; a reviewed guide is not a complete runnable protocol."}}}}
{"id":"catalog-task-community-profiling","kind":"benchmark","name":"Community profiling","description":"","status":"discovered","facets":{"areas":["microbes-communities"]},"source_ids":["catalog-source-metaphlan"],"links":[],"attributes":{"entity_level":"task","version":null,"task":"Community profiling","scope_note":"Estimate taxon abundances in metagenomic samples.","historical_missing_metadata":{"protocol_version":"not_yet_extracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Community profiling estimates which microorganisms are present in a sample and their relative abundances.","summary_source_ids":["src-discovery-cami-challenge-opal"],"summary_source_locator":"README: Overview, Computed metrics and Inputs","sections":[{"title":"Choosing an evaluation","body":"Compare a predicted taxonomic profile with a specified reference profile at declared taxonomic ranks. OPAL separates detection errors from abundance errors and diversity summaries. The reference taxonomy, filtering and abundance normalisation must accompany a result; changing these can change what is scored.","source_ids":["src-discovery-cami-challenge-opal"],"source_locator":"README: Overview, Computed metrics and Inputs"},{"title":"Task scope","body":"This is a task guide, not a single versioned benchmark protocol. The connected resources provide examples or concrete procedures. A candidate method or proposed control is not evidence that an evaluation has been completed.","source_ids":["src-discovery-cami-challenge-opal"],"source_locator":"README: Overview, Computed metrics and Inputs"}],"facts":[{"label":"Entity type","value":"Task guide; concrete protocol identities remain separate.","status":"source_checked","source_ids":["src-discovery-cami-challenge-opal"],"source_locator":"README: Overview, Computed metrics and Inputs"},{"label":"Datasets","value":"No single dataset is fixed by this guide. Select a linked protocol and its versioned data release.","status":"inapplicable","source_ids":["src-discovery-cami-challenge-opal"],"source_locator":"README: Overview, Computed metrics and Inputs"},{"label":"Organisms","value":"No shared organism population is defined at this guide level. Record it for each selected dataset.","status":"inapplicable","source_ids":["src-discovery-cami-challenge-opal"],"source_locator":"README: Overview, Computed metrics and Inputs"},{"label":"Assays","value":"No single measurement assay is fixed by this guide; the endpoint and assay belong to the selected protocol.","status":"inapplicable","source_ids":["src-discovery-cami-challenge-opal"],"source_locator":"README: Overview, Computed metrics and Inputs"},{"label":"Splits","value":"No executable split is attached to this task identity. Use the selected protocol’s split manifest.","status":"inapplicable","source_ids":["src-discovery-cami-challenge-opal"],"source_locator":"README: Overview, Computed metrics and Inputs"},{"label":"Allowed inputs","value":"Predicted and reference taxonomic profiles with consistent taxon identifiers and abundance definitions.","status":"source_checked","source_ids":["src-discovery-cami-challenge-opal"],"source_locator":"README: Overview, Computed metrics and Inputs"},{"label":"Adaptation","value":"No common fitting regime is imposed here. Keep pretrained, frozen, probed, fine-tuned and conventional methods distinct where applicable.","status":"inapplicable","source_ids":["src-discovery-cami-challenge-opal"],"source_locator":"README: Overview, Computed metrics and Inputs"},{"label":"Metrics","value":"OPAL reports precision, recall, F1, abundance errors such as L1 and Bray–Curtis, and diversity measures. These are distinct endpoints.","status":"source_checked","source_ids":["src-discovery-cami-challenge-opal"],"source_locator":"README: Overview, Computed metrics and Inputs"},{"label":"Baselines","value":"Compare profilers against the same reference profile and taxonomic ranks; OPAL is an evaluator, not a predictive baseline.","status":"source_checked","source_ids":["src-discovery-cami-challenge-opal"],"source_locator":"README: Overview, Computed metrics and Inputs"}],"strengths":[{"text":"Separate presence and abundance metrics reveal different failure modes instead of hiding them behind a single accuracy number.","source_ids":["src-discovery-cami-challenge-opal"],"source_locator":"README: Overview, Computed metrics and Inputs"}],"limitations":[{"text":"A profile-level result is not a read-binning result. Reference coverage and taxonomic rank affect interpretation.","source_ids":["src-discovery-cami-challenge-opal"],"source_locator":"README: Overview, Computed metrics and Inputs"}],"diagram":{"title":"Conceptual evaluation workflow","steps":["Fix reference profile and taxonomy","Generate candidate abundance profiles","Align taxa and ranks","Score detection and abundance separately"],"caption":"Conceptual task guide. Dataset preparation, parameters and scoring must come from a separately identified protocol.","source_ids":["src-discovery-cami-challenge-opal"],"source_locator":"README: Overview, Computed metrics and Inputs"},"coverage":"reviewed","gaps":["A concrete evaluation still needs a protocol version, dataset release, eligible/scored counts and documented uncertainty. This guide does not manufacture those run-specific values."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Reviewed named primary documentation and existing protocol records. Editorial task scope is distinguished from source-reported procedures. No model execution or new numerical result; a reviewed guide is not a complete runnable protocol."}}}}
{"id":"catalog-task-complex-structure","kind":"benchmark","name":"Biomolecular complex structure","description":"","status":"discovered","facets":{"areas":["molecular-interactions"]},"source_ids":["catalog-source-chai-1","catalog-source-boltz-2","catalog-source-alphafold-3-server"],"links":[],"attributes":{"entity_level":"task","version":null,"task":"Biomolecular complex structure","scope_note":"Predict joint structure for interacting proteins and other molecules.","historical_missing_metadata":{"protocol_version":"not_yet_extracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Complex-structure evaluation tests predicted molecular arrangements against experimentally determined structures.","summary_source_ids":["evidence-alphafold-paper"],"summary_source_locator":"Main text: Model architecture and Model limitations; Methods: Metrics and Recent PDB evaluation set","sections":[{"title":"Choosing an evaluation","body":"Declare the molecular partners and the input information available to each method. Assess the relevant chains and interfaces as well as the whole complex, using an explicit chain-assignment rule. Templates, alignments and sampling budgets belong to the evaluated configuration. A confidence estimate is not an experimental accuracy measurement.","source_ids":["evidence-alphafold-paper"],"source_locator":"Main text: Model architecture and Model limitations; Methods: Metrics and Recent PDB evaluation set"},{"title":"Task scope","body":"This is a task guide, not a single versioned benchmark protocol. The connected resources provide examples or concrete procedures. A candidate method or proposed control is not evidence that an evaluation has been completed.","source_ids":["evidence-alphafold-paper"],"source_locator":"Main text: Model architecture and Model limitations; Methods: Metrics and Recent PDB evaluation set"}],"facts":[{"label":"Entity type","value":"Task guide; concrete protocol identities remain separate.","status":"source_checked","source_ids":["evidence-alphafold-paper"],"source_locator":"Main text: Model architecture and Model limitations; Methods: Metrics and Recent PDB evaluation set"},{"label":"Datasets","value":"No single dataset is fixed by this guide. Select a linked protocol and its versioned data release.","status":"inapplicable","source_ids":["evidence-alphafold-paper"],"source_locator":"Main text: Model architecture and Model limitations; Methods: Metrics and Recent PDB evaluation set"},{"label":"Organisms","value":"No shared organism population is defined at this guide level. Record it for each selected dataset.","status":"inapplicable","source_ids":["evidence-alphafold-paper"],"source_locator":"Main text: Model architecture and Model limitations; Methods: Metrics and Recent PDB evaluation set"},{"label":"Assays","value":"No single measurement assay is fixed by this guide; the endpoint and assay belong to the selected protocol.","status":"inapplicable","source_ids":["evidence-alphafold-paper"],"source_locator":"Main text: Model architecture and Model limitations; Methods: Metrics and Recent PDB evaluation set"},{"label":"Splits","value":"No executable split is attached to this task identity. Use the selected protocol’s split manifest.","status":"inapplicable","source_ids":["evidence-alphafold-paper"],"source_locator":"Main text: Model architecture and Model limitations; Methods: Metrics and Recent PDB evaluation set"},{"label":"Allowed inputs","value":"Molecular sequences and chemical identities, with templates or alignments only where the protocol permits them; experimental reference structures for scoring.","status":"source_checked","source_ids":["evidence-alphafold-paper"],"source_locator":"Main text: Model architecture and Model limitations; Methods: Metrics and Recent PDB evaluation set"},{"label":"Adaptation","value":"No common fitting regime is imposed here. Keep pretrained, frozen, probed, fine-tuned and conventional methods distinct where applicable.","status":"inapplicable","source_ids":["evidence-alphafold-paper"],"source_locator":"Main text: Model architecture and Model limitations; Methods: Metrics and Recent PDB evaluation set"},{"label":"Metrics","value":"Protocol-specific structure and interface measures, such as LDDT, DockQ or interface LDDT. State the assessed entities and aggregation.","status":"source_checked","source_ids":["evidence-alphafold-paper"],"source_locator":"Main text: Model architecture and Model limitations; Methods: Metrics and Recent PDB evaluation set"},{"label":"Baselines","value":"Compare structure predictors with matched partner definitions and input information. Different sampling budgets require separate reporting.","status":"source_checked","source_ids":["evidence-alphafold-paper"],"source_locator":"Main text: Model architecture and Model limitations; Methods: Metrics and Recent PDB evaluation set"}],"strengths":[{"text":"Interface-specific scoring can reveal errors hidden by an otherwise accurate large component.","source_ids":["evidence-alphafold-paper"],"source_locator":"Main text: Model architecture and Model limitations; Methods: Metrics and Recent PDB evaluation set"}],"limitations":[{"text":"Agreement with one reference structure does not establish dynamics, binding affinity or every possible conformational state.","source_ids":["evidence-alphafold-paper"],"source_locator":"Main text: Model architecture and Model limitations; Methods: Metrics and Recent PDB evaluation set"}],"diagram":{"title":"Conceptual evaluation workflow","steps":["Fix partners, references and allowed inputs","Predict complete complexes","Match equivalent chains and atoms","Score structures and interfaces"],"caption":"Conceptual task guide. Dataset preparation, parameters and scoring must come from a separately identified protocol.","source_ids":["evidence-alphafold-paper"],"source_locator":"Main text: Model architecture and Model limitations; Methods: Metrics and Recent PDB evaluation set"},"coverage":"reviewed","gaps":["A concrete evaluation still needs a protocol version, dataset release, eligible/scored counts and documented uncertainty. This guide does not manufacture those run-specific values."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Reviewed named primary documentation and existing protocol records. Editorial task scope is distinguished from source-reported procedures. No model execution or new numerical result; a reviewed guide is not a complete runnable protocol."}}}}
{"id":"catalog-task-enhancer-effects","kind":"benchmark","name":"Enhancer / MPRA effects","description":"","status":"discovered","facets":{"areas":["dna-genomes"]},"source_ids":["catalog-source-dnabert-2","catalog-source-nt-v2","catalog-source-evo-2","catalog-source-alphagenome"],"links":[],"attributes":{"entity_level":"task","version":null,"task":"Enhancer / MPRA effects","scope_note":"Predict measured activity changes from regulatory sequence variants.","historical_missing_metadata":{"protocol_version":"not_yet_extracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Enhancer-effect evaluation asks whether sequence-based predictions track measured changes in regulatory activity.","summary_source_ids":["src-discovery-kundajelab-dart-eval","dart-eval-regulatory-2024"],"summary_source_locator":"DART-Eval README Task 5; paper evaluation tasks and ab initio baselines","sections":[{"title":"Choosing an evaluation","body":"Separate detection of regulatory DNA from prediction of the effect of a sequence change. A reporter assay and endogenous chromatin-accessibility assay are different endpoints. DART-Eval provides concrete regulatory prediction tasks and comparisons across frozen, probed and fine-tuned models. Use its exact task and data release rather than treating every enhancer test as equivalent.","source_ids":["src-discovery-kundajelab-dart-eval","dart-eval-regulatory-2024"],"source_locator":"DART-Eval README Task 5; paper evaluation tasks and ab initio baselines"},{"title":"Task scope","body":"This is a task guide, not a single versioned benchmark protocol. The connected resources provide examples or concrete procedures. A candidate method or proposed control is not evidence that an evaluation has been completed.","source_ids":["src-discovery-kundajelab-dart-eval","dart-eval-regulatory-2024"],"source_locator":"DART-Eval README Task 5; paper evaluation tasks and ab initio baselines"}],"facts":[{"label":"Entity type","value":"Task guide; concrete protocol identities remain separate.","status":"source_checked","source_ids":["src-discovery-kundajelab-dart-eval","dart-eval-regulatory-2024"],"source_locator":"DART-Eval README Task 5; paper evaluation tasks and ab initio baselines"},{"label":"Datasets","value":"No single dataset is fixed by this guide. Select a linked protocol and its versioned data release.","status":"inapplicable","source_ids":["src-discovery-kundajelab-dart-eval","dart-eval-regulatory-2024"],"source_locator":"DART-Eval README Task 5; paper evaluation tasks and ab initio baselines"},{"label":"Organisms","value":"No shared organism population is defined at this guide level. Record it for each selected dataset.","status":"inapplicable","source_ids":["src-discovery-kundajelab-dart-eval","dart-eval-regulatory-2024"],"source_locator":"DART-Eval README Task 5; paper evaluation tasks and ab initio baselines"},{"label":"Assays","value":"No single measurement assay is fixed by this guide; the endpoint and assay belong to the selected protocol.","status":"inapplicable","source_ids":["src-discovery-kundajelab-dart-eval","dart-eval-regulatory-2024"],"source_locator":"DART-Eval README Task 5; paper evaluation tasks and ab initio baselines"},{"label":"Splits","value":"No executable split is attached to this task identity. Use the selected protocol’s split manifest.","status":"inapplicable","source_ids":["src-discovery-kundajelab-dart-eval","dart-eval-regulatory-2024"],"source_locator":"DART-Eval README Task 5; paper evaluation tasks and ab initio baselines"},{"label":"Allowed inputs","value":"Regulatory DNA or matched alleles and a specified experimental activity endpoint, with genome assembly and window placement recorded.","status":"source_checked","source_ids":["src-discovery-kundajelab-dart-eval","dart-eval-regulatory-2024"],"source_locator":"DART-Eval README Task 5; paper evaluation tasks and ab initio baselines"},{"label":"Adaptation","value":"No common fitting regime is imposed here. Keep pretrained, frozen, probed, fine-tuned and conventional methods distinct where applicable.","status":"inapplicable","source_ids":["src-discovery-kundajelab-dart-eval","dart-eval-regulatory-2024"],"source_locator":"DART-Eval README Task 5; paper evaluation tasks and ab initio baselines"},{"label":"Metrics","value":"Use the metric attached to the concrete task: regulatory-element classification and quantitative effect prediction require different scoring.","status":"source_checked","source_ids":["src-discovery-kundajelab-dart-eval","dart-eval-regulatory-2024"],"source_locator":"DART-Eval README Task 5; paper evaluation tasks and ab initio baselines"},{"label":"Baselines","value":"DART-Eval includes supervised models trained from scratch, including ChromBPNet-related baselines, alongside language-model evaluation modes.","status":"source_checked","source_ids":["src-discovery-kundajelab-dart-eval","dart-eval-regulatory-2024"],"source_locator":"DART-Eval README Task 5; paper evaluation tasks and ab initio baselines"}],"strengths":[{"text":"A functional endpoint and matched learned baselines test usefulness beyond sequence-likelihood differences alone.","source_ids":["src-discovery-kundajelab-dart-eval","dart-eval-regulatory-2024"],"source_locator":"DART-Eval README Task 5; paper evaluation tasks and ab initio baselines"}],"limitations":[{"text":"Results from one reporter design or cellular context do not automatically transfer to endogenous regulation elsewhere.","source_ids":["src-discovery-kundajelab-dart-eval","dart-eval-regulatory-2024"],"source_locator":"DART-Eval README Task 5; paper evaluation tasks and ab initio baselines"}],"diagram":{"title":"Conceptual evaluation workflow","steps":["Choose assay endpoint and split","Define DNA windows and alleles","Apply declared prediction configuration","Compare scores with measured effects"],"caption":"Conceptual task guide. Dataset preparation, parameters and scoring must come from a separately identified protocol.","source_ids":["src-discovery-kundajelab-dart-eval","dart-eval-regulatory-2024"],"source_locator":"DART-Eval README Task 5; paper evaluation tasks and ab initio baselines"},"coverage":"reviewed","gaps":["A concrete evaluation still needs a protocol version, dataset release, eligible/scored counts and documented uncertainty. This guide does not manufacture those run-specific values."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Reviewed named primary documentation and existing protocol records. Editorial task scope is distinguished from source-reported procedures. No model execution or new numerical result; a reviewed guide is not a complete runnable protocol."}}}}
{"id":"catalog-task-heldout-clade","kind":"benchmark","name":"Held-out-clade classification","description":"","status":"discovered","facets":{"areas":["microbes-communities"]},"source_ids":["catalog-source-prokbert","catalog-source-metagene-1","catalog-source-evo-2","catalog-source-kraken2"],"links":[],"attributes":{"entity_level":"task","version":null,"task":"Held-out-clade classification","scope_note":"Hold clades out of downstream fitting and reference databases; evaluate known ancestor labels or unknown-taxon detection, and audit pretraining overlap separately.","historical_missing_metadata":{"protocol_version":"not_yet_extracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Held-out-clade classification asks how well sequence-based classification works when a defined taxonomic group is absent from training.","summary_source_ids":["barcodebert-2026"],"summary_source_locator":"Methods: dataset partition and 1-NN probing; Results: unseen-species genus prediction","sections":[{"title":"Choosing an evaluation","body":"Specify both the taxonomic level that is withheld and the level that is predicted. BarcodeBERT’s unseen-species genus probe illustrates why these differ: species can be unseen while their genera remain represented. A benchmark must document which reference sequences and labels remain available before interpreting a result as taxonomic generalisation.","source_ids":["barcodebert-2026"],"source_locator":"Methods: dataset partition and 1-NN probing; Results: unseen-species genus prediction"},{"title":"Task scope","body":"This is a task guide, not a single versioned benchmark protocol. The connected resources provide examples or concrete procedures. A candidate method or proposed control is not evidence that an evaluation has been completed.","source_ids":["barcodebert-2026"],"source_locator":"Methods: dataset partition and 1-NN probing; Results: unseen-species genus prediction"}],"facts":[{"label":"Entity type","value":"Task guide; concrete protocol identities remain separate.","status":"source_checked","source_ids":["barcodebert-2026"],"source_locator":"Methods: dataset partition and 1-NN probing; Results: unseen-species genus prediction"},{"label":"Datasets","value":"No single dataset is fixed by this guide. Select a linked protocol and its versioned data release.","status":"inapplicable","source_ids":["barcodebert-2026"],"source_locator":"Methods: dataset partition and 1-NN probing; Results: unseen-species genus prediction"},{"label":"Organisms","value":"No shared organism population is defined at this guide level. Record it for each selected dataset.","status":"inapplicable","source_ids":["barcodebert-2026"],"source_locator":"Methods: dataset partition and 1-NN probing; Results: unseen-species genus prediction"},{"label":"Assays","value":"No single measurement assay is fixed by this guide; the endpoint and assay belong to the selected protocol.","status":"inapplicable","source_ids":["barcodebert-2026"],"source_locator":"Methods: dataset partition and 1-NN probing; Results: unseen-species genus prediction"},{"label":"Splits","value":"No executable split is attached to this task identity. Use the selected protocol’s split manifest.","status":"inapplicable","source_ids":["barcodebert-2026"],"source_locator":"Methods: dataset partition and 1-NN probing; Results: unseen-species genus prediction"},{"label":"Allowed inputs","value":"Labelled reference sequences and a taxonomically held-out query set; exact rank and reference-library membership are protocol-specific.","status":"source_checked","source_ids":["barcodebert-2026"],"source_locator":"Methods: dataset partition and 1-NN probing; Results: unseen-species genus prediction"},{"label":"Adaptation","value":"No common fitting regime is imposed here. Keep pretrained, frozen, probed, fine-tuned and conventional methods distinct where applicable.","status":"inapplicable","source_ids":["barcodebert-2026"],"source_locator":"Methods: dataset partition and 1-NN probing; Results: unseen-species genus prediction"},{"label":"Metrics","value":"Classification accuracy or other declared label metrics at the target rank; results at different taxonomic ranks are not interchangeable.","status":"source_checked","source_ids":["barcodebert-2026"],"source_locator":"Methods: dataset partition and 1-NN probing; Results: unseen-species genus prediction"},{"label":"Baselines","value":"The BarcodeBERT protocol includes nearest-neighbour evaluation of sequence representations. A proposed alternative comparator is not a completed result.","status":"source_checked","source_ids":["barcodebert-2026"],"source_locator":"Methods: dataset partition and 1-NN probing; Results: unseen-species genus prediction"}],"strengths":[{"text":"Explicit taxonomic holdout distinguishes close-reference matching from transfer to less familiar sequences.","source_ids":["barcodebert-2026"],"source_locator":"Methods: dataset partition and 1-NN probing; Results: unseen-species genus prediction"}],"limitations":[{"text":"Unseen-species performance within known genera is not evidence for classifying entirely novel genera.","source_ids":["barcodebert-2026"],"source_locator":"Methods: dataset partition and 1-NN probing; Results: unseen-species genus prediction"}],"diagram":{"title":"Conceptual evaluation workflow","steps":["Choose held-out and predicted ranks","Construct reference and query sets","Apply the declared classifier or probe","Score at the specified taxonomic rank"],"caption":"Conceptual task guide. Dataset preparation, parameters and scoring must come from a separately identified protocol.","source_ids":["barcodebert-2026"],"source_locator":"Methods: dataset partition and 1-NN probing; Results: unseen-species genus prediction"},"coverage":"reviewed","gaps":["A concrete evaluation still needs a protocol version, dataset release, eligible/scored counts and documented uncertainty. This guide does not manufacture those run-specific values."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Reviewed named primary documentation and existing protocol records. Editorial task scope is distinguished from source-reported procedures. No model execution or new numerical result; a reviewed guide is not a complete runnable protocol."}}}}
{"id":"catalog-task-ligand-affinity","kind":"benchmark","name":"Small-molecule affinity","description":"","status":"discovered","facets":{"areas":["molecular-interactions"]},"source_ids":["catalog-source-boltz-2"],"links":[],"attributes":{"entity_level":"task","version":null,"task":"Small-molecule affinity","scope_note":"Predict measured binding affinity; pose confidence is not an affinity value.","historical_missing_metadata":{"protocol_version":"not_yet_extracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Affinity evaluation tests predictions of molecular binding measurements or binder labels, depending on the protocol.","summary_source_ids":["evidence-official-5a9e55abdf288d9dd8da","src-discovery-plinder-org-plinder"],"summary_source_locator":"Boltz docs/prediction.md: Properties (affinity), Affinity prediction outputs; PLINDER README: Known bugs","sections":[{"title":"Choosing an evaluation","body":"Define the endpoint before comparing numbers. Boltz-2 distinguishes a binder-versus-decoy probability from a quantitative affinity output, with different supervision. Preserve measurement units, transforms and assay conditions. Structural plausibility alone does not validate affinity; a structural dataset also needs checked affinity labels before it can support this task.","source_ids":["evidence-official-5a9e55abdf288d9dd8da","src-discovery-plinder-org-plinder"],"source_locator":"Boltz docs/prediction.md: Properties (affinity), Affinity prediction outputs; PLINDER README: Known bugs"},{"title":"Task scope","body":"This is a task guide, not a single versioned benchmark protocol. The connected resources provide examples or concrete procedures. A candidate method or proposed control is not evidence that an evaluation has been completed.","source_ids":["evidence-official-5a9e55abdf288d9dd8da","src-discovery-plinder-org-plinder"],"source_locator":"Boltz docs/prediction.md: Properties (affinity), Affinity prediction outputs; PLINDER README: Known bugs"}],"facts":[{"label":"Entity type","value":"Task guide; concrete protocol identities remain separate.","status":"source_checked","source_ids":["evidence-official-5a9e55abdf288d9dd8da","src-discovery-plinder-org-plinder"],"source_locator":"Boltz docs/prediction.md: Properties (affinity), Affinity prediction outputs; PLINDER README: Known bugs"},{"label":"Datasets","value":"No single dataset is fixed by this guide. Select a linked protocol and its versioned data release.","status":"inapplicable","source_ids":["evidence-official-5a9e55abdf288d9dd8da","src-discovery-plinder-org-plinder"],"source_locator":"Boltz docs/prediction.md: Properties (affinity), Affinity prediction outputs; PLINDER README: Known bugs"},{"label":"Organisms","value":"No shared organism population is defined at this guide level. Record it for each selected dataset.","status":"inapplicable","source_ids":["evidence-official-5a9e55abdf288d9dd8da","src-discovery-plinder-org-plinder"],"source_locator":"Boltz docs/prediction.md: Properties (affinity), Affinity prediction outputs; PLINDER README: Known bugs"},{"label":"Assays","value":"No single measurement assay is fixed by this guide; the endpoint and assay belong to the selected protocol.","status":"inapplicable","source_ids":["evidence-official-5a9e55abdf288d9dd8da","src-discovery-plinder-org-plinder"],"source_locator":"Boltz docs/prediction.md: Properties (affinity), Affinity prediction outputs; PLINDER README: Known bugs"},{"label":"Splits","value":"No executable split is attached to this task identity. Use the selected protocol’s split manifest.","status":"inapplicable","source_ids":["evidence-official-5a9e55abdf288d9dd8da","src-discovery-plinder-org-plinder"],"source_locator":"Boltz docs/prediction.md: Properties (affinity), Affinity prediction outputs; PLINDER README: Known bugs"},{"label":"Allowed inputs","value":"A protein–small-molecule pair plus any permitted structural information; matched experimental measurements or declared binary labels.","status":"source_checked","source_ids":["evidence-official-5a9e55abdf288d9dd8da","src-discovery-plinder-org-plinder"],"source_locator":"Boltz docs/prediction.md: Properties (affinity), Affinity prediction outputs; PLINDER README: Known bugs"},{"label":"Adaptation","value":"No common fitting regime is imposed here. Keep pretrained, frozen, probed, fine-tuned and conventional methods distinct where applicable.","status":"inapplicable","source_ids":["evidence-official-5a9e55abdf288d9dd8da","src-discovery-plinder-org-plinder"],"source_locator":"Boltz docs/prediction.md: Properties (affinity), Affinity prediction outputs; PLINDER README: Known bugs"},{"label":"Metrics","value":"Use endpoint-appropriate regression or classification metrics. Do not merge differently transformed affinity quantities into one table without documenting conversions.","status":"source_checked","source_ids":["evidence-official-5a9e55abdf288d9dd8da","src-discovery-plinder-org-plinder"],"source_locator":"Boltz docs/prediction.md: Properties (affinity), Affinity prediction outputs; PLINDER README: Known bugs"},{"label":"Baselines","value":"Select comparators evaluated on the same measured endpoint and split; a docking score is not automatically a calibrated affinity measurement.","status":"source_checked","source_ids":["evidence-official-5a9e55abdf288d9dd8da","src-discovery-plinder-org-plinder"],"source_locator":"Boltz docs/prediction.md: Properties (affinity), Affinity prediction outputs; PLINDER README: Known bugs"}],"strengths":[{"text":"Separating binding classification from quantitative affinity avoids conflating two different uses of a model output.","source_ids":["evidence-official-5a9e55abdf288d9dd8da","src-discovery-plinder-org-plinder"],"source_locator":"Boltz docs/prediction.md: Properties (affinity), Affinity prediction outputs; PLINDER README: Known bugs"}],"limitations":[{"text":"The inspected PLINDER revision flags its affinity query as disabled after a parsing bug. A dataset name alone does not verify an affinity label.","source_ids":["evidence-official-5a9e55abdf288d9dd8da","src-discovery-plinder-org-plinder"],"source_locator":"Boltz docs/prediction.md: Properties (affinity), Affinity prediction outputs; PLINDER README: Known bugs"}],"diagram":{"title":"Conceptual evaluation workflow","steps":["Fix assay endpoint, units and split","Specify model inputs","Predict the declared affinity quantity","Score matched measurements or labels"],"caption":"Conceptual task guide. Dataset preparation, parameters and scoring must come from a separately identified protocol.","source_ids":["evidence-official-5a9e55abdf288d9dd8da","src-discovery-plinder-org-plinder"],"source_locator":"Boltz docs/prediction.md: Properties (affinity), Affinity prediction outputs; PLINDER README: Known bugs"},"coverage":"reviewed","gaps":["A concrete evaluation still needs a protocol version, dataset release, eligible/scored counts and documented uncertainty. This guide does not manufacture those run-specific values."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Reviewed named primary documentation and existing protocol records. Editorial task scope is distinguished from source-reported procedures. No model execution or new numerical result; a reviewed guide is not a complete runnable protocol."}}}}
{"id":"catalog-task-ligand-pose","kind":"benchmark","name":"Protein–ligand pose","description":"","status":"discovered","facets":{"areas":["molecular-interactions"]},"source_ids":["catalog-source-chai-1","catalog-source-boltz-2","catalog-source-diffdock-l","catalog-source-alphafold-3-server","catalog-source-vina"],"links":[],"attributes":{"entity_level":"task","version":null,"task":"Protein–ligand pose","scope_note":"Predict the bound ligand geometry from prepared molecular inputs.","historical_missing_metadata":{"protocol_version":"not_yet_extracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Protein–ligand pose evaluation checks both agreement with a reference pose and the plausibility of the predicted geometry.","summary_source_ids":["src-discovery-maabuu-posebusters","src-discovery-plinder-org-plinder"],"summary_source_locator":"PoseBusters README: Usage; PLINDER README: Gold standard benchmark sets and Plinder versions","sections":[{"title":"Choosing an evaluation","body":"Record whether a method receives an experimental bound protein, an unbound structure or a predicted receptor. PLINDER supplies related systems and versioned splits; PoseBusters supplies plausibility checks. These resources have different roles. Keep geometric validity separate from pose recovery and state any filtering of failed predictions.","source_ids":["src-discovery-maabuu-posebusters","src-discovery-plinder-org-plinder"],"source_locator":"PoseBusters README: Usage; PLINDER README: Gold standard benchmark sets and Plinder versions"},{"title":"Task scope","body":"This is a task guide, not a single versioned benchmark protocol. The connected resources provide examples or concrete procedures. A candidate method or proposed control is not evidence that an evaluation has been completed.","source_ids":["src-discovery-maabuu-posebusters","src-discovery-plinder-org-plinder"],"source_locator":"PoseBusters README: Usage; PLINDER README: Gold standard benchmark sets and Plinder versions"}],"facts":[{"label":"Entity type","value":"Task guide; concrete protocol identities remain separate.","status":"source_checked","source_ids":["src-discovery-maabuu-posebusters","src-discovery-plinder-org-plinder"],"source_locator":"PoseBusters README: Usage; PLINDER README: Gold standard benchmark sets and Plinder versions"},{"label":"Datasets","value":"No single dataset is fixed by this guide. Select a linked protocol and its versioned data release.","status":"inapplicable","source_ids":["src-discovery-maabuu-posebusters","src-discovery-plinder-org-plinder"],"source_locator":"PoseBusters README: Usage; PLINDER README: Gold standard benchmark sets and Plinder versions"},{"label":"Organisms","value":"No shared organism population is defined at this guide level. Record it for each selected dataset.","status":"inapplicable","source_ids":["src-discovery-maabuu-posebusters","src-discovery-plinder-org-plinder"],"source_locator":"PoseBusters README: Usage; PLINDER README: Gold standard benchmark sets and Plinder versions"},{"label":"Assays","value":"No single measurement assay is fixed by this guide; the endpoint and assay belong to the selected protocol.","status":"inapplicable","source_ids":["src-discovery-maabuu-posebusters","src-discovery-plinder-org-plinder"],"source_locator":"PoseBusters README: Usage; PLINDER README: Gold standard benchmark sets and Plinder versions"},{"label":"Splits","value":"No executable split is attached to this task identity. Use the selected protocol’s split manifest.","status":"inapplicable","source_ids":["src-discovery-maabuu-posebusters","src-discovery-plinder-org-plinder"],"source_locator":"PoseBusters README: Usage; PLINDER README: Gold standard benchmark sets and Plinder versions"},{"label":"Allowed inputs","value":"Predicted ligand coordinates, receptor information allowed by the protocol, and an experimental reference pose where pose recovery is scored.","status":"source_checked","source_ids":["src-discovery-maabuu-posebusters","src-discovery-plinder-org-plinder"],"source_locator":"PoseBusters README: Usage; PLINDER README: Gold standard benchmark sets and Plinder versions"},{"label":"Adaptation","value":"No common fitting regime is imposed here. Keep pretrained, frozen, probed, fine-tuned and conventional methods distinct where applicable.","status":"inapplicable","source_ids":["src-discovery-maabuu-posebusters","src-discovery-plinder-org-plinder"],"source_locator":"PoseBusters README: Usage; PLINDER README: Gold standard benchmark sets and Plinder versions"},{"label":"Metrics","value":"Reference-pose error and explicitly versioned plausibility checks. A validity pass without a matching pose is a different outcome from correct pose recovery.","status":"source_checked","source_ids":["src-discovery-maabuu-posebusters","src-discovery-plinder-org-plinder"],"source_locator":"PoseBusters README: Usage; PLINDER README: Gold standard benchmark sets and Plinder versions"},{"label":"Baselines","value":"Use classical docking or learned comparators only under matched receptor and pocket information. Candidate applicability does not establish a completed run.","status":"source_checked","source_ids":["src-discovery-maabuu-posebusters","src-discovery-plinder-org-plinder"],"source_locator":"PoseBusters README: Usage; PLINDER README: Gold standard benchmark sets and Plinder versions"}],"strengths":[{"text":"Plausibility checks catch physically problematic outputs that a geometric alignment score alone may miss.","source_ids":["src-discovery-maabuu-posebusters","src-discovery-plinder-org-plinder"],"source_locator":"PoseBusters README: Usage; PLINDER README: Gold standard benchmark sets and Plinder versions"}],"limitations":[{"text":"Results depend on receptor state, reference quality and failure handling. Pose accuracy does not establish binding affinity.","source_ids":["src-discovery-maabuu-posebusters","src-discovery-plinder-org-plinder"],"source_locator":"PoseBusters README: Usage; PLINDER README: Gold standard benchmark sets and Plinder versions"}],"diagram":{"title":"Conceptual evaluation workflow","steps":["Choose receptor state and split","Predict ligand pose","Check reference-pose agreement","Check geometry and report failures"],"caption":"Conceptual task guide. Dataset preparation, parameters and scoring must come from a separately identified protocol.","source_ids":["src-discovery-maabuu-posebusters","src-discovery-plinder-org-plinder"],"source_locator":"PoseBusters README: Usage; PLINDER README: Gold standard benchmark sets and Plinder versions"},"coverage":"reviewed","gaps":["A concrete evaluation still needs a protocol version, dataset release, eligible/scored counts and documented uncertainty. This guide does not manufacture those run-specific values."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Reviewed named primary documentation and existing protocol records. Editorial task scope is distinguished from source-reported procedures. No model execution or new numerical result; a reviewed guide is not a complete runnable protocol."}}}}
{"id":"catalog-task-long-range-regulation","kind":"benchmark","name":"Long-range regulation","description":"","status":"discovered","facets":{"areas":["dna-genomes"]},"source_ids":["catalog-source-evo-2","catalog-source-alphagenome"],"links":[],"attributes":{"entity_level":"task","version":null,"task":"Long-range regulation","scope_note":"Predict gene-expression or chromatin effects from long-context DNA sequence.","historical_missing_metadata":{"protocol_version":"not_yet_extracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Long-range regulatory evaluation asks whether useful predictions depend on information beyond a short local DNA window.","summary_source_ids":["src-discovery-frederikkemarin-bend","dart-eval-regulatory-2024"],"summary_source_locator":"BEND README: Data format and training/evaluating supervised models; DART-Eval paper: task and context distinctions","sections":[{"title":"Choosing an evaluation","body":"A large supported input window is a model capability, not by itself a test of distant regulation. Select a concrete regulatory endpoint and retain genomic coordinates, split membership and permitted context. BEND’s coordinate-based task format illustrates why those details matter. Compare configurations only when their prediction target and available information are explicit.","source_ids":["src-discovery-frederikkemarin-bend","dart-eval-regulatory-2024"],"source_locator":"BEND README: Data format and training/evaluating supervised models; DART-Eval paper: task and context distinctions"},{"title":"Task scope","body":"This is a task guide, not a single versioned benchmark protocol. The connected resources provide examples or concrete procedures. A candidate method or proposed control is not evidence that an evaluation has been completed.","source_ids":["src-discovery-frederikkemarin-bend","dart-eval-regulatory-2024"],"source_locator":"BEND README: Data format and training/evaluating supervised models; DART-Eval paper: task and context distinctions"}],"facts":[{"label":"Entity type","value":"Task guide; concrete protocol identities remain separate.","status":"source_checked","source_ids":["src-discovery-frederikkemarin-bend","dart-eval-regulatory-2024"],"source_locator":"BEND README: Data format and training/evaluating supervised models; DART-Eval paper: task and context distinctions"},{"label":"Datasets","value":"No single dataset is fixed by this guide. Select a linked protocol and its versioned data release.","status":"inapplicable","source_ids":["src-discovery-frederikkemarin-bend","dart-eval-regulatory-2024"],"source_locator":"BEND README: Data format and training/evaluating supervised models; DART-Eval paper: task and context distinctions"},{"label":"Organisms","value":"No shared organism population is defined at this guide level. Record it for each selected dataset.","status":"inapplicable","source_ids":["src-discovery-frederikkemarin-bend","dart-eval-regulatory-2024"],"source_locator":"BEND README: Data format and training/evaluating supervised models; DART-Eval paper: task and context distinctions"},{"label":"Assays","value":"No single measurement assay is fixed by this guide; the endpoint and assay belong to the selected protocol.","status":"inapplicable","source_ids":["src-discovery-frederikkemarin-bend","dart-eval-regulatory-2024"],"source_locator":"BEND README: Data format and training/evaluating supervised models; DART-Eval paper: task and context distinctions"},{"label":"Splits","value":"No executable split is attached to this task identity. Use the selected protocol’s split manifest.","status":"inapplicable","source_ids":["src-discovery-frederikkemarin-bend","dart-eval-regulatory-2024"],"source_locator":"BEND README: Data format and training/evaluating supervised models; DART-Eval paper: task and context distinctions"},{"label":"Allowed inputs","value":"Genomic sequences with coordinates, a declared context window and the labels of a specific regulatory task.","status":"source_checked","source_ids":["src-discovery-frederikkemarin-bend","dart-eval-regulatory-2024"],"source_locator":"BEND README: Data format and training/evaluating supervised models; DART-Eval paper: task and context distinctions"},{"label":"Adaptation","value":"No common fitting regime is imposed here. Keep pretrained, frozen, probed, fine-tuned and conventional methods distinct where applicable.","status":"inapplicable","source_ids":["src-discovery-frederikkemarin-bend","dart-eval-regulatory-2024"],"source_locator":"BEND README: Data format and training/evaluating supervised models; DART-Eval paper: task and context distinctions"},{"label":"Metrics","value":"Use the concrete task’s metric and aggregation; there is no single score for all long-range regulation.","status":"source_checked","source_ids":["src-discovery-frederikkemarin-bend","dart-eval-regulatory-2024"],"source_locator":"BEND README: Data format and training/evaluating supervised models; DART-Eval paper: task and context distinctions"},{"label":"Baselines","value":"Matched short-context or task-specific methods can be proposed as controls, but require an explicit protocol and measured results before comparison.","status":"source_checked","source_ids":["src-discovery-frederikkemarin-bend","dart-eval-regulatory-2024"],"source_locator":"BEND README: Data format and training/evaluating supervised models; DART-Eval paper: task and context distinctions"}],"strengths":[{"text":"Recording context and coordinates makes the intended information advantage inspectable.","source_ids":["src-discovery-frederikkemarin-bend","dart-eval-regulatory-2024"],"source_locator":"BEND README: Data format and training/evaluating supervised models; DART-Eval paper: task and context distinctions"}],"limitations":[{"text":"This guide does not establish that longer context improves a model, nor does it provide a shared enhancer–gene or contact-map protocol.","source_ids":["src-discovery-frederikkemarin-bend","dart-eval-regulatory-2024"],"source_locator":"BEND README: Data format and training/evaluating supervised models; DART-Eval paper: task and context distinctions"}],"diagram":{"title":"Conceptual evaluation workflow","steps":["Define endpoint and genomic split","Declare local and distant information","Evaluate specified configurations","Interpret scores within that endpoint"],"caption":"Conceptual task guide. Dataset preparation, parameters and scoring must come from a separately identified protocol.","source_ids":["src-discovery-frederikkemarin-bend","dart-eval-regulatory-2024"],"source_locator":"BEND README: Data format and training/evaluating supervised models; DART-Eval paper: task and context distinctions"},"coverage":"reviewed","gaps":["A concrete evaluation still needs a protocol version, dataset release, eligible/scored counts and documented uncertainty. This guide does not manufacture those run-specific values."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Reviewed named primary documentation and existing protocol records. Editorial task scope is distinguished from source-reported procedures. No model execution or new numerical result; a reviewed guide is not a complete runnable protocol."}}}}
{"id":"catalog-task-mfass-splice","kind":"benchmark","name":"MFASS splice-variant prioritisation","description":"","status":"discovered","facets":{"areas":["dna-genomes"]},"source_ids":["catalog-source-dnabert-2","catalog-source-nt-v2","catalog-source-spliceai","catalog-source-pangolin"],"links":[],"attributes":{"entity_level":"task","version":null,"task":"MFASS splice-variant prioritisation","scope_note":"Functional exon-recognition assay; mfass-v2 reports a corrected baseline and one complete local DNABERT-2 protocol.","historical_missing_metadata":{"protocol_version":"not_yet_extracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"MFASS prioritisation asks whether variant scores enrich for experimentally disrupted exon recognition.","summary_source_ids":["rewire-mfass-v2-source"],"summary_source_locator":"Pinned MFASS v2 README: Correction, Dataset, Cohort reconciliation and Split; committed results files","sections":[{"title":"Choosing an evaluation","body":"The corrected rewire v2 protocol validates assay-oriented reference and mutant pairs and uses a fixed grouped split. Its baseline window is placed around the validated variant position. This task guide links that concrete protocol without replacing its identity. Preserve specialist sequence context, missing predictions and the chosen review capacity when comparing methods.","source_ids":["rewire-mfass-v2-source"],"source_locator":"Pinned MFASS v2 README: Correction, Dataset, Cohort reconciliation and Split; committed results files"},{"title":"Task scope","body":"This is a task guide, not a single versioned benchmark protocol. The connected resources provide examples or concrete procedures. A candidate method or proposed control is not evidence that an evaluation has been completed.","source_ids":["rewire-mfass-v2-source"],"source_locator":"Pinned MFASS v2 README: Correction, Dataset, Cohort reconciliation and Split; committed results files"}],"facts":[{"label":"Entity type","value":"Task guide; concrete protocol identities remain separate.","status":"source_checked","source_ids":["rewire-mfass-v2-source"],"source_locator":"Pinned MFASS v2 README: Correction, Dataset, Cohort reconciliation and Split; committed results files"},{"label":"Datasets","value":"The linked rewire MFASS v2 protocol defines its reconciled eligible variants and fixed test cohort; this guide is not another dataset release.","status":"source_checked","source_ids":["rewire-mfass-v2-source"],"source_locator":"Pinned MFASS v2 README: Correction, Dataset, Cohort reconciliation and Split; committed results files"},{"label":"Organisms","value":"Human variants evaluated through the MFASS reporter assay; this is not a population-level clinical validation.","status":"source_checked","source_ids":["rewire-mfass-v2-source"],"source_locator":"Pinned MFASS v2 README: Correction, Dataset, Cohort reconciliation and Split; committed results files"},{"label":"Assays","value":"MFASS reporter-based exon-recognition measurements, using the labels and eligibility rules preserved by the v2 protocol.","status":"source_checked","source_ids":["rewire-mfass-v2-source"],"source_locator":"Pinned MFASS v2 README: Correction, Dataset, Cohort reconciliation and Split; committed results files"},{"label":"Splits","value":"Use the grouped split-v2 manifests in the pinned runner revision. Preserve the train/test assignment and exclusions.","status":"source_checked","source_ids":["rewire-mfass-v2-source"],"source_locator":"Pinned MFASS v2 README: Correction, Dataset, Cohort reconciliation and Split; committed results files"},{"label":"Allowed inputs","value":"Variant scores aligned to the MFASS reporter-assay labels; assay-oriented sequence pairs for the corrected local baseline.","status":"source_checked","source_ids":["rewire-mfass-v2-source"],"source_locator":"Pinned MFASS v2 README: Correction, Dataset, Cohort reconciliation and Split; committed results files"},{"label":"Adaptation","value":"The corrected baseline and frozen-encoder logistic pipeline fit the training arm; the specialist scorers retain their published configurations.","status":"source_checked","source_ids":["rewire-mfass-v2-source"],"source_locator":"Pinned MFASS v2 README: Correction, Dataset, Cohort reconciliation and Split; committed results files"},{"label":"Metrics","value":"Precision at 100, average precision and AUROC in the pinned v2 evaluation. Coverage and paired uncertainty belong beside each comparison.","status":"source_checked","source_ids":["rewire-mfass-v2-source"],"source_locator":"Pinned MFASS v2 README: Correction, Dataset, Cohort reconciliation and Split; committed results files"},{"label":"Baselines","value":"The v2 k-mer/position baseline is a trained comparator. SpliceAI, Pangolin and the frozen DNABERT-2 logistic pipeline retain their own input and fitting definitions.","status":"source_checked","source_ids":["rewire-mfass-v2-source"],"source_locator":"Pinned MFASS v2 README: Correction, Dataset, Cohort reconciliation and Split; committed results files"}],"strengths":[{"text":"Functional reporter labels provide an assay endpoint separate from clinical assertions.","source_ids":["rewire-mfass-v2-source"],"source_locator":"Pinned MFASS v2 README: Correction, Dataset, Cohort reconciliation and Split; committed results files"}],"limitations":[{"text":"Reporter effects are not clinical diagnoses. Different sequence context and missing-prediction coverage prevent an identical-input interpretation.","source_ids":["rewire-mfass-v2-source"],"source_locator":"Pinned MFASS v2 README: Correction, Dataset, Cohort reconciliation and Split; committed results files"}],"diagram":{"title":"Conceptual evaluation workflow","steps":["Validate assay-oriented sequence pairs","Apply the fixed grouped split","Score each declared method","Report prioritisation, ranking and coverage"],"caption":"Conceptual task guide. Dataset preparation, parameters and scoring must come from a separately identified protocol.","source_ids":["rewire-mfass-v2-source"],"source_locator":"Pinned MFASS v2 README: Correction, Dataset, Cohort reconciliation and Split; committed results files"},"coverage":"reviewed","gaps":["A concrete evaluation still needs a protocol version, dataset release, eligible/scored counts and documented uncertainty. This guide does not manufacture those run-specific values."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Reviewed named primary documentation and existing protocol records. Editorial task scope is distinguished from source-reported procedures. No model execution or new numerical result; a reviewed guide is not a complete runnable protocol."}}}}
{"id":"catalog-task-microbial-promoters","kind":"benchmark","name":"Bacterial promoter prediction","description":"","status":"discovered","facets":{"areas":["microbes-communities"]},"source_ids":["catalog-source-prokbert","catalog-source-evo-2"],"links":[],"attributes":{"entity_level":"task","version":null,"task":"Bacterial promoter prediction","scope_note":"Classify promoter activity from microbial DNA sequence.","historical_missing_metadata":{"protocol_version":"not_yet_extracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Bacterial promoter prediction evaluates whether a sequence model can distinguish promoter-labelled sequences under a specified dataset definition.","summary_source_ids":["evidence-official-2efbfaba5a1c09f4f7fa"],"summary_source_locator":"ProkBERT README: promoter prediction task and fine-tuning examples","sections":[{"title":"Choosing an evaluation","body":"A concrete study must define what counts as a promoter, how negative examples are selected and which organisms or sequence groups are held out. Keep those decisions with the score. ProkBERT documents a promoter-prediction use case, but that implementation does not make every promoter dataset or split interchangeable.","source_ids":["evidence-official-2efbfaba5a1c09f4f7fa"],"source_locator":"ProkBERT README: promoter prediction task and fine-tuning examples"},{"title":"Task scope","body":"This is a task guide, not a single versioned benchmark protocol. The connected resources provide examples or concrete procedures. A candidate method or proposed control is not evidence that an evaluation has been completed.","source_ids":["evidence-official-2efbfaba5a1c09f4f7fa"],"source_locator":"ProkBERT README: promoter prediction task and fine-tuning examples"}],"facts":[{"label":"Entity type","value":"Task guide; concrete protocol identities remain separate.","status":"source_checked","source_ids":["evidence-official-2efbfaba5a1c09f4f7fa"],"source_locator":"ProkBERT README: promoter prediction task and fine-tuning examples"},{"label":"Datasets","value":"No single dataset is fixed by this guide. Select a linked protocol and its versioned data release.","status":"inapplicable","source_ids":["evidence-official-2efbfaba5a1c09f4f7fa"],"source_locator":"ProkBERT README: promoter prediction task and fine-tuning examples"},{"label":"Organisms","value":"No shared organism population is defined at this guide level. Record it for each selected dataset.","status":"inapplicable","source_ids":["evidence-official-2efbfaba5a1c09f4f7fa"],"source_locator":"ProkBERT README: promoter prediction task and fine-tuning examples"},{"label":"Assays","value":"No single measurement assay is fixed by this guide; the endpoint and assay belong to the selected protocol.","status":"inapplicable","source_ids":["evidence-official-2efbfaba5a1c09f4f7fa"],"source_locator":"ProkBERT README: promoter prediction task and fine-tuning examples"},{"label":"Splits","value":"No executable split is attached to this task identity. Use the selected protocol’s split manifest.","status":"inapplicable","source_ids":["evidence-official-2efbfaba5a1c09f4f7fa"],"source_locator":"ProkBERT README: promoter prediction task and fine-tuning examples"},{"label":"Allowed inputs","value":"DNA sequences and promoter labels from a named reference dataset; organism and negative-set construction are protocol-specific.","status":"source_checked","source_ids":["evidence-official-2efbfaba5a1c09f4f7fa"],"source_locator":"ProkBERT README: promoter prediction task and fine-tuning examples"},{"label":"Adaptation","value":"No common fitting regime is imposed here. Keep pretrained, frozen, probed, fine-tuned and conventional methods distinct where applicable.","status":"inapplicable","source_ids":["evidence-official-2efbfaba5a1c09f4f7fa"],"source_locator":"ProkBERT README: promoter prediction task and fine-tuning examples"},{"label":"Metrics","value":"Classification metrics defined by the selected dataset and protocol; decision thresholds and class balance must be reported.","status":"source_checked","source_ids":["evidence-official-2efbfaba5a1c09f4f7fa"],"source_locator":"ProkBERT README: promoter prediction task and fine-tuning examples"},{"label":"Baselines","value":"Documented sequence classifiers or simple composition controls may be candidates; no measured baseline value is supplied by this guide.","status":"source_checked","source_ids":["evidence-official-2efbfaba5a1c09f4f7fa"],"source_locator":"ProkBERT README: promoter prediction task and fine-tuning examples"}],"strengths":[{"text":"An explicit promoter-label task gives a sequence representation a measurable downstream endpoint.","source_ids":["evidence-official-2efbfaba5a1c09f4f7fa"],"source_locator":"ProkBERT README: promoter prediction task and fine-tuning examples"}],"limitations":[{"text":"A classifier can exploit how negatives were sampled. Transfer to another organism or promoter definition needs its own evaluation.","source_ids":["evidence-official-2efbfaba5a1c09f4f7fa"],"source_locator":"ProkBERT README: promoter prediction task and fine-tuning examples"}],"diagram":{"title":"Conceptual evaluation workflow","steps":["Choose promoter labels and negatives","Fix organism or sequence holdout","Evaluate the defined classifier","Report classification and coverage"],"caption":"Conceptual task guide. Dataset preparation, parameters and scoring must come from a separately identified protocol.","source_ids":["evidence-official-2efbfaba5a1c09f4f7fa"],"source_locator":"ProkBERT README: promoter prediction task and fine-tuning examples"},"coverage":"reviewed","gaps":["A concrete evaluation still needs a protocol version, dataset release, eligible/scored counts and documented uncertainty. This guide does not manufacture those run-specific values."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Reviewed named primary documentation and existing protocol records. Editorial task scope is distinguished from source-reported procedures. No model execution or new numerical result; a reviewed guide is not a complete runnable protocol."}}}}
{"id":"catalog-task-phage-pathogen-reads","kind":"benchmark","name":"Phage / pathogen reads","description":"","status":"discovered","facets":{"areas":["microbes-communities"]},"source_ids":["catalog-source-prokbert","catalog-source-metagene-1","catalog-source-kraken2"],"links":[],"attributes":{"entity_level":"task","version":null,"task":"Phage / pathogen reads","scope_note":"Classify held-out phage or pathogen sequences and record taxonomic distance.","historical_missing_metadata":{"protocol_version":"not_yet_extracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"This task area covers microbial sequence-read classification. Identifying a phage and identifying a pathogen are distinct endpoints.","summary_source_ids":["evidence-official-2efbfaba5a1c09f4f7fa"],"summary_source_locator":"ProkBERT README: phage identification and downstream tasks. The combined task label is catalogue scope, not a claim that phage identity establishes pathogenicity.","sections":[{"title":"Choosing an evaluation","body":"Choose a concrete label definition before evaluating read classifications. A phage-versus-non-phage benchmark does not establish clinical pathogenicity or organism abundance. Record reference-library coverage and the taxonomic or sequence separation between training and assessment. This guide describes evaluation scope and introduces no sequence-design procedure.","source_ids":["evidence-official-2efbfaba5a1c09f4f7fa"],"source_locator":"ProkBERT README: phage identification and downstream tasks. The combined task label is catalogue scope, not a claim that phage identity establishes pathogenicity."},{"title":"Task scope","body":"This is a task guide, not a single versioned benchmark protocol. The connected resources provide examples or concrete procedures. A candidate method or proposed control is not evidence that an evaluation has been completed.","source_ids":["evidence-official-2efbfaba5a1c09f4f7fa"],"source_locator":"ProkBERT README: phage identification and downstream tasks. The combined task label is catalogue scope, not a claim that phage identity establishes pathogenicity."}],"facts":[{"label":"Entity type","value":"Task guide; concrete protocol identities remain separate.","status":"source_checked","source_ids":["evidence-official-2efbfaba5a1c09f4f7fa"],"source_locator":"ProkBERT README: phage identification and downstream tasks. The combined task label is catalogue scope, not a claim that phage identity establishes pathogenicity."},{"label":"Datasets","value":"No single dataset is fixed by this guide. Select a linked protocol and its versioned data release.","status":"inapplicable","source_ids":["evidence-official-2efbfaba5a1c09f4f7fa"],"source_locator":"ProkBERT README: phage identification and downstream tasks. The combined task label is catalogue scope, not a claim that phage identity establishes pathogenicity."},{"label":"Organisms","value":"No shared organism population is defined at this guide level. Record it for each selected dataset.","status":"inapplicable","source_ids":["evidence-official-2efbfaba5a1c09f4f7fa"],"source_locator":"ProkBERT README: phage identification and downstream tasks. The combined task label is catalogue scope, not a claim that phage identity establishes pathogenicity."},{"label":"Assays","value":"No single measurement assay is fixed by this guide; the endpoint and assay belong to the selected protocol.","status":"inapplicable","source_ids":["evidence-official-2efbfaba5a1c09f4f7fa"],"source_locator":"ProkBERT README: phage identification and downstream tasks. The combined task label is catalogue scope, not a claim that phage identity establishes pathogenicity."},{"label":"Splits","value":"No executable split is attached to this task identity. Use the selected protocol’s split manifest.","status":"inapplicable","source_ids":["evidence-official-2efbfaba5a1c09f4f7fa"],"source_locator":"ProkBERT README: phage identification and downstream tasks. The combined task label is catalogue scope, not a claim that phage identity establishes pathogenicity."},{"label":"Allowed inputs","value":"Sequence reads or fragments with task-specific reference labels; reference database identity and version are part of the protocol.","status":"source_checked","source_ids":["evidence-official-2efbfaba5a1c09f4f7fa"],"source_locator":"ProkBERT README: phage identification and downstream tasks. The combined task label is catalogue scope, not a claim that phage identity establishes pathogenicity."},{"label":"Adaptation","value":"No common fitting regime is imposed here. Keep pretrained, frozen, probed, fine-tuned and conventional methods distinct where applicable.","status":"inapplicable","source_ids":["evidence-official-2efbfaba5a1c09f4f7fa"],"source_locator":"ProkBERT README: phage identification and downstream tasks. The combined task label is catalogue scope, not a claim that phage identity establishes pathogenicity."},{"label":"Metrics","value":"Label-specific classification metrics and explicit false-positive/negative counts. State the unit scored and the treatment of unclassified reads.","status":"source_checked","source_ids":["evidence-official-2efbfaba5a1c09f4f7fa"],"source_locator":"ProkBERT README: phage identification and downstream tasks. The combined task label is catalogue scope, not a claim that phage identity establishes pathogenicity."},{"label":"Baselines","value":"Compare against the conventional classifier specified by the selected protocol using the same reference information; candidate tools alone are not evidence.","status":"source_checked","source_ids":["evidence-official-2efbfaba5a1c09f4f7fa"],"source_locator":"ProkBERT README: phage identification and downstream tasks. The combined task label is catalogue scope, not a claim that phage identity establishes pathogenicity."}],"strengths":[{"text":"A defined label and held-out reference relationship make sequence-classification claims testable.","source_ids":["evidence-official-2efbfaba5a1c09f4f7fa"],"source_locator":"ProkBERT README: phage identification and downstream tasks. The combined task label is catalogue scope, not a claim that phage identity establishes pathogenicity."}],"limitations":[{"text":"Do not use a phage-detection score as evidence of pathogenicity, diagnostic accuracy or community abundance.","source_ids":["evidence-official-2efbfaba5a1c09f4f7fa"],"source_locator":"ProkBERT README: phage identification and downstream tasks. The combined task label is catalogue scope, not a claim that phage identity establishes pathogenicity."}],"diagram":{"title":"Conceptual evaluation workflow","steps":["Define the classification endpoint","Document reference and holdout scope","Classify held-out reads","Report label-specific errors"],"caption":"Conceptual task guide. Dataset preparation, parameters and scoring must come from a separately identified protocol.","source_ids":["evidence-official-2efbfaba5a1c09f4f7fa"],"source_locator":"ProkBERT README: phage identification and downstream tasks. The combined task label is catalogue scope, not a claim that phage identity establishes pathogenicity."},"coverage":"reviewed","gaps":["A concrete evaluation still needs a protocol version, dataset release, eligible/scored counts and documented uncertainty. This guide does not manufacture those run-specific values."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Reviewed named primary documentation and existing protocol records. Editorial task scope is distinguished from source-reported procedures. No model execution or new numerical result; a reviewed guide is not a complete runnable protocol."}}}}
{"id":"catalog-task-protein-design","kind":"benchmark","name":"Protein design / inverse folding","description":"","status":"discovered","facets":{"areas":["proteins-complexes"]},"source_ids":["catalog-source-proteinmpnn"],"links":[],"attributes":{"entity_level":"task","version":null,"task":"Protein design / inverse folding","scope_note":"Score or design sequences conditional on a known structure.","historical_missing_metadata":{"protocol_version":"not_yet_extracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Inverse-folding evaluation asks whether a method can propose amino-acid sequences compatible with a specified protein structure.","summary_source_ids":["src-discovery-dauparas-proteinmpnn"],"summary_source_locator":"ProteinMPNN README: overview, scoring outputs and training documentation","sections":[{"title":"Choosing an evaluation","body":"Keep inverse folding separate from sequence-only generation and experimental function. ProteinMPNN takes structural context and produces sequence probabilities or candidate sequences. A concrete benchmark must state the held-out structures, permitted constraints and assessment. Recovery of a reference sequence and experimental success answer different questions.","source_ids":["src-discovery-dauparas-proteinmpnn"],"source_locator":"ProteinMPNN README: overview, scoring outputs and training documentation"},{"title":"Task scope","body":"This is a task guide, not a single versioned benchmark protocol. The connected resources provide examples or concrete procedures. A candidate method or proposed control is not evidence that an evaluation has been completed.","source_ids":["src-discovery-dauparas-proteinmpnn"],"source_locator":"ProteinMPNN README: overview, scoring outputs and training documentation"}],"facts":[{"label":"Entity type","value":"Task guide; concrete protocol identities remain separate.","status":"source_checked","source_ids":["src-discovery-dauparas-proteinmpnn"],"source_locator":"ProteinMPNN README: overview, scoring outputs and training documentation"},{"label":"Datasets","value":"No single dataset is fixed by this guide. Select a linked protocol and its versioned data release.","status":"inapplicable","source_ids":["src-discovery-dauparas-proteinmpnn"],"source_locator":"ProteinMPNN README: overview, scoring outputs and training documentation"},{"label":"Organisms","value":"No shared organism population is defined at this guide level. Record it for each selected dataset.","status":"inapplicable","source_ids":["src-discovery-dauparas-proteinmpnn"],"source_locator":"ProteinMPNN README: overview, scoring outputs and training documentation"},{"label":"Assays","value":"No single measurement assay is fixed by this guide; the endpoint and assay belong to the selected protocol.","status":"inapplicable","source_ids":["src-discovery-dauparas-proteinmpnn"],"source_locator":"ProteinMPNN README: overview, scoring outputs and training documentation"},{"label":"Splits","value":"No executable split is attached to this task identity. Use the selected protocol’s split manifest.","status":"inapplicable","source_ids":["src-discovery-dauparas-proteinmpnn"],"source_locator":"ProteinMPNN README: overview, scoring outputs and training documentation"},{"label":"Allowed inputs","value":"A target protein backbone and declared structural or sequence constraints for an inverse-folding protocol.","status":"source_checked","source_ids":["src-discovery-dauparas-proteinmpnn"],"source_locator":"ProteinMPNN README: overview, scoring outputs and training documentation"},{"label":"Adaptation","value":"No common fitting regime is imposed here. Keep pretrained, frozen, probed, fine-tuned and conventional methods distinct where applicable.","status":"inapplicable","source_ids":["src-discovery-dauparas-proteinmpnn"],"source_locator":"ProteinMPNN README: overview, scoring outputs and training documentation"},{"label":"Metrics","value":"Protocol-specific sequence recovery or structural assessment; experimental validation, when available, is a separate endpoint.","status":"source_checked","source_ids":["src-discovery-dauparas-proteinmpnn"],"source_locator":"ProteinMPNN README: overview, scoring outputs and training documentation"},{"label":"Baselines","value":"Use inverse-folding comparators supplied with the selected benchmark and match structural inputs and constraints.","status":"source_checked","source_ids":["src-discovery-dauparas-proteinmpnn"],"source_locator":"ProteinMPNN README: overview, scoring outputs and training documentation"}],"strengths":[{"text":"A specified structural target provides a clear conditioning context for evaluating sequence proposals.","source_ids":["src-discovery-dauparas-proteinmpnn"],"source_locator":"ProteinMPNN README: overview, scoring outputs and training documentation"}],"limitations":[{"text":"Reference-sequence recovery does not by itself establish folding, function or experimental success.","source_ids":["src-discovery-dauparas-proteinmpnn"],"source_locator":"ProteinMPNN README: overview, scoring outputs and training documentation"}],"diagram":{"title":"Conceptual evaluation workflow","steps":["Select held-out structural targets","Declare allowed conditioning","Generate or score candidate sequences","Evaluate the specified endpoint"],"caption":"Conceptual task guide. Dataset preparation, parameters and scoring must come from a separately identified protocol.","source_ids":["src-discovery-dauparas-proteinmpnn"],"source_locator":"ProteinMPNN README: overview, scoring outputs and training documentation"},"coverage":"reviewed","gaps":["A concrete evaluation still needs a protocol version, dataset release, eligible/scored counts and documented uncertainty. This guide does not manufacture those run-specific values."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Reviewed named primary documentation and existing protocol records. Editorial task scope is distinguished from source-reported procedures. No model execution or new numerical result; a reviewed guide is not a complete runnable protocol."}}}}
{"id":"catalog-task-protein-monomer-structure","kind":"benchmark","name":"Monomer structure","description":"","status":"discovered","facets":{"areas":["proteins-complexes"]},"source_ids":["catalog-source-esm-2","catalog-source-esmfold","catalog-source-chai-1"],"links":[],"attributes":{"entity_level":"task","version":null,"task":"Monomer structure","scope_note":"Predict single-chain structure from sequence.","historical_missing_metadata":{"protocol_version":"not_yet_extracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Monomer-structure evaluation measures the accuracy of a predicted individual protein structure against a specified reference.","summary_source_ids":["evidence-alphafold-paper"],"summary_source_locator":"AlphaFold 3 Methods: Metrics and recent PDB comparisons. Monomer scope is defined by this catalogue guide.","sections":[{"title":"Choosing an evaluation","body":"Define whether the method receives only sequence or also evolutionary and template information. Score the individual chain with the chosen alignment and residue-coverage rules. A monomer score is not a protein-interface score, and confidence estimates are not experimental reference measurements.","source_ids":["evidence-alphafold-paper"],"source_locator":"AlphaFold 3 Methods: Metrics and recent PDB comparisons. Monomer scope is defined by this catalogue guide."},{"title":"Task scope","body":"This is a task guide, not a single versioned benchmark protocol. The connected resources provide examples or concrete procedures. A candidate method or proposed control is not evidence that an evaluation has been completed.","source_ids":["evidence-alphafold-paper"],"source_locator":"AlphaFold 3 Methods: Metrics and recent PDB comparisons. Monomer scope is defined by this catalogue guide."}],"facts":[{"label":"Entity type","value":"Task guide; concrete protocol identities remain separate.","status":"source_checked","source_ids":["evidence-alphafold-paper"],"source_locator":"AlphaFold 3 Methods: Metrics and recent PDB comparisons. Monomer scope is defined by this catalogue guide."},{"label":"Datasets","value":"No single dataset is fixed by this guide. Select a linked protocol and its versioned data release.","status":"inapplicable","source_ids":["evidence-alphafold-paper"],"source_locator":"AlphaFold 3 Methods: Metrics and recent PDB comparisons. Monomer scope is defined by this catalogue guide."},{"label":"Organisms","value":"No shared organism population is defined at this guide level. Record it for each selected dataset.","status":"inapplicable","source_ids":["evidence-alphafold-paper"],"source_locator":"AlphaFold 3 Methods: Metrics and recent PDB comparisons. Monomer scope is defined by this catalogue guide."},{"label":"Assays","value":"No single measurement assay is fixed by this guide; the endpoint and assay belong to the selected protocol.","status":"inapplicable","source_ids":["evidence-alphafold-paper"],"source_locator":"AlphaFold 3 Methods: Metrics and recent PDB comparisons. Monomer scope is defined by this catalogue guide."},{"label":"Splits","value":"No executable split is attached to this task identity. Use the selected protocol’s split manifest.","status":"inapplicable","source_ids":["evidence-alphafold-paper"],"source_locator":"AlphaFold 3 Methods: Metrics and recent PDB comparisons. Monomer scope is defined by this catalogue guide."},{"label":"Allowed inputs","value":"A protein sequence with any protocol-permitted alignments or templates; an experimental chain structure for reference-based scoring.","status":"source_checked","source_ids":["evidence-alphafold-paper"],"source_locator":"AlphaFold 3 Methods: Metrics and recent PDB comparisons. Monomer scope is defined by this catalogue guide."},{"label":"Adaptation","value":"No common fitting regime is imposed here. Keep pretrained, frozen, probed, fine-tuned and conventional methods distinct where applicable.","status":"inapplicable","source_ids":["evidence-alphafold-paper"],"source_locator":"AlphaFold 3 Methods: Metrics and recent PDB comparisons. Monomer scope is defined by this catalogue guide."},{"label":"Metrics","value":"Structure-quality measures such as LDDT or TM-score where specified by the protocol; exact residue inclusion and aggregation remain explicit.","status":"source_checked","source_ids":["evidence-alphafold-paper"],"source_locator":"AlphaFold 3 Methods: Metrics and recent PDB comparisons. Monomer scope is defined by this catalogue guide."},{"label":"Baselines","value":"Compare structure predictors under declared input information and sampling budgets rather than treating all sequence-to-structure runs as equivalent.","status":"source_checked","source_ids":["evidence-alphafold-paper"],"source_locator":"AlphaFold 3 Methods: Metrics and recent PDB comparisons. Monomer scope is defined by this catalogue guide."}],"strengths":[{"text":"Chain-level reference comparison isolates a clearly defined structural endpoint.","source_ids":["evidence-alphafold-paper"],"source_locator":"AlphaFold 3 Methods: Metrics and recent PDB comparisons. Monomer scope is defined by this catalogue guide."}],"limitations":[{"text":"A correct monomer fold does not establish the arrangement or accuracy of a molecular complex.","source_ids":["evidence-alphafold-paper"],"source_locator":"AlphaFold 3 Methods: Metrics and recent PDB comparisons. Monomer scope is defined by this catalogue guide."}],"diagram":{"title":"Conceptual evaluation workflow","steps":["Fix reference chain and allowed inputs","Predict the monomer structure","Align and match scored residues","Measure declared structural accuracy"],"caption":"Conceptual task guide. Dataset preparation, parameters and scoring must come from a separately identified protocol.","source_ids":["evidence-alphafold-paper"],"source_locator":"AlphaFold 3 Methods: Metrics and recent PDB comparisons. Monomer scope is defined by this catalogue guide."},"coverage":"reviewed","gaps":["A concrete evaluation still needs a protocol version, dataset release, eligible/scored counts and documented uncertainty. This guide does not manufacture those run-specific values."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Reviewed named primary documentation and existing protocol records. Editorial task scope is distinguished from source-reported procedures. No model execution or new numerical result; a reviewed guide is not a complete runnable protocol."}}}}
{"id":"catalog-task-proteingym-effects","kind":"benchmark","name":"ProteinGym mutation effects","description":"","status":"discovered","facets":{"areas":["proteins-complexes"]},"source_ids":["catalog-source-mimic","catalog-source-esm-2","catalog-source-proteinmpnn"],"links":[],"attributes":{"entity_level":"task","version":null,"task":"ProteinGym mutation effects","scope_note":"Rank substitution effects within held-out deep-mutational-scanning assays.","historical_missing_metadata":{"protocol_version":"not_yet_extracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"ProteinGym mutation-effect evaluation compares variant scores with measurements from individual functional assays.","summary_source_ids":["src-discovery-oatml-markslab-proteingym"],"summary_source_locator":"README: Overview, Results, benchmark baselines and contribution notes on aggregation","sections":[{"title":"Choosing an evaluation","body":"Select the suite release, assay subset and zero-shot or supervised track. ProteinGym reports within-assay metrics and further aggregation by protein and functional category. Use the published aggregation rules; a simple mean across all assays is not automatically the suite’s reported score. Preserve the distinction between molecular assay effects and clinical labels.","source_ids":["src-discovery-oatml-markslab-proteingym"],"source_locator":"README: Overview, Results, benchmark baselines and contribution notes on aggregation"},{"title":"Task scope","body":"This is a task guide, not a single versioned benchmark protocol. The connected resources provide examples or concrete procedures. A candidate method or proposed control is not evidence that an evaluation has been completed.","source_ids":["src-discovery-oatml-markslab-proteingym"],"source_locator":"README: Overview, Results, benchmark baselines and contribution notes on aggregation"}],"facts":[{"label":"Entity type","value":"Task guide; concrete protocol identities remain separate.","status":"source_checked","source_ids":["src-discovery-oatml-markslab-proteingym"],"source_locator":"README: Overview, Results, benchmark baselines and contribution notes on aggregation"},{"label":"Datasets","value":"Select a particular ProteinGym release, assay subset and evaluation track; the suite contains separate substitution and indel resources.","status":"source_checked","source_ids":["src-discovery-oatml-markslab-proteingym"],"source_locator":"README: Overview, Results, benchmark baselines and contribution notes on aggregation"},{"label":"Organisms","value":"No shared organism population is defined at this guide level. Record it for each selected dataset.","status":"inapplicable","source_ids":["src-discovery-oatml-markslab-proteingym"],"source_locator":"README: Overview, Results, benchmark baselines and contribution notes on aggregation"},{"label":"Assays","value":"Deep mutational scanning measurements for molecular effects; assay metadata and functional categories remain distinct.","status":"source_checked","source_ids":["src-discovery-oatml-markslab-proteingym"],"source_locator":"README: Overview, Results, benchmark baselines and contribution notes on aggregation"},{"label":"Splits","value":"No executable split is attached to this task identity. Use the selected protocol’s split manifest.","status":"inapplicable","source_ids":["src-discovery-oatml-markslab-proteingym"],"source_locator":"README: Overview, Results, benchmark baselines and contribution notes on aggregation"},{"label":"Allowed inputs","value":"Protein variants, permitted sequence or structure information, and the selected assay’s reference measurements.","status":"source_checked","source_ids":["src-discovery-oatml-markslab-proteingym"],"source_locator":"README: Overview, Results, benchmark baselines and contribution notes on aggregation"},{"label":"Adaptation","value":"No common fitting regime is imposed here. Keep pretrained, frozen, probed, fine-tuned and conventional methods distinct where applicable.","status":"inapplicable","source_ids":["src-discovery-oatml-markslab-proteingym"],"source_locator":"README: Overview, Results, benchmark baselines and contribution notes on aggregation"},{"label":"Metrics","value":"DMS zero-shot tracks include Spearman, NDCG, AUC, MCC and top-k recall; supervised tracks include Spearman and MSE. Choose the track-specific definition.","status":"source_checked","source_ids":["src-discovery-oatml-markslab-proteingym"],"source_locator":"README: Overview, Results, benchmark baselines and contribution notes on aggregation"},{"label":"Baselines","value":"ProteinGym supplies single-sequence, alignment-based and other comparator scores; their extra information and supervision must remain visible.","status":"source_checked","source_ids":["src-discovery-oatml-markslab-proteingym"],"source_locator":"README: Overview, Results, benchmark baselines and contribution notes on aggregation"}],"strengths":[{"text":"Per-assay reporting exposes how performance varies across proteins and measurement types.","source_ids":["src-discovery-oatml-markslab-proteingym"],"source_locator":"README: Overview, Results, benchmark baselines and contribution notes on aggregation"}],"limitations":[{"text":"Assays measure different properties and have different coverage. A pooled ranking can hide those differences and depends on the aggregation rule.","source_ids":["src-discovery-oatml-markslab-proteingym"],"source_locator":"README: Overview, Results, benchmark baselines and contribution notes on aggregation"}],"diagram":{"title":"Conceptual evaluation workflow","steps":["Pin release, track and assays","Score the eligible variants","Compute per-assay metrics","Apply the declared aggregation"],"caption":"Conceptual task guide. Dataset preparation, parameters and scoring must come from a separately identified protocol.","source_ids":["src-discovery-oatml-markslab-proteingym"],"source_locator":"README: Overview, Results, benchmark baselines and contribution notes on aggregation"},"coverage":"reviewed","gaps":["A concrete evaluation still needs a protocol version, dataset release, eligible/scored counts and documented uncertainty. This guide does not manufacture those run-specific values."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Reviewed named primary documentation and existing protocol records. Editorial task scope is distinguished from source-reported procedures. No model execution or new numerical result; a reviewed guide is not a complete runnable protocol."}}}}
{"id":"catalog-task-rna-secondary-structure","kind":"benchmark","name":"RNA secondary structure","description":"","status":"discovered","facets":{"areas":["rna-transcriptomes"]},"source_ids":["catalog-source-rna-fm","catalog-source-rhofold"],"links":[],"attributes":{"entity_level":"task","version":null,"task":"RNA secondary structure","scope_note":"Compare predicted base pairs against held-out RNA structures.","historical_missing_metadata":{"protocol_version":"not_yet_extracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"RNA secondary-structure prediction evaluates proposed nucleotide pairing rather than complete three-dimensional geometry.","summary_source_ids":["src-discovery-terry-r123-rnabenchmark","src-discovery-ml4bio-rna-fm"],"summary_source_locator":"BEACON README: Tasks and Datasets; RNA-FM README: secondary structure prediction and foundation-model outputs","sections":[{"title":"Choosing an evaluation","body":"Use reference pairing labels and a declared scoring rule. BEACON includes a dedicated secondary-structure task, while RNA-FM supplies representations and pairing-related outputs used by downstream models. A representation, a pairing predictor and a complete benchmark are different records. Keep sequence-family overlap and reference-label construction visible.","source_ids":["src-discovery-terry-r123-rnabenchmark","src-discovery-ml4bio-rna-fm"],"source_locator":"BEACON README: Tasks and Datasets; RNA-FM README: secondary structure prediction and foundation-model outputs"},{"title":"Task scope","body":"This is a task guide, not a single versioned benchmark protocol. The connected resources provide examples or concrete procedures. A candidate method or proposed control is not evidence that an evaluation has been completed.","source_ids":["src-discovery-terry-r123-rnabenchmark","src-discovery-ml4bio-rna-fm"],"source_locator":"BEACON README: Tasks and Datasets; RNA-FM README: secondary structure prediction and foundation-model outputs"}],"facts":[{"label":"Entity type","value":"Task guide; concrete protocol identities remain separate.","status":"source_checked","source_ids":["src-discovery-terry-r123-rnabenchmark","src-discovery-ml4bio-rna-fm"],"source_locator":"BEACON README: Tasks and Datasets; RNA-FM README: secondary structure prediction and foundation-model outputs"},{"label":"Datasets","value":"No single dataset is fixed by this guide. Select a linked protocol and its versioned data release.","status":"inapplicable","source_ids":["src-discovery-terry-r123-rnabenchmark","src-discovery-ml4bio-rna-fm"],"source_locator":"BEACON README: Tasks and Datasets; RNA-FM README: secondary structure prediction and foundation-model outputs"},{"label":"Organisms","value":"No shared organism population is defined at this guide level. Record it for each selected dataset.","status":"inapplicable","source_ids":["src-discovery-terry-r123-rnabenchmark","src-discovery-ml4bio-rna-fm"],"source_locator":"BEACON README: Tasks and Datasets; RNA-FM README: secondary structure prediction and foundation-model outputs"},{"label":"Assays","value":"No single measurement assay is fixed by this guide; the endpoint and assay belong to the selected protocol.","status":"inapplicable","source_ids":["src-discovery-terry-r123-rnabenchmark","src-discovery-ml4bio-rna-fm"],"source_locator":"BEACON README: Tasks and Datasets; RNA-FM README: secondary structure prediction and foundation-model outputs"},{"label":"Splits","value":"No executable split is attached to this task identity. Use the selected protocol’s split manifest.","status":"inapplicable","source_ids":["src-discovery-terry-r123-rnabenchmark","src-discovery-ml4bio-rna-fm"],"source_locator":"BEACON README: Tasks and Datasets; RNA-FM README: secondary structure prediction and foundation-model outputs"},{"label":"Allowed inputs","value":"RNA sequence and any permitted auxiliary information; reference secondary-structure labels for scoring.","status":"source_checked","source_ids":["src-discovery-terry-r123-rnabenchmark","src-discovery-ml4bio-rna-fm"],"source_locator":"BEACON README: Tasks and Datasets; RNA-FM README: secondary structure prediction and foundation-model outputs"},{"label":"Adaptation","value":"No common fitting regime is imposed here. Keep pretrained, frozen, probed, fine-tuned and conventional methods distinct where applicable.","status":"inapplicable","source_ids":["src-discovery-terry-r123-rnabenchmark","src-discovery-ml4bio-rna-fm"],"source_locator":"BEACON README: Tasks and Datasets; RNA-FM README: secondary structure prediction and foundation-model outputs"},{"label":"Metrics","value":"Pairing or secondary-structure metrics from the selected protocol. This guide defines no common pseudoknot policy, matching tolerance or aggregation.","status":"source_checked","source_ids":["src-discovery-terry-r123-rnabenchmark","src-discovery-ml4bio-rna-fm"],"source_locator":"BEACON README: Tasks and Datasets; RNA-FM README: secondary structure prediction and foundation-model outputs"},{"label":"Baselines","value":"Compare the learned predictor with the conventional folding or learned baselines of the concrete task; do not infer a measured result from model availability.","status":"source_checked","source_ids":["src-discovery-terry-r123-rnabenchmark","src-discovery-ml4bio-rna-fm"],"source_locator":"BEACON README: Tasks and Datasets; RNA-FM README: secondary structure prediction and foundation-model outputs"}],"strengths":[{"text":"Pairing labels give a distinct structural endpoint that can be evaluated separately from full 3D prediction.","source_ids":["src-discovery-terry-r123-rnabenchmark","src-discovery-ml4bio-rna-fm"],"source_locator":"BEACON README: Tasks and Datasets; RNA-FM README: secondary structure prediction and foundation-model outputs"}],"limitations":[{"text":"Secondary-structure accuracy does not establish tertiary-structure accuracy, and close sequence relatives can weaken a claimed generalisation test.","source_ids":["src-discovery-terry-r123-rnabenchmark","src-discovery-ml4bio-rna-fm"],"source_locator":"BEACON README: Tasks and Datasets; RNA-FM README: secondary structure prediction and foundation-model outputs"}],"diagram":{"title":"Conceptual evaluation workflow","steps":["Select RNA references and split","Declare permitted sequence context","Predict pairing structure","Apply the declared matching and scoring"],"caption":"Conceptual task guide. Dataset preparation, parameters and scoring must come from a separately identified protocol.","source_ids":["src-discovery-terry-r123-rnabenchmark","src-discovery-ml4bio-rna-fm"],"source_locator":"BEACON README: Tasks and Datasets; RNA-FM README: secondary structure prediction and foundation-model outputs"},"coverage":"reviewed","gaps":["A concrete evaluation still needs a protocol version, dataset release, eligible/scored counts and documented uncertainty. This guide does not manufacture those run-specific values."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Reviewed named primary documentation and existing protocol records. Editorial task scope is distinguished from source-reported procedures. No model execution or new numerical result; a reviewed guide is not a complete runnable protocol."}}}}
{"id":"catalog-task-rna-splice-sites","kind":"benchmark","name":"RNA splice-site mapping","description":"","status":"discovered","facets":{"areas":["rna-transcriptomes"]},"source_ids":["catalog-source-rna-fm","catalog-source-mimic"],"links":[],"attributes":{"entity_level":"task","version":null,"task":"RNA splice-site mapping","scope_note":"Predict splice-site classes from transcript sequence, using a held-out gene split.","historical_missing_metadata":{"protocol_version":"not_yet_extracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Splice-site mapping evaluates where splice-related labels occur along a sequence. It is distinct from scoring the effect of a particular variant.","summary_source_ids":["src-discovery-terry-r123-rnabenchmark"],"summary_source_locator":"README: Tasks and Datasets (SpliceAI task); model inventory including SpliceBERT variants","sections":[{"title":"Choosing an evaluation","body":"Choose the task’s coordinate system, sequence window and label definition before interpreting scores. BEACON lists a SpliceAI-labelled downstream task and several splice-oriented representations. Their presence in the suite identifies a task area, not a shared checkpoint or proof of performance for every listed model.","source_ids":["src-discovery-terry-r123-rnabenchmark"],"source_locator":"README: Tasks and Datasets (SpliceAI task); model inventory including SpliceBERT variants"},{"title":"Task scope","body":"This is a task guide, not a single versioned benchmark protocol. The connected resources provide examples or concrete procedures. A candidate method or proposed control is not evidence that an evaluation has been completed.","source_ids":["src-discovery-terry-r123-rnabenchmark"],"source_locator":"README: Tasks and Datasets (SpliceAI task); model inventory including SpliceBERT variants"}],"facts":[{"label":"Entity type","value":"Task guide; concrete protocol identities remain separate.","status":"source_checked","source_ids":["src-discovery-terry-r123-rnabenchmark"],"source_locator":"README: Tasks and Datasets (SpliceAI task); model inventory including SpliceBERT variants"},{"label":"Datasets","value":"No single dataset is fixed by this guide. Select a linked protocol and its versioned data release.","status":"inapplicable","source_ids":["src-discovery-terry-r123-rnabenchmark"],"source_locator":"README: Tasks and Datasets (SpliceAI task); model inventory including SpliceBERT variants"},{"label":"Organisms","value":"No shared organism population is defined at this guide level. Record it for each selected dataset.","status":"inapplicable","source_ids":["src-discovery-terry-r123-rnabenchmark"],"source_locator":"README: Tasks and Datasets (SpliceAI task); model inventory including SpliceBERT variants"},{"label":"Assays","value":"No single measurement assay is fixed by this guide; the endpoint and assay belong to the selected protocol.","status":"inapplicable","source_ids":["src-discovery-terry-r123-rnabenchmark"],"source_locator":"README: Tasks and Datasets (SpliceAI task); model inventory including SpliceBERT variants"},{"label":"Splits","value":"No executable split is attached to this task identity. Use the selected protocol’s split manifest.","status":"inapplicable","source_ids":["src-discovery-terry-r123-rnabenchmark"],"source_locator":"README: Tasks and Datasets (SpliceAI task); model inventory including SpliceBERT variants"},{"label":"Allowed inputs","value":"Sequence windows and position-level splice labels under the selected task’s coordinate and strand conventions.","status":"source_checked","source_ids":["src-discovery-terry-r123-rnabenchmark"],"source_locator":"README: Tasks and Datasets (SpliceAI task); model inventory including SpliceBERT variants"},{"label":"Adaptation","value":"No common fitting regime is imposed here. Keep pretrained, frozen, probed, fine-tuned and conventional methods distinct where applicable.","status":"inapplicable","source_ids":["src-discovery-terry-r123-rnabenchmark"],"source_locator":"README: Tasks and Datasets (SpliceAI task); model inventory including SpliceBERT variants"},{"label":"Metrics","value":"The chosen protocol must define site-level scoring, matching tolerance and class balance; this guide does not supply one universal metric.","status":"source_checked","source_ids":["src-discovery-terry-r123-rnabenchmark"],"source_locator":"README: Tasks and Datasets (SpliceAI task); model inventory including SpliceBERT variants"},{"label":"Baselines","value":"Use the protocol’s sequence and splice-specialist comparators with matched inputs. Proposed applicability is separate from completed evaluation.","status":"source_checked","source_ids":["src-discovery-terry-r123-rnabenchmark"],"source_locator":"README: Tasks and Datasets (SpliceAI task); model inventory including SpliceBERT variants"}],"strengths":[{"text":"Position-level labels test localisation, rather than only a sequence-wide classification.","source_ids":["src-discovery-terry-r123-rnabenchmark"],"source_locator":"README: Tasks and Datasets (SpliceAI task); model inventory including SpliceBERT variants"}],"limitations":[{"text":"A splice-site score is not a measured variant-effect score and does not establish clinical interpretation.","source_ids":["src-discovery-terry-r123-rnabenchmark"],"source_locator":"README: Tasks and Datasets (SpliceAI task); model inventory including SpliceBERT variants"}],"diagram":{"title":"Conceptual evaluation workflow","steps":["Pin site labels and coordinate system","Define sequence windows and split","Predict positional splice outputs","Score sites using the declared rule"],"caption":"Conceptual task guide. Dataset preparation, parameters and scoring must come from a separately identified protocol.","source_ids":["src-discovery-terry-r123-rnabenchmark"],"source_locator":"README: Tasks and Datasets (SpliceAI task); model inventory including SpliceBERT variants"},"coverage":"reviewed","gaps":["A concrete evaluation still needs a protocol version, dataset release, eligible/scored counts and documented uncertainty. This guide does not manufacture those run-specific values."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Reviewed named primary documentation and existing protocol records. Editorial task scope is distinguished from source-reported procedures. No model execution or new numerical result; a reviewed guide is not a complete runnable protocol."}}}}
{"id":"catalog-task-rna-tertiary-structure","kind":"benchmark","name":"RNA tertiary structure","description":"","status":"discovered","facets":{"areas":["rna-transcriptomes"]},"source_ids":["catalog-source-rna-fm","catalog-source-rhofold"],"links":[],"attributes":{"entity_level":"task","version":null,"task":"RNA tertiary structure","scope_note":"Compare predicted 3D folds against independently held-out structures.","historical_missing_metadata":{"protocol_version":"not_yet_extracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"RNA tertiary-structure evaluation compares predicted three-dimensional RNA geometry with reference structures.","summary_source_ids":["src-discovery-ml4bio-rna-fm","evidence-alphafold-paper"],"summary_source_locator":"RNA-FM README: RhoFold; AlphaFold 3 Methods: Nucleic acid prediction baseline and CASP15 RNA targets","sections":[{"title":"Choosing an evaluation","body":"Declare the target molecules and available inputs, including whether an alignment is allowed. Compare coordinates against the relevant reference state using an explicit alignment and atom-selection rule. The AlphaFold 3 paper describes a CASP15 RNA comparison; the RNA-FM project documents RhoFold as a separate complete predictor. A base RNA embedding model is not itself that evaluated pipeline.","source_ids":["src-discovery-ml4bio-rna-fm","evidence-alphafold-paper"],"source_locator":"RNA-FM README: RhoFold; AlphaFold 3 Methods: Nucleic acid prediction baseline and CASP15 RNA targets"},{"title":"Task scope","body":"This is a task guide, not a single versioned benchmark protocol. The connected resources provide examples or concrete procedures. A candidate method or proposed control is not evidence that an evaluation has been completed.","source_ids":["src-discovery-ml4bio-rna-fm","evidence-alphafold-paper"],"source_locator":"RNA-FM README: RhoFold; AlphaFold 3 Methods: Nucleic acid prediction baseline and CASP15 RNA targets"}],"facts":[{"label":"Entity type","value":"Task guide; concrete protocol identities remain separate.","status":"source_checked","source_ids":["src-discovery-ml4bio-rna-fm","evidence-alphafold-paper"],"source_locator":"RNA-FM README: RhoFold; AlphaFold 3 Methods: Nucleic acid prediction baseline and CASP15 RNA targets"},{"label":"Datasets","value":"No single dataset is fixed by this guide. Select a linked protocol and its versioned data release.","status":"inapplicable","source_ids":["src-discovery-ml4bio-rna-fm","evidence-alphafold-paper"],"source_locator":"RNA-FM README: RhoFold; AlphaFold 3 Methods: Nucleic acid prediction baseline and CASP15 RNA targets"},{"label":"Organisms","value":"No shared organism population is defined at this guide level. Record it for each selected dataset.","status":"inapplicable","source_ids":["src-discovery-ml4bio-rna-fm","evidence-alphafold-paper"],"source_locator":"RNA-FM README: RhoFold; AlphaFold 3 Methods: Nucleic acid prediction baseline and CASP15 RNA targets"},{"label":"Assays","value":"No single measurement assay is fixed by this guide; the endpoint and assay belong to the selected protocol.","status":"inapplicable","source_ids":["src-discovery-ml4bio-rna-fm","evidence-alphafold-paper"],"source_locator":"RNA-FM README: RhoFold; AlphaFold 3 Methods: Nucleic acid prediction baseline and CASP15 RNA targets"},{"label":"Splits","value":"No executable split is attached to this task identity. Use the selected protocol’s split manifest.","status":"inapplicable","source_ids":["src-discovery-ml4bio-rna-fm","evidence-alphafold-paper"],"source_locator":"RNA-FM README: RhoFold; AlphaFold 3 Methods: Nucleic acid prediction baseline and CASP15 RNA targets"},{"label":"Allowed inputs","value":"RNA sequence and permitted auxiliary information, plus a reference structure for geometric evaluation.","status":"source_checked","source_ids":["src-discovery-ml4bio-rna-fm","evidence-alphafold-paper"],"source_locator":"RNA-FM README: RhoFold; AlphaFold 3 Methods: Nucleic acid prediction baseline and CASP15 RNA targets"},{"label":"Adaptation","value":"No common fitting regime is imposed here. Keep pretrained, frozen, probed, fine-tuned and conventional methods distinct where applicable.","status":"inapplicable","source_ids":["src-discovery-ml4bio-rna-fm","evidence-alphafold-paper"],"source_locator":"RNA-FM README: RhoFold; AlphaFold 3 Methods: Nucleic acid prediction baseline and CASP15 RNA targets"},{"label":"Metrics","value":"Protocol-defined structural accuracy. Atom selection, reference states and treatment of missing residues must accompany a number.","status":"source_checked","source_ids":["src-discovery-ml4bio-rna-fm","evidence-alphafold-paper"],"source_locator":"RNA-FM README: RhoFold; AlphaFold 3 Methods: Nucleic acid prediction baseline and CASP15 RNA targets"},{"label":"Baselines","value":"Use complete RNA-structure predictors and their stated input information; a language-model family name alone does not identify a structural pipeline.","status":"source_checked","source_ids":["src-discovery-ml4bio-rna-fm","evidence-alphafold-paper"],"source_locator":"RNA-FM README: RhoFold; AlphaFold 3 Methods: Nucleic acid prediction baseline and CASP15 RNA targets"}],"strengths":[{"text":"Coordinate-level scoring tests a structural endpoint beyond nucleotide pairing or embedding quality.","source_ids":["src-discovery-ml4bio-rna-fm","evidence-alphafold-paper"],"source_locator":"RNA-FM README: RhoFold; AlphaFold 3 Methods: Nucleic acid prediction baseline and CASP15 RNA targets"}],"limitations":[{"text":"Alternative conformations and incomplete experimental structures complicate reference comparison. Predictions are not a calibrated dynamic ensemble.","source_ids":["src-discovery-ml4bio-rna-fm","evidence-alphafold-paper"],"source_locator":"RNA-FM README: RhoFold; AlphaFold 3 Methods: Nucleic acid prediction baseline and CASP15 RNA targets"}],"diagram":{"title":"Conceptual evaluation workflow","steps":["Select targets and reference states","Declare inputs and full prediction pipeline","Predict RNA coordinates","Align and score the specified atoms"],"caption":"Conceptual task guide. Dataset preparation, parameters and scoring must come from a separately identified protocol.","source_ids":["src-discovery-ml4bio-rna-fm","evidence-alphafold-paper"],"source_locator":"RNA-FM README: RhoFold; AlphaFold 3 Methods: Nucleic acid prediction baseline and CASP15 RNA targets"},"coverage":"reviewed","gaps":["A concrete evaluation still needs a protocol version, dataset release, eligible/scored counts and documented uncertainty. This guide does not manufacture those run-specific values."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Reviewed named primary documentation and existing protocol records. Editorial task scope is distinguished from source-reported procedures. No model execution or new numerical result; a reviewed guide is not a complete runnable protocol."}}}}
{"id":"catalog-task-utr-translation","kind":"benchmark","name":"Translation / RNA stability","description":"","status":"discovered","facets":{"areas":["rna-transcriptomes"]},"source_ids":["catalog-source-mrna-fm","catalog-source-mimic"],"links":[],"attributes":{"entity_level":"task","version":null,"task":"Translation / RNA stability","scope_note":"Predict measured translation or stability effects; choose UTR or coding-sequence assays to match each model’s input modality.","historical_missing_metadata":{"protocol_version":"not_yet_extracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Translation and RNA-stability tasks test different measurable properties of transcripts and should be reported as separate endpoints.","summary_source_ids":["src-discovery-morrislab-mrnabench"],"summary_source_locator":"README: Dataset Catalog / Translation Regulation and RNA Stability; Model Catalog / Baseline Models","sections":[{"title":"Choosing an evaluation","body":"Select the endpoint and biological context before comparing methods. mRNABench distinguishes mean ribosome load, translation efficiency and RNA half-life datasets. Reporter constructs, native transcripts and different organisms are not interchangeable datasets. Record which transcript regions and supervision are available to the model.","source_ids":["src-discovery-morrislab-mrnabench"],"source_locator":"README: Dataset Catalog / Translation Regulation and RNA Stability; Model Catalog / Baseline Models"},{"title":"Task scope","body":"This is a task guide, not a single versioned benchmark protocol. The connected resources provide examples or concrete procedures. A candidate method or proposed control is not evidence that an evaluation has been completed.","source_ids":["src-discovery-morrislab-mrnabench"],"source_locator":"README: Dataset Catalog / Translation Regulation and RNA Stability; Model Catalog / Baseline Models"}],"facts":[{"label":"Entity type","value":"Task guide; concrete protocol identities remain separate.","status":"source_checked","source_ids":["src-discovery-morrislab-mrnabench"],"source_locator":"README: Dataset Catalog / Translation Regulation and RNA Stability; Model Catalog / Baseline Models"},{"label":"Datasets","value":"No single dataset is fixed by this guide. Select a linked protocol and its versioned data release.","status":"inapplicable","source_ids":["src-discovery-morrislab-mrnabench"],"source_locator":"README: Dataset Catalog / Translation Regulation and RNA Stability; Model Catalog / Baseline Models"},{"label":"Organisms","value":"No shared organism population is defined at this guide level. Record it for each selected dataset.","status":"inapplicable","source_ids":["src-discovery-morrislab-mrnabench"],"source_locator":"README: Dataset Catalog / Translation Regulation and RNA Stability; Model Catalog / Baseline Models"},{"label":"Assays","value":"No single measurement assay is fixed by this guide; the endpoint and assay belong to the selected protocol.","status":"inapplicable","source_ids":["src-discovery-morrislab-mrnabench"],"source_locator":"README: Dataset Catalog / Translation Regulation and RNA Stability; Model Catalog / Baseline Models"},{"label":"Splits","value":"No executable split is attached to this task identity. Use the selected protocol’s split manifest.","status":"inapplicable","source_ids":["src-discovery-morrislab-mrnabench"],"source_locator":"README: Dataset Catalog / Translation Regulation and RNA Stability; Model Catalog / Baseline Models"},{"label":"Allowed inputs","value":"Transcript or UTR sequences with task-specific experimental measurements and declared transcript-region boundaries.","status":"source_checked","source_ids":["src-discovery-morrislab-mrnabench"],"source_locator":"README: Dataset Catalog / Translation Regulation and RNA Stability; Model Catalog / Baseline Models"},{"label":"Adaptation","value":"No common fitting regime is imposed here. Keep pretrained, frozen, probed, fine-tuned and conventional methods distinct where applicable.","status":"inapplicable","source_ids":["src-discovery-morrislab-mrnabench"],"source_locator":"README: Dataset Catalog / Translation Regulation and RNA Stability; Model Catalog / Baseline Models"},{"label":"Metrics","value":"Endpoint-specific regression or classification as defined by the selected dataset; retain measurement units and any target transformation.","status":"source_checked","source_ids":["src-discovery-morrislab-mrnabench"],"source_locator":"README: Dataset Catalog / Translation Regulation and RNA Stability; Model Catalog / Baseline Models"},{"label":"Baselines","value":"mRNABench supplies conventional baselines alongside RNA and DNA representation models. Compare only within the same endpoint, split and input scope.","status":"source_checked","source_ids":["src-discovery-morrislab-mrnabench"],"source_locator":"README: Dataset Catalog / Translation Regulation and RNA Stability; Model Catalog / Baseline Models"}],"strengths":[{"text":"Separate molecular endpoints allow evaluation of specific transcript properties instead of an undefined general RNA score.","source_ids":["src-discovery-morrislab-mrnabench"],"source_locator":"README: Dataset Catalog / Translation Regulation and RNA Stability; Model Catalog / Baseline Models"}],"limitations":[{"text":"Ribosome loading, translation efficiency and stability are related but distinct measurements. Good performance on one is not evidence for the others.","source_ids":["src-discovery-morrislab-mrnabench"],"source_locator":"README: Dataset Catalog / Translation Regulation and RNA Stability; Model Catalog / Baseline Models"}],"diagram":{"title":"Conceptual evaluation workflow","steps":["Choose molecular endpoint and dataset","Fix sequence regions and held-out split","Fit or apply the declared predictor","Score the matching measurement"],"caption":"Conceptual task guide. Dataset preparation, parameters and scoring must come from a separately identified protocol.","source_ids":["src-discovery-morrislab-mrnabench"],"source_locator":"README: Dataset Catalog / Translation Regulation and RNA Stability; Model Catalog / Baseline Models"},"coverage":"reviewed","gaps":["A concrete evaluation still needs a protocol version, dataset release, eligible/scored counts and documented uncertainty. This guide does not manufacture those run-specific values."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Reviewed named primary documentation and existing protocol records. Editorial task scope is distinguished from source-reported procedures. No model execution or new numerical result; a reviewed guide is not a complete runnable protocol."}}}}
{"id":"cathe2-2025","kind":"source","name":"CATHe2: Enhanced CATH superfamily detection using ProstT5 and structural alphabets","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["proteins-complexes"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12631783/","version":"journal full text in PMC","retrieved_at":"2026-09-15T23:33:26Z","doi":"10.1093/biomethods/bpaf080","publication_status":"peer_reviewed","year":2025,"artifact_sha256":"713dbfb6ec1cc1aa85c0543eb93aafa0b45b8873df28053b765dd0a1b6d9b563","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12631783/fullTextXML","artifact_retrieved_at":"2026-09-16T10:33:38.366Z","legacy_paper":{"id":"cathe2-2025","title":"CATHe2: Enhanced CATH superfamily detection using ProstT5 and structural alphabets","year":2025,"publication_status":"peer_reviewed","version":"journal full text in PMC","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12631783/","primary_domain":"proteins-complexes","retrieved_utc":"2026-09-15T23:33:26Z","doi":"10.1093/biomethods/bpaf080","notes":"Numeric result checked against Table 3. in primary full-text XML; journal/source: Biology Methods & Protocols."},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"cell-dino-2025","kind":"source","name":"Cell-DINO: Self-supervised image-based embeddings for cell fluorescent microscopy","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["cells-tissues"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12826486/","version":"version of record","retrieved_at":"2026-09-15T23:37:05Z","doi":"10.1371/journal.pcbi.1013828","publication_status":"peer_reviewed","year":2025,"artifact_sha256":"12a53a78c70b3033c3351cf7afd4da42ebc98bb3281308f07e71e5baffc153a0","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12826486/fullTextXML","artifact_retrieved_at":"2026-09-16T10:41:06Z","legacy_paper":{"id":"cell-dino-2025","title":"Cell-DINO: Self-supervised image-based embeddings for cell fluorescent microscopy","year":2025,"publication_status":"peer_reviewed","version":"version of record","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12826486/","primary_domain":"cells-tissues","retrieved_utc":"2026-09-15T23:37:05Z","notes":"Primary full text verified via Europe PMC fullTextXML; venue: PLOS Computational Biology; PMC ID: PMC12826486. PL column is F1-score reported on a 0–100 scale; Cell-DINO is a vision encoder plus downstream classifier.","doi":"10.1371/journal.pcbi.1013828"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"cell2sentence-2024","kind":"source","name":"Cell2Sentence: Teaching Large Language Models the Language of Biology","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["cells-tissues"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11565894/","version":"preprint archived 2024-10-29","retrieved_at":"2026-09-15T23:25:00Z","doi":"10.1101/2023.09.11.557287","publication_status":"preprint","year":2024,"artifact_sha256":"e088727d6e04857fccb7033a9b074e1850f775e86e7d2e99e603dde09558ab02","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11565894/fullTextXML","artifact_retrieved_at":"2026-09-16T10:41:16.533640+00:00","legacy_paper":{"id":"cell2sentence-2024","title":"Cell2Sentence: Teaching Large Language Models the Language of Biology","year":2024,"publication_status":"preprint","version":"preprint archived 2024-10-29","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11565894/","primary_domain":"cells-tissues","retrieved_utc":"2026-09-15T23:25:00Z","notes":"Primary full text via Europe PMC XML; venue: bioRxiv; PMC ID: PMC11565894.","doi":"10.1101/2023.09.11.557287"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"claim-b2-2ome-lm-2025","kind":"claim","name":"Reported AUC for 2OMe-LM","description":"","status":"source_checked","facets":{"areas":["rna-transcriptomes"],"tasks":["human RNA 2-prime-O-methylation site prediction"]},"source_ids":["2ome-lm-2025"],"links":[{"relation":"subject","target_id":"b2-2ome-lm-2025"}],"attributes":{"field":"attributes.printed_value","value":"0.919","source_locator":"Table 1, 2OMe-LM row, AUC column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:38.332Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness."}}}
{"id":"claim-b2-antibody-deamidation-plm-2024","kind":"claim","name":"Reported accuracy for ESM-2 650M embeddings + classifier","description":"","status":"source_checked","facets":{"areas":["proteins-complexes"],"tasks":["antibody deamidation-site prediction"]},"source_ids":["antibody-deamidation-plm-2024"],"links":[{"relation":"subject","target_id":"b2-antibody-deamidation-plm-2024"}],"attributes":{"field":"attributes.printed_value","value":"0.944","source_locator":"Table 1, Global embeddings only row, Accuracy column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:38.478Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness."}}}
{"id":"claim-b2-barcodebert-2026","kind":"claim","name":"Reported accuracy for BarcodeBERT (4–4-4)","description":"","status":"source_checked","facets":{"areas":["dna-genomes"],"tasks":["unseen-species genus classification"]},"source_ids":["barcodebert-2026"],"links":[{"relation":"subject","target_id":"b2-barcodebert-2026"}],"attributes":{"field":"attributes.printed_value","value":"78.5","source_locator":"Table 1, BarcodeBERT (4–4-4) row, unseen-species genus-level 1-NN Acc (%) column","review":{"method":"independent_ai_table_review","reviewer":"Codex secondary table review","reviewed_at":"2026-09-16T10:38:57.558051+00:00","notes":"Resolved the two-level column header: Acc (%) falls under genus-level 1-NN probe of unseen species, not seen-species classification or BIN reconstruction. BarcodeBERT (4–4-4) has 78.5 in this cell."}}}
{"id":"claim-b2-birna-bert-2025","kind":"claim","name":"Reported F1 for BiRNA-BERT","description":"","status":"source_checked","facets":{"areas":["rna-transcriptomes"],"tasks":["extremely long RNA species classification"]},"source_ids":["birna-bert-2025"],"links":[{"relation":"subject","target_id":"b2-birna-bert-2025"}],"attributes":{"field":"attributes.printed_value","value":"0.804","source_locator":"Table 2, BiRNA-BERT row, F1 Score column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:38.292Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness."}}}
{"id":"claim-b2-cathe2-2025","kind":"claim","name":"Reported F1 for CATHe2 + ProstT5","description":"","status":"source_checked","facets":{"areas":["proteins-complexes"],"tasks":["CATH superfamily annotation"]},"source_ids":["cathe2-2025"],"links":[{"relation":"subject","target_id":"b2-cathe2-2025"}],"attributes":{"field":"attributes.printed_value","value":"82.3","source_locator":"Table 3, ProstT5 full row, F1 score column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:38.366Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness."}}}
{"id":"claim-b2-clathrin-plm-2025","kind":"claim","name":"Reported accuracy for ESM-2 embedding + paper classifier","description":"","status":"source_checked","facets":{"areas":["proteins-complexes"],"tasks":["clathrin protein classification"]},"source_ids":["clathrin-plm-2025"],"links":[{"relation":"subject","target_id":"b2-clathrin-plm-2025"}],"attributes":{"field":"attributes.printed_value","value":"0.916","source_locator":"Table 2, Independent test / ESM-2 row, ACC column","review":{"method":"independent_ai_table_review","reviewer":"Codex secondary table review","reviewed_at":"2026-09-16T10:38:57.558194+00:00","notes":"Resolved the blank evaluation-strategy cells by their independent-test row group. ESM-2 ACC is 0.916 there; the cross-validation ESM-2 ACC is instead 0.873. This is the paper classifier using embeddings, not a standalone checkpoint."}}}
{"id":"claim-b2-cobra-rna-binding-2026","kind":"claim","name":"Reported MCC for ERNIE-RNA + CoBRA","description":"","status":"source_checked","facets":{"areas":["rna-transcriptomes"],"tasks":["RNA compound-binding site prediction"]},"source_ids":["cobra-rna-binding-2026"],"links":[{"relation":"subject","target_id":"b2-cobra-rna-binding-2026"}],"attributes":{"field":"attributes.printed_value","value":"0.657","source_locator":"Table 2, ERNIE-RNA / TCL focal row, MCC column","review":{"method":"independent_ai_table_review","reviewer":"Codex secondary table review","reviewed_at":"2026-09-16T10:38:57.558197+00:00","notes":"Matched ERNIE-RNA jointly with TCL focal loss, then the MCC column. Table 2 explicitly reports test-set models. The cell is 0.657, distinct from AUROC 0.868."}}}
{"id":"claim-b2-codonbert-vaccines-2024","kind":"claim","name":"Reported Spearman rho for CodonBERT","description":"","status":"source_checked","facets":{"areas":["rna-transcriptomes"],"tasks":["flu-vaccine mRNA property prediction"]},"source_ids":["codonbert-vaccines-2024"],"links":[{"relation":"subject","target_id":"b2-codonbert-vaccines-2024"}],"attributes":{"field":"attributes.printed_value","value":"0.81","source_locator":"Table 2, CodonBERT row, Flu vaccines Spearman correlation column","review":{"method":"independent_ai_table_review","reviewer":"Codex secondary table review","reviewed_at":"2026-09-16T10:38:57.558201+00:00","notes":"Matched the CodonBERT row and Flu vaccines column (0.81). The table footnote identifies regression columns as Spearman rank correlation and singles out E. coli as classification; this is not a flu-vaccine accuracy score."}}}
{"id":"claim-b2-dart-eval-regulatory-2024","kind":"claim","name":"Reported accuracy for DNABERT-2","description":"","status":"source_checked","facets":{"areas":["dna-genomes"],"tasks":["regulatory element identification"]},"source_ids":["dart-eval-regulatory-2024"],"links":[{"relation":"subject","target_id":"b2-dart-eval-regulatory-2024"}],"attributes":{"field":"attributes.printed_value","value":"0.876","source_locator":"Table 3 (PDF page 5), DNABERT-2 row, Zero-Shot Accuracy column","review":{"method":"independent_ai_table_review","reviewer":"Codex secondary table review","reviewed_at":"2026-09-16T10:38:57.558203+00:00","notes":"Inspected the pinned NeurIPS primary PDF table and explanatory text. The DNABERT-2 row reports 0.876 under Zero-Shot Accuracy. The caption defines this as pairwise prioritization of positives over matched controls, distinct from supervised absolute accuracy."}}}
{"id":"claim-b2-dnabert2-enhancer-2025","kind":"claim","name":"Reported AUC for DNABERT2-Enhancer","description":"","status":"source_checked","facets":{"areas":["dna-genomes"],"tasks":["enhancer recognition"]},"source_ids":["dnabert2-enhancer-2025"],"links":[{"relation":"subject","target_id":"b2-dnabert2-enhancer-2025"}],"attributes":{"field":"attributes.printed_value","value":"0.965","source_locator":"Table 4, first-layer DNABERT2-Enhancer row, AUC column","review":{"method":"independent_ai_table_review","reviewer":"Codex secondary table review","reviewed_at":"2026-09-16T10:38:57.558204+00:00","notes":"Resolved the first-layer row group. DNABERT2-Enhancer AUC is 0.965, whereas second-layer AUC is 0.933. The caption explicitly describes 5-fold cross-validation on Liu training data, not an independent held-out test."}}}
{"id":"claim-b2-eden-genomic-classification-2026","kind":"claim","name":"Reported MCC for DNABERT-2","description":"","status":"source_checked","facets":{"areas":["dna-genomes"],"tasks":["human core-promoter classification"]},"source_ids":["eden-genomic-classification-2026"],"links":[{"relation":"subject","target_id":"b2-eden-genomic-classification-2026"}],"attributes":{"field":"attributes.printed_value","value":"70.52","source_locator":"Table 5, DNABERT-2 row, H-CPD (MCC) column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:37.531Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness."}}}
{"id":"claim-b2-ernie-rna-2025","kind":"claim","name":"Reported binary F1 for ERNIE-RNA","description":"","status":"source_checked","facets":{"areas":["rna-transcriptomes"],"tasks":["RNA secondary-structure prediction"]},"source_ids":["ernie-rna-2025"],"links":[{"relation":"subject","target_id":"b2-ernie-rna-2025"}],"attributes":{"field":"attributes.printed_value","value":"0.575","source_locator":"Table 2, ERNIE-RNA zero-shot row, bpRNA-new F1-Score (binary) column","review":{"method":"independent_ai_table_review","reviewer":"Codex secondary table review","reviewed_at":"2026-09-16T10:38:57.558206+00:00","notes":"Resolved bpRNA-new as the first three-column dataset group and F1-Score (binary) as its third metric. ERNIE-RNA zero shot is 86M and reports 0.575; RNA3DB-2D F1 is instead 0.542."}}}
{"id":"claim-b2-esm2-ofs-fitness-2025","kind":"claim","name":"Reported Spearman rho for ESM2 OFS pseudo-perplexity","description":"","status":"source_checked","facets":{"areas":["proteins-complexes"],"tasks":["protein variant fitness prediction"]},"source_ids":["esm2-ofs-fitness-2025"],"links":[{"relation":"subject","target_id":"b2-esm2-ofs-fitness-2025"}],"attributes":{"field":"attributes.printed_value","value":"0.403","source_locator":"Published PDF page 6 (033014-6), Table I, ESM2: OFS PP row, Aggregate mean column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source review, not human review","reviewed_at":"2026-09-16T10:45:41.099916+00:00","notes":"Publisher PDF retrieved through official APS harvest endpoint after direct download returned403. Table I is ProteinGym substitutions, not indels TableII. Last column aggregate mean0.403; separate function categories precede it. This verifies reported score, not experimental reproduction. Comparator rows in this table are sourced from ProteinGym; OFS PP is authors own method."}}}
{"id":"claim-b2-fusion-breakpoint-foundation-models-2026","kind":"claim","name":"Reported ROC AUC for Nucleotide Transformer + NN (middle)","description":"","status":"source_checked","facets":{"areas":["dna-genomes"],"tasks":["gene fusion breakpoint classification"]},"source_ids":["fusion-breakpoint-foundation-models-2026"],"links":[{"relation":"subject","target_id":"b2-fusion-breakpoint-foundation-models-2026"}],"attributes":{"field":"attributes.printed_value","value":"0.994","source_locator":"Table 2, NT / NN (middle) row, ROC AUC column","review":{"method":"independent_ai_table_review","reviewer":"Codex secondary table review","reviewed_at":"2026-09-16T10:38:57.558209+00:00","notes":"Matched NT jointly with NN (middle) and ROC AUC 0.994 in the full-test-set table. NT with SVM reports 0.995 and is a separate pipeline."}}}
{"id":"claim-b2-genomic-tokenizer-selection-2025","kind":"claim","name":"Reported MCC for Caduceus (character tokens)","description":"","status":"source_checked","facets":{"areas":["dna-genomes"],"tasks":["regulatory sequence classification"]},"source_ids":["genomic-tokenizer-selection-2025"],"links":[{"relation":"subject","target_id":"b2-genomic-tokenizer-selection-2025"}],"attributes":{"field":"attributes.printed_value","value":"0.778","source_locator":"Table 2, Regulatory row, Caduceus (char) MCC column","review":{"method":"independent_ai_table_review","reviewer":"Codex secondary table review","reviewed_at":"2026-09-16T10:38:57.558210+00:00","notes":"Matched Regulatory row with Caduceus (char) column, 0.778. Caption establishes these as MCC summaries by category; model-size row identifies 3.9M parameters. This is an aggregated category result, not a single unspecified split."}}}
{"id":"claim-b2-gsmformer-ppi-2026","kind":"claim","name":"Reported AUROC for GSMFormer-PPI + ProstT5","description":"","status":"source_checked","facets":{"areas":["proteins-complexes"],"tasks":["protein-protein interaction prediction"]},"source_ids":["gsmformer-ppi-2026"],"links":[{"relation":"subject","target_id":"b2-gsmformer-ppi-2026"}],"attributes":{"field":"attributes.printed_value","value":"0.988","source_locator":"Table 6, ProstT5 embedding row, AUROC column","review":{"method":"independent_ai_table_review","reviewer":"Codex secondary table review","reviewed_at":"2026-09-16T10:38:57.558212+00:00","notes":"Matched ProstT5 embedding row and AUROC column, 0.988. Caption explicitly describes GSMFormer-PPI using embeddings as node features, not standalone ProstT5 prediction."}}}
{"id":"claim-b2-megsite-2025","kind":"claim","name":"Reported AUC for MegSite + ESM3","description":"","status":"source_checked","facets":{"areas":["proteins-complexes"],"tasks":["DNA-binding residue prediction"]},"source_ids":["megsite-2025"],"links":[{"relation":"subject","target_id":"b2-megsite-2025"}],"attributes":{"field":"attributes.printed_value","value":"0.948","source_locator":"Table 2, DNA-129_Test / ESM3 row, AUC column","review":{"method":"independent_ai_table_review","reviewer":"Codex secondary table review","reviewed_at":"2026-09-16T10:38:57.558213+00:00","notes":"Resolved DNA-129_Test row group and ESM3 row. AUC is 0.948; the next numeric cell 0.582 is AP. Caption states an embedding comparison within MegSite."}}}
{"id":"claim-b2-mrna-lm-2025","kind":"claim","name":"Reported Spearman rho for mRNA-LM","description":"","status":"source_checked","facets":{"areas":["rna-transcriptomes"],"tasks":["mRNA half-life prediction"]},"source_ids":["mrna-lm-2025"],"links":[{"relation":"subject","target_id":"b2-mrna-lm-2025"}],"attributes":{"field":"attributes.printed_value","value":"0.696","source_locator":"Table 1, mRNA-LM row, mRNA half-life Spearman column","review":{"method":"independent_ai_table_review","reviewer":"Codex secondary table review","reviewed_at":"2026-09-16T10:38:57.558214+00:00","notes":"Resolved mRNA half-life column under the Spearman header spanning three tasks. mRNA-LM gives 0.696. Caption identifies average test performance across cross-validation splits; protein-expression AUROC is a different column."}}}
{"id":"claim-b2-mrnabert-2025","kind":"claim","name":"Reported R-squared for mRNABERT","description":"","status":"source_checked","facets":{"areas":["rna-transcriptomes"],"tasks":["translation-efficiency prediction"]},"source_ids":["mrnabert-2025"],"links":[{"relation":"subject","target_id":"b2-mrnabert-2025"}],"attributes":{"field":"attributes.printed_value","value":"0.669","source_locator":"Table 2, mRNABERT (3066) row, Human R-squared column","review":{"method":"independent_ai_table_review","reviewer":"Codex secondary table review","reviewed_at":"2026-09-16T10:38:57.558216+00:00","notes":"Resolved Human group and its R-squared subcolumn. mRNABERT (3066) reports 0.669; Human Spearman is 0.814 and Mouse R-squared is 0.649. Caption specifies ultra-long mRNA translation-efficiency prediction."}}}
{"id":"claim-b2-mulan-2025","kind":"claim","name":"Reported AUC for MULAN-ESM2 S","description":"","status":"source_checked","facets":{"areas":["proteins-complexes"],"tasks":["human protein-protein interaction prediction"]},"source_ids":["mulan-2025"],"links":[{"relation":"subject","target_id":"b2-mulan-2025"}],"attributes":{"field":"attributes.printed_value","value":"0.717","source_locator":"Table 2, MULAN-ESM2 S row, HumanPPI AUC column","review":{"method":"independent_ai_table_review","reviewer":"Codex secondary table review","reviewed_at":"2026-09-16T10:38:57.558217+00:00","notes":"Resolved the multirow header: HumanPPI uses AUC. MULAN-ESM2 S has 0.717; this is the small-model group, distinct from M and L variants."}}}
{"id":"claim-b2-phylogpn-2025","kind":"claim","name":"Reported AUROC for PhyloGPN","description":"","status":"source_checked","facets":{"areas":["dna-genomes"],"tasks":["ClinVar 3-prime UTR variant classification"]},"source_ids":["phylogpn-2025"],"links":[{"relation":"subject","target_id":"b2-phylogpn-2025"}],"attributes":{"field":"attributes.printed_value","value":"0.94","source_locator":"Table 1, 3-prime UTR row, PhyloGPN AUROC column","review":{"method":"independent_ai_table_review","reviewer":"Codex secondary table review","reviewed_at":"2026-09-16T10:38:57.558218+00:00","notes":"Matched 3-prime UTR row and PhyloGPN column (0.94). Caption specifies log-likelihood-ratio predictions of ClinVar classes and explicitly defines each cell as AUROC."}}}
{"id":"claim-b2-polya-glm-2025","kind":"claim","name":"Reported AUC for HyenaDNA","description":"","status":"source_checked","facets":{"areas":["dna-genomes"],"tasks":["polyadenylation site detection"]},"source_ids":["polya-glm-2025"],"links":[{"relation":"subject","target_id":"b2-polya-glm-2025"}],"attributes":{"field":"attributes.printed_value","value":"0.7510","source_locator":"Table 1, Few-shot HyenaDNA row, G-G AUC column","review":{"method":"independent_ai_table_review","reviewer":"Codex secondary table review","reviewed_at":"2026-09-16T10:38:57.558220+00:00","notes":"Resolved Few-shot group, HyenaDNA row, and G-G subcolumn under AUC (0.7510). IG-G AUC is 0.7541. Caption states averages over five-fold cross-validation and distinguishes negative sampling regions."}}}
{"id":"claim-b2-rlsite-rna-binding-2025","kind":"claim","name":"Reported AUC for RLsite","description":"","status":"source_checked","facets":{"areas":["rna-transcriptomes"],"tasks":["RNA-small-molecule binding-site prediction"]},"source_ids":["rlsite-rna-binding-2025"],"links":[{"relation":"subject","target_id":"b2-rlsite-rna-binding-2025"}],"attributes":{"field":"attributes.printed_value","value":"0.828","source_locator":"Table 1, RLsite row, T18 AUC column","review":{"method":"independent_ai_table_review","reviewer":"Codex secondary table review","reviewed_at":"2026-09-16T10:38:57.558222+00:00","notes":"Matched RLsite and AUC (0.828). Caption explicitly identifies dataset T18; MCC 0.474 is a different metric."}}}
{"id":"claim-b2-rnaret-2026","kind":"claim","name":"Reported F1 for RNAret","description":"","status":"source_checked","facets":{"areas":["rna-transcriptomes"],"tasks":["miRNA-mRNA interaction prediction"]},"source_ids":["rnaret-2026"],"links":[{"relation":"subject","target_id":"b2-rnaret-2026"}],"attributes":{"field":"attributes.printed_value","value":"0.9622","source_locator":"Table 1, MirTarRAW / 5-mer RNAret row, F1 column","review":{"method":"independent_ai_table_review","reviewer":"Codex secondary table review","reviewed_at":"2026-09-16T10:38:57.558224+00:00","notes":"Resolved the MirTarRAW section, 5-mer RNAret row, and F1 column (0.9622), distinct from DeepMirTarLeft F1 0.9728. Methods confirm 72/8/20 train/validation/test partition for MirTarRAW."}}}
{"id":"claim-b2-spin-protein-function-2026","kind":"claim","name":"Reported F1 macro-weighted for SPIN + ESM2-35M","description":"","status":"source_checked","facets":{"areas":["proteins-complexes"],"tasks":["protein function annotation"]},"source_ids":["spin-protein-function-2026"],"links":[{"relation":"subject","target_id":"b2-spin-protein-function-2026"}],"attributes":{"field":"attributes.printed_value","value":"0.796","source_locator":"Table 1, ESM2-35M Test row, F1_m-w column","review":{"method":"independent_ai_table_review","reviewer":"Codex secondary table review","reviewed_at":"2026-09-16T10:38:57.558225+00:00","notes":"Resolved Test group and macro-weighted F1 subcolumn (0.796) for frozen ESM2-35M in SPIN. Test weighted accuracy is 0.798. Methods define inverse-frequency class weighting for macro-weighted F1."}}}
{"id":"claim-b2-structure-informed-plm-2025","kind":"claim","name":"Reported AUROC for structure-informed pLM","description":"","status":"source_checked","facets":{"areas":["proteins-complexes"],"tasks":["protein variant-effect classification"]},"source_ids":["structure-informed-plm-2025"],"links":[{"relation":"subject","target_id":"b2-structure-informed-plm-2025"}],"attributes":{"field":"attributes.printed_value","value":".803","source_locator":"PMC12068927 HTML, Table4, AA+SS+RSA+CM row, AUROC column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source review, not human review","reviewed_at":"2026-09-16T10:45:41.099916+00:00","notes":"Full-text HTML succeeds although EuropePMC XMLreturned404. Row is mutation-site variables AA+SS+RSA+CM, not neighbouring environment variant. AUROC .803 is numerically equivalent to preserved legacy0.803. Source check, not experimental reproduction; do not claim original source printed leading zero."}}}
{"id":"claim-lit-001","kind":"claim","name":"Reported AUC for Caduceus-Ph","description":"","status":"source_checked","facets":{"areas":["dna-genomes"],"tasks":["Human 5mC detection"]},"source_ids":["dna-foundation-models-2025"],"links":[{"relation":"subject","target_id":"lit-001"}],"attributes":{"field":"attributes.printed_value","value":"0.783","source_locator":"Table 3, Human 5mC row, Caduceus-Ph column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:35.327Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness."}}}
{"id":"claim-lit-002","kind":"claim","name":"Reported AUC for NT-v2","description":"","status":"source_checked","facets":{"areas":["dna-genomes"],"tasks":["Human 5mC detection"]},"source_ids":["dna-foundation-models-2025"],"links":[{"relation":"subject","target_id":"lit-002"}],"attributes":{"field":"attributes.printed_value","value":"0.7377","source_locator":"Table 3, Human 5mC row, NT-v2 column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:35.327Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness."}}}
{"id":"claim-lit-003","kind":"claim","name":"Reported Accuracy for ENBED","description":"","status":"source_checked","facets":{"areas":["dna-genomes"],"tasks":["Enhancer classification"]},"source_ids":["enbed-2024"],"links":[{"relation":"subject","target_id":"lit-003"}],"attributes":{"field":"attributes.printed_value","value":"90.3","source_locator":"Table 2, Mouse Enhancers row, ENBED column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:35.378Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness."}}}
{"id":"claim-lit-004","kind":"claim","name":"Reported Accuracy for ENBED (GRCh38)","description":"","status":"source_checked","facets":{"areas":["dna-genomes"],"tasks":["Enhancer classification"]},"source_ids":["enbed-2024"],"links":[{"relation":"subject","target_id":"lit-004"}],"attributes":{"field":"attributes.printed_value","value":"81.1","source_locator":"Table 2, Mouse Enhancers row, ENBED (GRCh38) column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:35.378Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness."}}}
{"id":"claim-lit-005","kind":"claim","name":"Reported Accuracy for DNABERT-2","description":"","status":"source_checked","facets":{"areas":["dna-genomes"],"tasks":["G-quadruplex classification"]},"source_ids":["quadruplex-llm-benchmark-2025"],"links":[{"relation":"subject","target_id":"lit-005"}],"attributes":{"field":"attributes.printed_value","value":"97.0","source_locator":"Table 5, DNABERT-2 (117 M) row, Accuracy column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:35.379Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness."}}}
{"id":"claim-lit-006","kind":"claim","name":"Reported Accuracy for Caduceus","description":"","status":"source_checked","facets":{"areas":["dna-genomes"],"tasks":["G-quadruplex classification"]},"source_ids":["quadruplex-llm-benchmark-2025"],"links":[{"relation":"subject","target_id":"lit-006"}],"attributes":{"field":"attributes.printed_value","value":"95.0","source_locator":"Table 5, Caduceus (8 M) row, Accuracy column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:35.379Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness."}}}
{"id":"claim-lit-007","kind":"claim","name":"Reported AUROC for HyenaDNA","description":"","status":"source_checked","facets":{"areas":["dna-genomes"],"tasks":["Enhancer-target gene prediction"]},"source_ids":["dnalongbench-2025"],"links":[{"relation":"subject","target_id":"lit-007"}],"attributes":{"field":"attributes.printed_value","value":"0.828","source_locator":"Table 3, HyenaDNA row, ETGP column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.492545+00:00","notes":"ETGP is the first numeric column, separate from the six CMP cell-type columns and average; caption defines ETGP AUROC. This verifies the central score at its source location, not every metadata field or an experimental reproduction."}}}
{"id":"claim-lit-008","kind":"claim","name":"Reported AUROC for Caduceus-Ph","description":"","status":"source_checked","facets":{"areas":["dna-genomes"],"tasks":["Enhancer-target gene prediction"]},"source_ids":["dnalongbench-2025"],"links":[{"relation":"subject","target_id":"lit-008"}],"attributes":{"field":"attributes.printed_value","value":"0.826","source_locator":"Table 3, Caduceus-Ph row, ETGP column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.493765+00:00","notes":"ETGP is the first numeric column, separate from the six CMP cell-type columns and average; caption defines ETGP AUROC. This verifies the central score at its source location, not every metadata field or an experimental reproduction."}}}
{"id":"claim-lit-009","kind":"claim","name":"Reported Pearson R for RiNALMo","description":"","status":"source_checked","facets":{"areas":["rna-transcriptomes"],"tasks":["Mean ribosome load from MPRA"]},"source_ids":["mrnabench-2025"],"links":[{"relation":"subject","target_id":"lit-009"}],"attributes":{"field":"attributes.printed_value","value":"0.74","source_locator":"Table 2, RiNALMo row, MRL MPRA column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.497221+00:00","notes":"MRL MPRA is the second task under Local and uses R. Caption specifies mean over ten random seeds and selected best model per family, not a fully identified checkpoint. This verifies the central score at its source location, not every metadata field or an experimental reproduction."}}}
{"id":"claim-lit-010","kind":"claim","name":"Reported Pearson R for RNA-FM","description":"","status":"source_checked","facets":{"areas":["rna-transcriptomes"],"tasks":["Mean ribosome load from MPRA"]},"source_ids":["mrnabench-2025"],"links":[{"relation":"subject","target_id":"lit-010"}],"attributes":{"field":"attributes.printed_value","value":"0.49","source_locator":"Table 2, RNA-FM row, MRL MPRA column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.500211+00:00","notes":"MRL MPRA is the second task under Local and uses R. Caption specifies mean over ten random seeds and selected best model per family, not a fully identified checkpoint. This verifies the central score at its source location, not every metadata field or an experimental reproduction."}}}
{"id":"claim-lit-011","kind":"claim","name":"Reported F1 for BPfold","description":"","status":"source_checked","facets":{"areas":["rna-transcriptomes"],"tasks":["RNA secondary structure"]},"source_ids":["bpfold-2025"],"links":[{"relation":"subject","target_id":"lit-011"}],"attributes":{"field":"attributes.printed_value","value":"0.814","source_locator":"Table 2, BPfold row, PDB F1 column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.502000+00:00","notes":"PDB is the second four-metric block; its F1 is numeric column six, not Rfam F1. This verifies the central score at its source location, not every metadata field or an experimental reproduction."}}}
{"id":"claim-lit-012","kind":"claim","name":"Reported F1 for RNAfold","description":"","status":"source_checked","facets":{"areas":["rna-transcriptomes"],"tasks":["RNA secondary structure"]},"source_ids":["bpfold-2025"],"links":[{"relation":"subject","target_id":"lit-012"}],"attributes":{"field":"attributes.printed_value","value":"0.747","source_locator":"Table 2, RNAfold row, PDB F1 column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.504220+00:00","notes":"PDB is the second four-metric block; its F1 is numeric column six, not Rfam F1. This verifies the central score at its source location, not every metadata field or an experimental reproduction."}}}
{"id":"claim-lit-013","kind":"claim","name":"Reported F1 for TU-Fold (aug)","description":"","status":"source_checked","facets":{"areas":["rna-transcriptomes"],"tasks":["RNA secondary structure"]},"source_ids":["tu-fold-2025"],"links":[{"relation":"subject","target_id":"lit-013"}],"attributes":{"field":"attributes.printed_value","value":"0.947","source_locator":"Table 2, TU-Fold (aug) row, Overall F1 column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.505799+00:00","notes":"Overall is the first two-metric block. F1 is first numeric column; source cell includes uncertainty after the preserved central value. This verifies the central score at its source location, not every metadata field or an experimental reproduction."}}}
{"id":"claim-lit-014","kind":"claim","name":"Reported F1 for UFold","description":"","status":"source_checked","facets":{"areas":["rna-transcriptomes"],"tasks":["RNA secondary structure"]},"source_ids":["tu-fold-2025"],"links":[{"relation":"subject","target_id":"lit-014"}],"attributes":{"field":"attributes.printed_value","value":"0.938","source_locator":"Table 2, UFold row, Overall F1 column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.507031+00:00","notes":"Overall is the first two-metric block. F1 is first numeric column; source cell includes uncertainty after the preserved central value. This verifies the central score at its source location, not every metadata field or an experimental reproduction."}}}
{"id":"claim-lit-015","kind":"claim","name":"Reported Median F1 for DEBFold","description":"","status":"source_checked","facets":{"areas":["rna-transcriptomes"],"tasks":["RNA secondary structure"]},"source_ids":["debfold-2024"],"links":[{"relation":"subject","target_id":"lit-015"}],"attributes":{"field":"attributes.printed_value","value":"55.7","source_locator":"Table 1, DEBFold row, TestSetβ F1 (%) column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.509290+00:00","notes":"TestSet beta is the second four-column block. F1 (%) is its first column; caption reports test-set median F1. This verifies the central score at its source location, not every metadata field or an experimental reproduction."}}}
{"id":"claim-lit-016","kind":"claim","name":"Reported Median F1 for RNAfold","description":"","status":"source_checked","facets":{"areas":["rna-transcriptomes"],"tasks":["RNA secondary structure"]},"source_ids":["debfold-2024"],"links":[{"relation":"subject","target_id":"lit-016"}],"attributes":{"field":"attributes.printed_value","value":"52.3","source_locator":"Table 1, RNAfold row, TestSetβ F1 (%) column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.511166+00:00","notes":"TestSet beta is the second four-column block. F1 (%) is its first column; caption reports test-set median F1. This verifies the central score at its source location, not every metadata field or an experimental reproduction."}}}
{"id":"claim-lit-017","kind":"claim","name":"Reported Mean Spearman rho for ESM-2","description":"","status":"source_checked","facets":{"areas":["proteins-complexes"],"tasks":["Zero-shot substitution mutation effects: stability"]},"source_ids":["proteingym-2023"],"links":[{"relation":"subject","target_id":"lit-017"}],"attributes":{"field":"attributes.printed_value","value":"0.488","source_locator":"Table A7, ESM-2 (15B) row, Stability column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.517323+00:00","notes":"Table A7 is zero-shot substitution DMS grouped by function. Stability is the fifth numeric column; model-type row spans do not change its placement. This verifies the central score at its source location, not every metadata field or an experimental reproduction."}}}
{"id":"claim-lit-018","kind":"claim","name":"Reported Mean Spearman rho for ProteinMPNN","description":"","status":"source_checked","facets":{"areas":["proteins-complexes"],"tasks":["Zero-shot substitution mutation effects: stability"]},"source_ids":["proteingym-2023"],"links":[{"relation":"subject","target_id":"lit-018"}],"attributes":{"field":"attributes.printed_value","value":"0.566","source_locator":"Table A7, ProteinMPNN row, Stability column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.523422+00:00","notes":"Table A7 is zero-shot substitution DMS grouped by function. Stability is the fifth numeric column; model-type row spans do not change its placement. This verifies the central score at its source location, not every metadata field or an experimental reproduction."}}}
{"id":"claim-lit-019","kind":"claim","name":"Reported AUROC for FUJISAN","description":"","status":"source_checked","facets":{"areas":["proteins-complexes"],"tasks":["Enzyme functional identity prediction"]},"source_ids":["fujisan-2024"],"links":[{"relation":"subject","target_id":"lit-019"}],"attributes":{"field":"attributes.printed_value","value":"0.9427","source_locator":"Table 1, FUJISAN row, AUROC column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:35.728Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness."}}}
{"id":"claim-lit-020","kind":"claim","name":"Reported AUROC for ESM2","description":"","status":"source_checked","facets":{"areas":["proteins-complexes"],"tasks":["Enzyme functional identity prediction"]},"source_ids":["fujisan-2024"],"links":[{"relation":"subject","target_id":"lit-020"}],"attributes":{"field":"attributes.printed_value","value":"0.7991","source_locator":"Table 1, ESM2 row, AUROC column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:35.728Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness."}}}
{"id":"claim-lit-021","kind":"claim","name":"Reported R² for ESM-2","description":"","status":"source_checked","facets":{"areas":["proteins-complexes"],"tasks":["Mutated RBD binding prediction"]},"source_ids":["prime-2026"],"links":[{"relation":"subject","target_id":"lit-021"}],"attributes":{"field":"attributes.printed_value","value":"0.0248","source_locator":"Table 1, ESM-2 8M / Mean / not fine-tuned row, Position-Stratified Split Binding R² column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.525183+00:00","notes":"Resolved model row spans and Mean/CLS subrows in JATS: selected Mean, not fine-tuned (cross), Position-Stratified Split > Binding > R-squared. Central value agrees; uncertainty is retained in evidence. This verifies the central score at its source location, not every metadata field or an experimental reproduction."}}}
{"id":"claim-lit-022","kind":"claim","name":"Reported R² for ESM-C","description":"","status":"source_checked","facets":{"areas":["proteins-complexes"],"tasks":["Mutated RBD binding prediction"]},"source_ids":["prime-2026"],"links":[{"relation":"subject","target_id":"lit-022"}],"attributes":{"field":"attributes.printed_value","value":"-0.0162","source_locator":"Table 1, ESM-C 300M / Mean / not fine-tuned row, Position-Stratified Split Binding R² column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.526387+00:00","notes":"Resolved model row spans and Mean/CLS subrows in JATS: selected Mean, not fine-tuned (cross), Position-Stratified Split > Binding > R-squared. Central value agrees; uncertainty is retained in evidence. This verifies the central score at its source location, not every metadata field or an experimental reproduction."}}}
{"id":"claim-lit-023","kind":"claim","name":"Reported Mean |Spearman rho| for PST","description":"","status":"source_checked","facets":{"areas":["proteins-complexes"],"tasks":["Zero-shot variant effect prediction"]},"source_ids":["pst-2025"],"links":[{"relation":"subject","target_id":"lit-023"}],"attributes":{"field":"attributes.printed_value","value":"0.501","source_locator":"Table 2, PST row, Zero-shot VEP Mean |ρ| column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.527633+00:00","notes":"Zero-shot VEP is the last metric group; selected Mean absolute rho, not GO/EC/binding-site metrics. This verifies the central score at its source location, not every metadata field or an experimental reproduction."}}}
{"id":"claim-lit-024","kind":"claim","name":"Reported Mean |Spearman rho| for ESM-2","description":"","status":"source_checked","facets":{"areas":["proteins-complexes"],"tasks":["Zero-shot variant effect prediction"]},"source_ids":["pst-2025"],"links":[{"relation":"subject","target_id":"lit-024"}],"attributes":{"field":"attributes.printed_value","value":"0.489","source_locator":"Table 2, ESM-2 row, Zero-shot VEP Mean |ρ| column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.528656+00:00","notes":"Zero-shot VEP is the last metric group; selected Mean absolute rho, not GO/EC/binding-site metrics. This verifies the central score at its source location, not every metadata field or an experimental reproduction."}}}
{"id":"claim-lit-025","kind":"claim","name":"Reported F1-Score for scGPT","description":"","status":"source_checked","facets":{"areas":["cells-tissues"],"tasks":["Cell-type identification"]},"source_ids":["single-cell-peft-2024"],"links":[{"relation":"subject","target_id":"lit-025"}],"attributes":{"field":"attributes.printed_value","value":"0.734","source_locator":"Table 2, M.S. / scGPT row, F1-Score column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.530269+00:00","notes":"Selected M.S. dataset block, first scGPT/Geneformer occurrences. F1-Score is last column; later dataset blocks deliberately excluded. This verifies the central score at its source location, not every metadata field or an experimental reproduction."}}}
{"id":"claim-lit-026","kind":"claim","name":"Reported F1-Score for Geneformer","description":"","status":"source_checked","facets":{"areas":["cells-tissues"],"tasks":["Cell-type identification"]},"source_ids":["single-cell-peft-2024"],"links":[{"relation":"subject","target_id":"lit-026"}],"attributes":{"field":"attributes.printed_value","value":"0.388","source_locator":"Table 2, M.S. / Geneformer row, F1-Score column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.531756+00:00","notes":"Selected M.S. dataset block, first scGPT/Geneformer occurrences. F1-Score is last column; later dataset blocks deliberately excluded. This verifies the central score at its source location, not every metadata field or an experimental reproduction."}}}
{"id":"claim-lit-027","kind":"claim","name":"Reported Partial-label accuracy for C2S (GPT-2 Large)","description":"","status":"source_checked","facets":{"areas":["cells-tissues"],"tasks":["Combinatorial cell-label classification"]},"source_ids":["cell2sentence-2024"],"links":[{"relation":"subject","target_id":"lit-027"}],"attributes":{"field":"attributes.printed_value","value":"0.631","source_locator":"Table 3, Partial label / C2S (GPT-2 Large) row, L1000 Acc column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.533640+00:00","notes":"Read inline small-caps/bold XML in document order, restoring Geneformer and GPT-2 Large labels. Selected Partial label (first block), L1000 > Acc, not AUROC or Full label. This verifies the central score at its source location, not every metadata field or an experimental reproduction."}}}
{"id":"claim-lit-028","kind":"claim","name":"Reported Partial-label accuracy for Geneformer","description":"","status":"source_checked","facets":{"areas":["cells-tissues"],"tasks":["Combinatorial cell-label classification"]},"source_ids":["cell2sentence-2024"],"links":[{"relation":"subject","target_id":"lit-028"}],"attributes":{"field":"attributes.printed_value","value":"0.419","source_locator":"Table 3, Partial label / Geneformer row, L1000 Acc column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.535220+00:00","notes":"Read inline small-caps/bold XML in document order, restoring Geneformer and GPT-2 Large labels. Selected Partial label (first block), L1000 > Acc, not AUROC or Full label. This verifies the central score at its source location, not every metadata field or an experimental reproduction."}}}
{"id":"claim-lit-029","kind":"claim","name":"Reported F1 for scGPT","description":"","status":"source_checked","facets":{"areas":["cells-tissues"],"tasks":["Cell-type annotation"]},"source_ids":["scelmo-2025"],"links":[{"relation":"subject","target_id":"lit-029"}],"attributes":{"field":"attributes.printed_value","value":"0.550","source_locator":"Table 1, hPancreas zero-shot / scGPT (z) row, F1 column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.537541+00:00","notes":"Selected first hPancreas zero-shot block and F1 last column. Caption says some scores are copied from GenePT; this is source checking of the reported table, not independent experimental evidence. This verifies the central score at its source location, not every metadata field or an experimental reproduction."}}}
{"id":"claim-lit-030","kind":"claim","name":"Reported F1 for Geneformer","description":"","status":"source_checked","facets":{"areas":["cells-tissues"],"tasks":["Cell-type annotation"]},"source_ids":["scelmo-2025"],"links":[{"relation":"subject","target_id":"lit-030"}],"attributes":{"field":"attributes.printed_value","value":"0.270","source_locator":"Table 1, hPancreas zero-shot / Geneformer (z) row, F1 column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.539630+00:00","notes":"Selected first hPancreas zero-shot block and F1 last column. Caption says some scores are copied from GenePT; this is source checking of the reported table, not independent experimental evidence. This verifies the central score at its source location, not every metadata field or an experimental reproduction."}}}
{"id":"claim-lit-031","kind":"claim","name":"Reported AUROC for scRegNet (Geneformer backbone)","description":"","status":"source_checked","facets":{"areas":["cells-tissues"],"tasks":["Gene-regulatory link prediction"]},"source_ids":["scregnet-2025"],"links":[{"relation":"subject","target_id":"lit-031"}],"attributes":{"field":"attributes.printed_value","value":"0.89","source_locator":"Table 2, scRegNet (w/ Geneformer) row, hESC AUROC entry","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.541287+00:00","notes":"Read break elements: cells contain AUROC on first line then AUPRC. Selected hESC (first cell type), first line. Caption specifies 500 most-variable genes and 50 independent evaluations. This verifies the central score at its source location, not every metadata field or an experimental reproduction."}}}
{"id":"claim-lit-032","kind":"claim","name":"Reported AUROC for scRegNet (scBERT backbone)","description":"","status":"source_checked","facets":{"areas":["cells-tissues"],"tasks":["Gene-regulatory link prediction"]},"source_ids":["scregnet-2025"],"links":[{"relation":"subject","target_id":"lit-032"}],"attributes":{"field":"attributes.printed_value","value":"0.88","source_locator":"Table 2, scRegNet (w/ scBERT) row, hESC AUROC entry","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.542807+00:00","notes":"Read break elements: cells contain AUROC on first line then AUPRC. Selected hESC (first cell type), first line. Caption specifies 500 most-variable genes and 50 independent evaluations. This verifies the central score at its source location, not every metadata field or an experimental reproduction."}}}
{"id":"claim-lit-033","kind":"claim","name":"Reported Accuracy for ProkBERT-mini","description":"","status":"source_checked","facets":{"areas":["microbes-communities"],"tasks":["E. coli sigma70 promoter prediction"]},"source_ids":["prokbert-2024"],"links":[{"relation":"subject","target_id":"lit-033"}],"attributes":{"field":"attributes.printed_value","value":"0.87","source_locator":"Table 3, ProkBERT-mini row, Accuracy column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:36.197Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness."}}}
{"id":"claim-lit-034","kind":"claim","name":"Reported Accuracy for Promotech","description":"","status":"source_checked","facets":{"areas":["microbes-communities"],"tasks":["E. coli sigma70 promoter prediction"]},"source_ids":["prokbert-2024"],"links":[{"relation":"subject","target_id":"lit-034"}],"attributes":{"field":"attributes.printed_value","value":"0.71","source_locator":"Table 3, Promotech row, Accuracy column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:36.197Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness."}}}
{"id":"claim-lit-035","kind":"claim","name":"Reported Promoter-class F1 for Eco70PromBERT","description":"","status":"source_checked","facets":{"areas":["microbes-communities"],"tasks":["E. coli sigma70 promoter prediction"]},"source_ids":["cyaprombert-2022"],"links":[{"relation":"subject","target_id":"lit-035"}],"attributes":{"field":"attributes.printed_value","value":"0.91","source_locator":"TABLE 3, Eco70PromBERT (BERT-base + 1bp tokenizer) row, F1 score Promoter column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.544033+00:00","notes":"F1 score is the third two-column group; selected Promoter subcolumn, not AUROC or precision. This verifies the central score at its source location, not every metadata field or an experimental reproduction."}}}
{"id":"claim-lit-036","kind":"claim","name":"Reported Promoter-class F1 for iPro70-FMWin","description":"","status":"source_checked","facets":{"areas":["microbes-communities"],"tasks":["E. coli sigma70 promoter prediction"]},"source_ids":["cyaprombert-2022"],"links":[{"relation":"subject","target_id":"lit-036"}],"attributes":{"field":"attributes.printed_value","value":"0.90","source_locator":"TABLE 3, iPro70-FMWin row, F1 score Promoter column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.544926+00:00","notes":"F1 score is the third two-column group; selected Promoter subcolumn, not AUROC or precision. This verifies the central score at its source location, not every metadata field or an experimental reproduction."}}}
{"id":"claim-lit-037","kind":"claim","name":"Reported MCC for EVO2","description":"","status":"source_checked","facets":{"areas":["microbes-communities"],"tasks":["Genome-wide prophage detection"]},"source_ids":["lambda-prophage-2026"],"links":[{"relation":"subject","target_id":"lit-037"}],"attributes":{"field":"attributes.printed_value","value":"0.680","source_locator":"Table 5, EVO2 row, MCC column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:36.240Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness."}}}
{"id":"claim-lit-038","kind":"claim","name":"Reported MCC for geNomad","description":"","status":"source_checked","facets":{"areas":["microbes-communities"],"tasks":["Genome-wide prophage detection"]},"source_ids":["lambda-prophage-2026"],"links":[{"relation":"subject","target_id":"lit-038"}],"attributes":{"field":"attributes.printed_value","value":"0.794","source_locator":"Table 5, geNomad row, MCC column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:36.240Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness."}}}
{"id":"claim-lit-039","kind":"claim","name":"Reported F1 score for NABAS+","description":"","status":"source_checked","facets":{"areas":["microbes-communities"],"tasks":["Metagenomic taxonomic classification"]},"source_ids":["nabas-plus-2025"],"links":[{"relation":"subject","target_id":"lit-039"}],"attributes":{"field":"attributes.printed_value","value":"0.719","source_locator":"Table 3, Sample19-new / NABAS+ row, F1 score column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.546107+00:00","notes":"Selected Sample19-new explicitly, not Sample19-old, and F1 rather than precision/recall. This verifies the central score at its source location, not every metadata field or an experimental reproduction."}}}
{"id":"claim-lit-040","kind":"claim","name":"Reported F1 score for MetaPhlAn3","description":"","status":"source_checked","facets":{"areas":["microbes-communities"],"tasks":["Metagenomic taxonomic classification"]},"source_ids":["nabas-plus-2025"],"links":[{"relation":"subject","target_id":"lit-040"}],"attributes":{"field":"attributes.printed_value","value":"0.753","source_locator":"Table 3, Sample19-new / MetaPhlAn3 row, F1 score column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.547023+00:00","notes":"Selected Sample19-new explicitly, not Sample19-old, and F1 rather than precision/recall. This verifies the central score at its source location, not every metadata field or an experimental reproduction."}}}
{"id":"claim-lit-041","kind":"claim","name":"Reported Success rate, ligand all-atom RMSD <2 Å for Chai-1","description":"","status":"source_checked","facets":{"areas":["molecular-interactions"],"tasks":["Lipid–protein binding pose"]},"source_ids":["lipp-2026"],"links":[{"relation":"subject","target_id":"lit-041"}],"attributes":{"field":"attributes.printed_value","value":"60.7","source_locator":"Table 2, Chai-1 row, LiPP (N=331) % Success Rate column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.548973+00:00","notes":"JATS label is bare 2, which caused original parser miss. LiPP N=331 full-set column selected, not N=36 test subset. Caption success is lipid all-atom RMSD <2 Angstrom; PB-valid is a separate table. This verifies the central score at its source location, not every metadata field or an experimental reproduction."}}}
{"id":"claim-lit-042","kind":"claim","name":"Reported Success rate, ligand all-atom RMSD <2 Å for DiffDock-L","description":"","status":"source_checked","facets":{"areas":["molecular-interactions"],"tasks":["Lipid–protein binding pose"]},"source_ids":["lipp-2026"],"links":[{"relation":"subject","target_id":"lit-042"}],"attributes":{"field":"attributes.printed_value","value":"46.8","source_locator":"Table 2, DiffDock-L row, LiPP (N=331) % Success Rate column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.550691+00:00","notes":"JATS label is bare 2, which caused original parser miss. LiPP N=331 full-set column selected, not N=36 test subset. Caption success is lipid all-atom RMSD <2 Angstrom; PB-valid is a separate table. This verifies the central score at its source location, not every metadata field or an experimental reproduction."}}}
{"id":"claim-lit-043","kind":"claim","name":"Reported Forward-screening success rate for DiffDock-NMDN","description":"","status":"source_checked","facets":{"areas":["molecular-interactions"],"tasks":["Protein–ligand virtual screening"]},"source_ids":["nmdn-2025"],"links":[{"relation":"subject","target_id":"lit-043"}],"attributes":{"field":"attributes.printed_value","value":"66.7","source_locator":"Table 2, DiffDock-NMDN / NMDN row, success rate (%) column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.552697+00:00","notes":"All scoring functions share DiffDock-NMDN poses via rowspan. Selected forward-screening success percentage, not docking pose success or scoring-power correlation. This verifies the central score at its source location, not every metadata field or an experimental reproduction."}}}
{"id":"claim-lit-044","kind":"claim","name":"Reported Forward-screening success rate for Vina","description":"","status":"source_checked","facets":{"areas":["molecular-interactions"],"tasks":["Protein–ligand virtual screening"]},"source_ids":["nmdn-2025"],"links":[{"relation":"subject","target_id":"lit-044"}],"attributes":{"field":"attributes.printed_value","value":"42.1","source_locator":"Table 2, Vina scoring row, success rate (%) column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.554518+00:00","notes":"All scoring functions share DiffDock-NMDN poses via rowspan. Selected forward-screening success percentage, not docking pose success or scoring-power correlation. This verifies the central score at its source location, not every metadata field or an experimental reproduction."}}}
{"id":"claim-lit-045","kind":"claim","name":"Reported Median ligand RMSD for Boltz-1","description":"","status":"source_checked","facets":{"areas":["molecular-interactions"],"tasks":["Protein–ligand pose prediction"]},"source_ids":["boltz-stereochemistry-2025"],"links":[{"relation":"subject","target_id":"lit-045"}],"attributes":{"field":"attributes.printed_value","value":"1.393","source_locator":"Table 1, Boltz-1 row, Ligand RMSD (Å) column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.555674+00:00","notes":"JATS label is bare 1. Selected Ligand RMSD column in Plinder-L95, not Protein RMSD; retained method row without refinement settings. This verifies the central score at its source location, not every metadata field or an experimental reproduction."}}}
{"id":"claim-lit-046","kind":"claim","name":"Reported Median ligand RMSD for DiffDock","description":"","status":"source_checked","facets":{"areas":["molecular-interactions"],"tasks":["Protein–ligand pose prediction"]},"source_ids":["boltz-stereochemistry-2025"],"links":[{"relation":"subject","target_id":"lit-046"}],"attributes":{"field":"attributes.printed_value","value":"1.342","source_locator":"Table 1, DiffDock row, Ligand RMSD (Å) column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.556572+00:00","notes":"JATS label is bare 1. Selected Ligand RMSD column in Plinder-L95, not Protein RMSD; retained method row without refinement settings. This verifies the central score at its source location, not every metadata field or an experimental reproduction."}}}
{"id":"claim-lit-047","kind":"claim","name":"Reported Pearson R for Boltz-2","description":"","status":"source_checked","facets":{"areas":["molecular-interactions"],"tasks":["Ligand potency prediction using generated poses"]},"source_ids":["mpro-pose-affinity-2025"],"links":[{"relation":"subject","target_id":"lit-047"}],"attributes":{"field":"attributes.printed_value","value":"0.800","source_locator":"Table 3, Boltz-2 row, Pearson’s R column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.557756+00:00","notes":"JATS label is bare 3. Selected SARS-CoV-2 Mpro potency Pearson R, not MERS-CoV table 2 or Boltz-2-Internal row; uncertainty retained in evidence. This verifies the central score at its source location, not every metadata field or an experimental reproduction."}}}
{"id":"claim-lit-048","kind":"claim","name":"Reported Pearson R for DiffDock","description":"","status":"source_checked","facets":{"areas":["molecular-interactions"],"tasks":["Ligand potency prediction using generated poses"]},"source_ids":["mpro-pose-affinity-2025"],"links":[{"relation":"subject","target_id":"lit-048"}],"attributes":{"field":"attributes.printed_value","value":"0.695","source_locator":"Table 3, DiffDock row, Pearson’s R column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.558815+00:00","notes":"JATS label is bare 3. Selected SARS-CoV-2 Mpro potency Pearson R, not MERS-CoV table 2 or Boltz-2-Internal row; uncertainty retained in evidence. This verifies the central score at its source location, not every metadata field or an experimental reproduction."}}}
{"id":"claim-lit-b3-003","kind":"claim","name":"Reported F1 for Mouse-Geneformer","description":"","status":"source_checked","facets":{"areas":["cells-tissues"],"tasks":["Human thymus cell-type classification"]},"source_ids":["mouse-geneformer-2025"],"links":[{"relation":"subject","target_id":"lit-b3-003"}],"attributes":{"field":"attributes.printed_value","value":"48.57","source_locator":"Table 4, h/ Thymus row, Mouse-Geneformer Zero-shot F1 column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:44:03.392488+00:00","notes":"h/ Thymus row, four human cell types; zero-shot model blocks use Acc then F1, not fine-tuned scores. Ortholog-based conversion evaluated, not mouse cell annotation. Source check verifies central value and table context, not experiment reproduction or all metadata."}}}
{"id":"claim-lit-b3-004","kind":"claim","name":"Reported F1 for Human-Geneformer","description":"","status":"source_checked","facets":{"areas":["cells-tissues"],"tasks":["Human thymus cell-type classification"]},"source_ids":["mouse-geneformer-2025"],"links":[{"relation":"subject","target_id":"lit-b3-004"}],"attributes":{"field":"attributes.printed_value","value":"74.48","source_locator":"Table 4, h/ Thymus row, Human-Geneformer Zero-shot F1 column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:44:03.394073+00:00","notes":"h/ Thymus row, four human cell types; zero-shot model blocks use Acc then F1, not fine-tuned scores. Ortholog-based conversion evaluated, not mouse cell annotation. Source check verifies central value and table context, not experiment reproduction or all metadata."}}}
{"id":"claim-lit-b3-005","kind":"claim","name":"Reported F1 for scLLMDA","description":"","status":"source_checked","facets":{"areas":["cells-tissues"],"tasks":["Cross-platform scATAC cell-type annotation"]},"source_ids":["scatac-llmda-2026"],"links":[{"relation":"subject","target_id":"lit-b3-005"}],"attributes":{"field":"attributes.printed_value","value":"0.6525","source_locator":"Table 2, scLLMDA row, Ref: MosA1 / Q: WholeBrainA F1 column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:44:03.395850+00:00","notes":"First reference/query block MosA1 to WholeBrainA, F1 second column in block; direction of transfer is part of protocol. Source check verifies central value and table context, not experiment reproduction or all metadata."}}}
{"id":"claim-lit-b3-006","kind":"claim","name":"Reported F1 for MINGLE","description":"","status":"source_checked","facets":{"areas":["cells-tissues"],"tasks":["Cross-platform scATAC cell-type annotation"]},"source_ids":["scatac-llmda-2026"],"links":[{"relation":"subject","target_id":"lit-b3-006"}],"attributes":{"field":"attributes.printed_value","value":"0.6256","source_locator":"Table 2, MINGLE row, Ref: MosA1 / Q: WholeBrainA F1 column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:44:03.397272+00:00","notes":"First reference/query block MosA1 to WholeBrainA, F1 second column in block; direction of transfer is part of protocol. Source check verifies central value and table context, not experiment reproduction or all metadata."}}}
{"id":"claim-lit-b3-011","kind":"claim","name":"Reported Adjusted Rand Index for GenePT-w","description":"","status":"source_checked","facets":{"areas":["cells-tissues"],"tasks":["Cell-type structure in frozen embeddings"]},"source_ids":["genept-2024"],"links":[{"relation":"subject","target_id":"lit-b3-011"}],"attributes":{"field":"attributes.printed_value","value":"0.54","source_locator":"Table 2, Aorta / Cell type row, GenePT-w ARI column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:44:03.399274+00:00","notes":"First Cell type row belongs to Aorta, not preceding Phenotype row or later organs. ARI is first in each three-metric method block, not AMI/ASW. Source check verifies central value and table context, not experiment reproduction or all metadata."}}}
{"id":"claim-lit-b3-012","kind":"claim","name":"Reported Adjusted Rand Index for scGPT","description":"","status":"source_checked","facets":{"areas":["cells-tissues"],"tasks":["Cell-type structure in frozen embeddings"]},"source_ids":["genept-2024"],"links":[{"relation":"subject","target_id":"lit-b3-012"}],"attributes":{"field":"attributes.printed_value","value":"0.47","source_locator":"Table 2, Aorta / Cell type row, scGPT ARI column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:44:03.402215+00:00","notes":"First Cell type row belongs to Aorta, not preceding Phenotype row or later organs. ARI is first in each three-metric method block, not AMI/ASW. Source check verifies central value and table context, not experiment reproduction or all metadata."}}}
{"id":"claim-lit-b3-013","kind":"claim","name":"Reported Balanced accuracy for Best frozen single-cell foundation model","description":"","status":"source_checked","facets":{"areas":["cells-tissues"],"tasks":["Donor-aware age-class prediction"]},"source_ids":["single-cell-aging-probes-2026"],"links":[{"relation":"subject","target_id":"lit-b3-013"}],"attributes":{"field":"attributes.printed_value","value":"0.322","source_locator":"Table 2, AIDA v2 row, scFM BA ± SD column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:44.421Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness."}}}
{"id":"claim-lit-b3-014","kind":"claim","name":"Reported Balanced accuracy for Gene-expression PCA","description":"","status":"source_checked","facets":{"areas":["cells-tissues"],"tasks":["Donor-aware age-class prediction"]},"source_ids":["single-cell-aging-probes-2026"],"links":[{"relation":"subject","target_id":"lit-b3-014"}],"attributes":{"field":"attributes.printed_value","value":"0.384","source_locator":"Table 2, AIDA v2 row, Gene-expr BA column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:44.421Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness."}}}
{"id":"claim-lit-b3-015","kind":"claim","name":"Reported Cell-type accuracy for scaLR","description":"","status":"source_checked","facets":{"areas":["cells-tissues"],"tasks":["PBMC cell-type classification"]},"source_ids":["scalr-2025"],"links":[{"relation":"subject","target_id":"lit-b3-015"}],"attributes":{"field":"attributes.printed_value","value":"0.942","source_locator":"Table 2, scaLR row, Cell type Accuracy column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:44:03.403844+00:00","notes":"PBMCs-BS all-feature/all-sample cell-type accuracy block, not cell-state accuracy or time. Footnote letters on model labels excluded from identity. Source check verifies central value and table context, not experiment reproduction or all metadata."}}}
{"id":"claim-lit-b3-016","kind":"claim","name":"Reported Cell-type accuracy for scVI + scANVI","description":"","status":"source_checked","facets":{"areas":["cells-tissues"],"tasks":["PBMC cell-type classification"]},"source_ids":["scalr-2025"],"links":[{"relation":"subject","target_id":"lit-b3-016"}],"attributes":{"field":"attributes.printed_value","value":"0.939","source_locator":"Table 2, Svi-tools (scVI & scANVI) row, Cell type Accuracy column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:44:03.405089+00:00","notes":"PBMCs-BS all-feature/all-sample cell-type accuracy block, not cell-state accuracy or time. Footnote letters on model labels excluded from identity. Source check verifies central value and table context, not experiment reproduction or all metadata."}}}
{"id":"claim-lit-b3-017","kind":"claim","name":"Reported AUC for scXDR","description":"","status":"source_checked","facets":{"areas":["cells-tissues"],"tasks":["Cross-dataset single-cell drug response transfer"]},"source_ids":["scxdr-2026"],"links":[{"relation":"subject","target_id":"lit-b3-017"}],"attributes":{"field":"attributes.printed_value","value":"0.8248","source_locator":"Table 2, scXDR row, Scenario 2 column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:50.056Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness."}}}
{"id":"claim-lit-b3-018","kind":"claim","name":"Reported AUC for scVI","description":"","status":"source_checked","facets":{"areas":["cells-tissues"],"tasks":["Cross-dataset single-cell drug response transfer"]},"source_ids":["scxdr-2026"],"links":[{"relation":"subject","target_id":"lit-b3-018"}],"attributes":{"field":"attributes.printed_value","value":"0.6970","source_locator":"Table 2, scVI row, Scenario 2 column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:50.056Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness."}}}
{"id":"claim-lit-b3-019","kind":"claim","name":"Reported L1 abundance error for CAMMiQ","description":"","status":"source_checked","facets":{"areas":["microbes-communities"],"tasks":["Strain-level abundance quantification"]},"source_ids":["cammiq-2022"],"links":[{"relation":"subject","target_id":"lit-b3-019"}],"attributes":{"field":"attributes.printed_value","value":"0.0517","source_locator":"Table 5, B. L1 Err. / HumanGut-all row, CAMMiQ L1 Err. column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:44:03.408237+00:00","notes":"HumanGut-all in B. L1 Err. block (second occurrence), not strain count or C. L2 Err. Numbers are errors; lower is better. Source check verifies central value and table context, not experiment reproduction or all metadata."}}}
{"id":"claim-lit-b3-020","kind":"claim","name":"Reported L1 abundance error for Kraken2","description":"","status":"source_checked","facets":{"areas":["microbes-communities"],"tasks":["Strain-level abundance quantification"]},"source_ids":["cammiq-2022"],"links":[{"relation":"subject","target_id":"lit-b3-020"}],"attributes":{"field":"attributes.printed_value","value":"0.2841","source_locator":"Table 5, B. L1 Err. / HumanGut-all row, Kraken2 L1 Err. column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:44:03.411180+00:00","notes":"HumanGut-all in B. L1 Err. block (second occurrence), not strain count or C. L2 Err. Numbers are errors; lower is better. Source check verifies central value and table context, not experiment reproduction or all metadata."}}}
{"id":"claim-lit-b3-021","kind":"claim","name":"Reported Genus-level F1 for Lazypipe-nt","description":"","status":"source_checked","facets":{"areas":["microbes-communities"],"tasks":["Simulated metagenome virus-taxon retrieval"]},"source_ids":["lazypipe-2020"],"links":[{"relation":"subject","target_id":"lit-b3-021"}],"attributes":{"field":"attributes.printed_value","value":"0.932","source_locator":"Table 1, Lazypipe-nt / Genus row, F column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:44:03.412605+00:00","notes":"First Genus block selected using rank row span, not Species. Final F column is F score, not precision or recall. Source check verifies central value and table context, not experiment reproduction or all metadata."}}}
{"id":"claim-lit-b3-022","kind":"claim","name":"Reported Genus-level F1 for Kraken2","description":"","status":"source_checked","facets":{"areas":["microbes-communities"],"tasks":["Simulated metagenome virus-taxon retrieval"]},"source_ids":["lazypipe-2020"],"links":[{"relation":"subject","target_id":"lit-b3-022"}],"attributes":{"field":"attributes.printed_value","value":"0.627","source_locator":"Table 1, Kraken2 / Genus row, F column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:44:03.413635+00:00","notes":"First Genus block selected using rank row span, not Species. Final F column is F score, not precision or recall. Source check verifies central value and table context, not experiment reproduction or all metadata."}}}
{"id":"claim-lit-b3-023","kind":"claim","name":"Reported Macro F1 for NCD-gzip","description":"","status":"source_checked","facets":{"areas":["microbes-communities"],"tasks":["CAMI II superkingdom read classification"]},"source_ids":["ncd-metagenomics-2026"],"links":[{"relation":"subject","target_id":"lit-b3-023"}],"attributes":{"field":"attributes.printed_value","value":"0.9804","source_locator":"Table 5, NCD Superkingdom row, F1 column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:44:03.414806+00:00","notes":"NCD rank-specific table 5, F1 column; Superkingdom and Phylum are different classification granularities. Source check verifies central value and table context, not experiment reproduction or all metadata."}}}
{"id":"claim-lit-b3-024","kind":"claim","name":"Reported Macro F1 for NCD-gzip","description":"","status":"source_checked","facets":{"areas":["microbes-communities"],"tasks":["CAMI II phylum read classification"]},"source_ids":["ncd-metagenomics-2026"],"links":[{"relation":"subject","target_id":"lit-b3-024"}],"attributes":{"field":"attributes.printed_value","value":"0.1263","source_locator":"Table 5, NCD Phylum row, F1 column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:44:03.415788+00:00","notes":"NCD rank-specific table 5, F1 column; Superkingdom and Phylum are different classification granularities. Source check verifies central value and table context, not experiment reproduction or all metadata."}}}
{"id":"claim-lit-b3-025","kind":"claim","name":"Reported Average prophage F1 for VIBRANT","description":"","status":"source_checked","facets":{"areas":["microbes-communities"],"tasks":["Simulated prophage-contig detection"]},"source_ids":["viral-contig-simulation-2021"],"links":[{"relation":"subject","target_id":"lit-b3-025"}],"attributes":{"field":"attributes.printed_value","value":"0.169","source_locator":"Table 3, Vibrant row, Prophage F1 column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:50.134Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness."}}}
{"id":"claim-lit-b3-026","kind":"claim","name":"Reported Average prophage F1 for VirSorter","description":"","status":"source_checked","facets":{"areas":["microbes-communities"],"tasks":["Simulated prophage-contig detection"]},"source_ids":["viral-contig-simulation-2021"],"links":[{"relation":"subject","target_id":"lit-b3-026"}],"attributes":{"field":"attributes.printed_value","value":"0.147","source_locator":"Table 3, VirSorter row, Prophage F1 column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:50.134Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness."}}}
{"id":"claim-lit-b3-027","kind":"claim","name":"Reported F1 for GenomeOcean","description":"","status":"source_checked","facets":{"areas":["microbes-communities"],"tasks":["Natural vs artificial microbial genome sequence"]},"source_ids":["genomeocean-2025"],"links":[{"relation":"subject","target_id":"lit-b3-027"}],"attributes":{"field":"attributes.printed_value","value":"99.03","source_locator":"Table 2, GenomeOcean row, F1 column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:55.224Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness."}}}
{"id":"claim-lit-b3-028","kind":"claim","name":"Reported F1 for DNABERT-2","description":"","status":"source_checked","facets":{"areas":["microbes-communities"],"tasks":["Natural vs artificial microbial genome sequence"]},"source_ids":["genomeocean-2025"],"links":[{"relation":"subject","target_id":"lit-b3-028"}],"attributes":{"field":"attributes.printed_value","value":"85.12","source_locator":"Table 2, DNABERT-2 row, F1 column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:55.224Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness."}}}
{"id":"claim-lit-b3-029","kind":"claim","name":"Reported Genus-level F1 for kMetaShot","description":"","status":"source_checked","facets":{"areas":["microbes-communities"],"tasks":["Mock-community MAG taxonomy classification"]},"source_ids":["kmetashot-2025"],"links":[{"relation":"subject","target_id":"lit-b3-029"}],"attributes":{"field":"attributes.printed_value","value":"95.83","source_locator":"Table 2, F1-score % row, Genus kMS column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:44:03.417367+00:00","notes":"Real mock sequencing table, F1-score percentage row; Genus is final three-column block, selecting kMS or Gtk rather than Species/Strain. Source check verifies central value and table context, not experiment reproduction or all metadata."}}}
{"id":"claim-lit-b3-030","kind":"claim","name":"Reported Genus-level F1 for GTDB-Tk","description":"","status":"source_checked","facets":{"areas":["microbes-communities"],"tasks":["Mock-community MAG taxonomy classification"]},"source_ids":["kmetashot-2025"],"links":[{"relation":"subject","target_id":"lit-b3-030"}],"attributes":{"field":"attributes.printed_value","value":"89.80","source_locator":"Table 2, F1-score % row, Genus Gtk column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:44:03.418914+00:00","notes":"Real mock sequencing table, F1-score percentage row; Genus is final three-column block, selecting kMS or Gtk rather than Species/Strain. Source check verifies central value and table context, not experiment reproduction or all metadata."}}}
{"id":"claim-lit-b3-031","kind":"claim","name":"Reported F1 for Lemur","description":"","status":"source_checked","facets":{"areas":["microbes-communities"],"tasks":["Long-read taxonomic profiling"]},"source_ids":["lemur-magnet-2024"],"links":[{"relation":"subject","target_id":"lit-b3-031"}],"attributes":{"field":"attributes.printed_value","value":"0.376","source_locator":"Table 3, LOG 10% / Lemur row, F1 column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:44:03.420493+00:00","notes":"LOG 10% first block, F1 column. Kraken 2 row inherits dataset via rowspan; not LOG 75% or abundance Spearman. Source check verifies central value and table context, not experiment reproduction or all metadata."}}}
{"id":"claim-lit-b3-032","kind":"claim","name":"Reported F1 for Kraken 2","description":"","status":"source_checked","facets":{"areas":["microbes-communities"],"tasks":["Long-read taxonomic profiling"]},"source_ids":["lemur-magnet-2024"],"links":[{"relation":"subject","target_id":"lit-b3-032"}],"attributes":{"field":"attributes.printed_value","value":"0.375","source_locator":"Table 3, LOG 10% / Kraken 2 row, F1 column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:44:03.421888+00:00","notes":"LOG 10% first block, F1 column. Kraken 2 row inherits dataset via rowspan; not LOG 75% or abundance Spearman. Source check verifies central value and table context, not experiment reproduction or all metadata."}}}
{"id":"claim-lit-b3-033","kind":"claim","name":"Reported Mean AUC for iPro-MP","description":"","status":"source_checked","facets":{"areas":["microbes-communities"],"tasks":["Multi-species prokaryotic promoter detection"]},"source_ids":["ipromp-2025"],"links":[{"relation":"subject","target_id":"lit-b3-033"}],"attributes":{"field":"attributes.printed_value","value":"0.935","source_locator":"Table 2, iPro-MP row, AUC column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:55.361Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness."}}}
{"id":"claim-lit-b3-034","kind":"claim","name":"Reported Mean AUC for Prompt","description":"","status":"source_checked","facets":{"areas":["microbes-communities"],"tasks":["Multi-species prokaryotic promoter detection"]},"source_ids":["ipromp-2025"],"links":[{"relation":"subject","target_id":"lit-b3-034"}],"attributes":{"field":"attributes.printed_value","value":"0.835","source_locator":"Table 2, Prompt row, AUC column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:55.361Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness."}}}
{"id":"claim-lit-b3-035","kind":"claim","name":"Reported Genus macro AveP for ICCTax","description":"","status":"source_checked","facets":{"areas":["microbes-communities"],"tasks":["Hierarchical metagenomic taxonomy classification"]},"source_ids":["icctax-2025"],"links":[{"relation":"subject","target_id":"lit-b3-035"}],"attributes":{"field":"attributes.printed_value","value":"67.20","source_locator":"Table 2, ICCTax row, Genus column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:55.373Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness."}}}
{"id":"claim-lit-b3-036","kind":"claim","name":"Reported Genus macro AveP for Kraken2","description":"","status":"source_checked","facets":{"areas":["microbes-communities"],"tasks":["Hierarchical metagenomic taxonomy classification"]},"source_ids":["icctax-2025"],"links":[{"relation":"subject","target_id":"lit-b3-036"}],"attributes":{"field":"attributes.printed_value","value":"70.56","source_locator":"Table 2, Kraken2 row, Genus column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:55.373Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness."}}}
{"id":"claim-lit-b3-037","kind":"claim","name":"Reported AUC-ROC for Chai-1","description":"","status":"source_checked","facets":{"areas":["molecular-interactions"],"tasks":["Antibody–antigen interaction prediction using folded complexes"]},"source_ids":["antibody-flexibility-2025"],"links":[{"relation":"subject","target_id":"lit-b3-037"}],"attributes":{"field":"attributes.printed_value","value":"0.86","source_locator":"Table 5, Folded row, Chai-1 (no MSA) column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:55.400Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness."}}}
{"id":"claim-lit-b3-038","kind":"claim","name":"Reported AUC-ROC for Boltz-1","description":"","status":"source_checked","facets":{"areas":["molecular-interactions"],"tasks":["Antibody–antigen interaction prediction using folded complexes"]},"source_ids":["antibody-flexibility-2025"],"links":[{"relation":"subject","target_id":"lit-b3-038"}],"attributes":{"field":"attributes.printed_value","value":"0.85","source_locator":"Table 5, Folded row, Boltz-1 (no MSA) column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:55.400Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness."}}}
{"id":"claim-lit-b3-039","kind":"claim","name":"Reported Top-1 ligand RMSD <2 Å rate for Boltz-1","description":"","status":"source_checked","facets":{"areas":["molecular-interactions"],"tasks":["Protein–ligand pose prediction"]},"source_ids":["boltz1-2025"],"links":[{"relation":"subject","target_id":"lit-b3-039"}],"attributes":{"field":"attributes.printed_value","value":"0.545","source_locator":"Table 1, 3 recycling rounds / 200 steps row, L-RMSD <2Å top-1 column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:44:03.424864+00:00","notes":"3 recycling rounds and 200 steps; L-RMSD <2 Angstrom top-1 (last column), not oracle. Five samples generated; top-1 means highest-confidence candidate. Repeated reference rows are one evaluation, not independent experiments. Source check verifies central value and table context, not experiment reproduction or all metadata."}}}
{"id":"claim-lit-b3-040","kind":"claim","name":"Reported Mean CDR H3 RMSD for Ibex","description":"","status":"source_checked","facets":{"areas":["molecular-interactions"],"tasks":["Antibody loop structure prediction"]},"source_ids":["ibex-2025"],"links":[{"relation":"subject","target_id":"lit-b3-040"}],"attributes":{"field":"attributes.printed_value","value":"2.72","source_locator":"Table 1, Antibodies / Ibex row, CDR H3 column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:44:03.426811+00:00","notes":"First Antibodies block, CDR H3 mean RMSD in Angstrom; excludes later Nanobodies and TCR blocks. Source check verifies central value and table context, not experiment reproduction or all metadata."}}}
{"id":"claim-lit-b3-041","kind":"claim","name":"Reported Mean CDR H3 RMSD for Chai-1","description":"","status":"source_checked","facets":{"areas":["molecular-interactions"],"tasks":["Antibody loop structure prediction"]},"source_ids":["ibex-2025"],"links":[{"relation":"subject","target_id":"lit-b3-041"}],"attributes":{"field":"attributes.printed_value","value":"2.65","source_locator":"Table 1, Antibodies / Chai-1 row, CDR H3 column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:44:03.428536+00:00","notes":"First Antibodies block, CDR H3 mean RMSD in Angstrom; excludes later Nanobodies and TCR blocks. Source check verifies central value and table context, not experiment reproduction or all metadata."}}}
{"id":"claim-lit-b3-042","kind":"claim","name":"Reported Pearson R for DEELIG","description":"","status":"source_checked","facets":{"areas":["molecular-interactions"],"tasks":["Protein–ligand binding affinity prediction"]},"source_ids":["deelig-2021"],"links":[{"relation":"subject","target_id":"lit-b3-042"}],"attributes":{"field":"attributes.printed_value","value":"0.889","source_locator":"Table 2, DEELIG row, PDBbind v2016 column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:55.586Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness."}}}
{"id":"claim-lit-b3-043","kind":"claim","name":"Reported Pearson R for TOPBP (Complex)","description":"","status":"source_checked","facets":{"areas":["molecular-interactions"],"tasks":["Protein–ligand binding affinity prediction"]},"source_ids":["deelig-2021"],"links":[{"relation":"subject","target_id":"lit-b3-043"}],"attributes":{"field":"attributes.printed_value","value":"0.861","source_locator":"Table 2, TOPBP (Complex) row, PDBbind v2016 column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:44:03.429669+00:00","notes":"TOPBP Complex reference row; PDBbind v2016 core-set Pearson correlation. Third-party comparator with cited reference; do not infer an independent new run from table inclusion. Source check verifies central value and table context, not experiment reproduction or all metadata."}}}
{"id":"claim-lit-b3-044","kind":"claim","name":"Reported RMSD ≤1 Å and PB-valid success for MolAS","description":"","status":"source_checked","facets":{"areas":["molecular-interactions"],"tasks":["Physically valid protein–ligand pose selection"]},"source_ids":["molas-2026"],"links":[{"relation":"subject","target_id":"lit-b3-044"}],"attributes":{"field":"attributes.printed_value","value":"36.69","source_locator":"Table 3, PoseBusters / Mixed / AutoDock row, MolAS success column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:44:03.432565+00:00","notes":"PoseBusters, Mixed, AutoDock row within jointly trained with/without relaxation block. Selected RMSD <=1 Angstrom AND PB-valid group; five-fold average success percentage, not <=2 Angstrom. Inline bold digit nodes joined in original order. Source check verifies central value and table context, not experiment reproduction or all metadata."}}}
{"id":"claim-lit-b3-045","kind":"claim","name":"Reported RMSD ≤1 Å and PB-valid success for Single best solver","description":"","status":"source_checked","facets":{"areas":["molecular-interactions"],"tasks":["Physically valid protein–ligand pose selection"]},"source_ids":["molas-2026"],"links":[{"relation":"subject","target_id":"lit-b3-045"}],"attributes":{"field":"attributes.printed_value","value":"34.34","source_locator":"Table 3, PoseBusters / Mixed / AutoDock row, SBS success column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:44:03.435454+00:00","notes":"PoseBusters, Mixed, AutoDock row within jointly trained with/without relaxation block. Selected RMSD <=1 Angstrom AND PB-valid group; five-fold average success percentage, not <=2 Angstrom. Inline bold digit nodes joined in original order. Source check verifies central value and table context, not experiment reproduction or all metadata."}}}
{"id":"claim-lit-b3-046","kind":"claim","name":"Reported Docked frames best-matched RMSD <3 Å for AutoDock Vina holo","description":"","status":"source_checked","facets":{"areas":["molecular-interactions"],"tasks":["Intrinsically disordered protein ensemble docking"]},"source_ids":["ensemble-idp-docking-2025"],"links":[{"relation":"subject","target_id":"lit-b3-046"}],"attributes":{"field":"attributes.printed_value","value":"27.96","source_locator":"Table 2, Ligand 47 row, AutoDock Vina Holo Docking column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:56.275Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness."}}}
{"id":"claim-lit-b3-047","kind":"claim","name":"Reported Docked frames best-matched RMSD <3 Å for DiffDock holo","description":"","status":"source_checked","facets":{"areas":["molecular-interactions"],"tasks":["Intrinsically disordered protein ensemble docking"]},"source_ids":["ensemble-idp-docking-2025"],"links":[{"relation":"subject","target_id":"lit-b3-047"}],"attributes":{"field":"attributes.printed_value","value":"21.32","source_locator":"Table 2, Ligand 47 row, DiffDock Holo Docking column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:56.275Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness."}}}
{"id":"claim-lit-b3-048","kind":"claim","name":"Reported Pearson R for AK-score-ensemble","description":"","status":"source_checked","facets":{"areas":["molecular-interactions"],"tasks":["Protein–ligand binding affinity scoring"]},"source_ids":["akscore-2020"],"links":[{"relation":"subject","target_id":"lit-b3-048"}],"attributes":{"field":"attributes.printed_value","value":"0.812","source_locator":"Table 2, AK-score-ensemble / learning rate 0.0007 row, Scoring Pearson (R) column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:44:03.436853+00:00","notes":"CASF-2016 scoring Pearson R with learning rate0.0007. Single-model versus ensemble blocks kept distinct; ranking/docking scores not substituted. Source check verifies central value and table context, not experiment reproduction or all metadata."}}}
{"id":"claim-lit-b3-049","kind":"claim","name":"Reported Pearson R for AK-score-single","description":"","status":"source_checked","facets":{"areas":["molecular-interactions"],"tasks":["Protein–ligand binding affinity scoring"]},"source_ids":["akscore-2020"],"links":[{"relation":"subject","target_id":"lit-b3-049"}],"attributes":{"field":"attributes.printed_value","value":"0.759","source_locator":"Table 2, AK-score-single / learning rate 0.0007 row, Scoring Pearson (R) column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:44:03.437894+00:00","notes":"CASF-2016 scoring Pearson R with learning rate0.0007. Single-model versus ensemble blocks kept distinct; ranking/docking scores not substituted. Source check verifies central value and table context, not experiment reproduction or all metadata."}}}
{"id":"claim-lit-b3-050","kind":"claim","name":"Reported Pearson R for PMF + ECFP + PF (LightGBM)","description":"","status":"source_checked","facets":{"areas":["molecular-interactions"],"tasks":["Protein–ligand binding energy prediction"]},"source_ids":["fingerprint-scoring-2022"],"links":[{"relation":"subject","target_id":"lit-b3-050"}],"attributes":{"field":"attributes.printed_value","value":"0.79","source_locator":"Table 1, PMF + ECFP + PF / LightGBM row, R column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:44:03.439460+00:00","notes":"Resolved two-row model rowspan: last LightGBM row is PMF+ECFP+PF; first LASSO row is PMF. Pearson R, not RMSE. Source check verifies central value and table context, not experiment reproduction or all metadata."}}}
{"id":"claim-lit-b3-051","kind":"claim","name":"Reported Pearson R for PMF (LASSO)","description":"","status":"source_checked","facets":{"areas":["molecular-interactions"],"tasks":["Protein–ligand binding energy prediction"]},"source_ids":["fingerprint-scoring-2022"],"links":[{"relation":"subject","target_id":"lit-b3-051"}],"attributes":{"field":"attributes.printed_value","value":"0.67","source_locator":"Table 1, PMF / LASSO row, R column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:44:03.440812+00:00","notes":"Resolved two-row model rowspan: last LightGBM row is PMF+ECFP+PF; first LASSO row is PMF. Pearson R, not RMSE. Source check verifies central value and table context, not experiment reproduction or all metadata."}}}
{"id":"claim-lit-b4-001","kind":"claim","name":"Reported AUROC for ARSENAL+ChromBPNet","description":"","status":"source_checked","facets":{"areas":["dna-genomes"],"tasks":["regulatory-variant scoring"]},"source_ids":["arsenal-regulatory-dna-2026"],"links":[{"relation":"subject","target_id":"lit-b4-001"}],"attributes":{"field":"attributes.printed_value","value":"0.896","source_locator":"Table 1, Yoruban LCL dsQTLs section, ARSENAL+ChromBPNet row, AUROC column","review":{"method":"independent_ai_table_review","reviewer":"Codex primary curator table review","reviewed_at":"2026-09-16T10:41:06Z","notes":"Yoruban LCL dsQTLs; ARSENAL+ChromBPNet AUROC Transcription verified; experimental claims not independently reproduced."}}}
{"id":"claim-lit-b4-002","kind":"claim","name":"Reported AUROC for PlantCAD2","description":"","status":"source_checked","facets":{"areas":["dna-genomes"],"tasks":["cross-species conservation prediction"]},"source_ids":["plantcad2-2025"],"links":[{"relation":"subject","target_id":"lit-b4-002"}],"attributes":{"field":"attributes.printed_value","value":"0.725","source_locator":"Table 1, Cross-species evolutionary conservation > Conservation within Andropogoneae (Genome-wide) row, PlantCAD2 AUROC entry","review":{"method":"independent_ai_table_review","reviewer":"Codex primary curator table review","reviewed_at":"2026-09-16T10:41:06Z","notes":"First comparison entry is PlantCAD2; AUROC is 0.725 versus comparator 0.691. Transcription verified; experimental claims not independently reproduced."}}}
{"id":"claim-lit-b4-003","kind":"claim","name":"Reported accuracy for Stacking-Auto","description":"","status":"source_checked","facets":{"areas":["dna-genomes"],"tasks":["enhancer prediction"]},"source_ids":["hi-enhancer-2025"],"links":[{"relation":"subject","target_id":"lit-b4-003"}],"attributes":{"field":"attributes.printed_value","value":"80.50","source_locator":"Table 2, Ours (Stacking-Auto) row, Accuracy column","review":{"method":"independent_ai_table_review","reviewer":"Codex primary curator table review","reviewed_at":"2026-09-16T10:41:06Z","notes":"Ours row, Accuracy column; original source method is the Stacking-Auto stage. Transcription verified; experimental claims not independently reproduced."}}}
{"id":"claim-lit-b4-004","kind":"claim","name":"Reported AUROC for position-aware CNN","description":"","status":"source_checked","facets":{"areas":["dna-genomes"],"tasks":["enhancer prediction"]},"source_ids":["enhancer-position-encoding-2024"],"links":[{"relation":"subject","target_id":"lit-b4-004"}],"attributes":{"field":"attributes.printed_value","value":"0.94","source_locator":"Table 2, Human section, CNN row, AUC column","review":{"method":"independent_ai_table_review","reviewer":"Codex primary curator table review","reviewed_at":"2026-09-16T10:41:06Z","notes":"Human dataset row, CNN, AUC column. Transcription verified; experimental claims not independently reproduced."}}}
{"id":"claim-lit-b4-005","kind":"claim","name":"Reported F1 for ADAR-GPT continual","description":"","status":"source_checked","facets":{"areas":["rna-transcriptomes"],"tasks":["A-to-I RNA editing site prediction"]},"source_ids":["adar-gpt-editing-2026"],"links":[{"relation":"subject","target_id":"lit-b4-005"}],"attributes":{"field":"attributes.printed_value","value":"0.763","source_locator":"Table 2, Adar-GPT (continual) row, F1 column","review":{"method":"independent_ai_table_review","reviewer":"Codex primary curator table review","reviewed_at":"2026-09-16T10:41:06Z","notes":"Adar-GPT continual row; XML inline decimal reordered by parser, original text verified separately. Transcription verified; experimental claims not independently reproduced."}}}
{"id":"claim-lit-b4-006","kind":"claim","name":"Reported sequence recovery for R3Design","description":"","status":"source_checked","facets":{"areas":["rna-transcriptomes"],"tasks":["RNA sequence design"]},"source_ids":["r3design-2025"],"links":[{"relation":"subject","target_id":"lit-b4-006"}],"attributes":{"field":"attributes.printed_value","value":"43.27","source_locator":"Table 3, R3Design row, Recovery (%) > Rfam column","review":{"method":"independent_ai_table_review","reviewer":"Codex primary curator table review","reviewed_at":"2026-09-16T10:41:06Z","notes":"R3Design row, first Recovery column Rfam; 43.27 plus/minus0.56. Transcription verified; experimental claims not independently reproduced."}}}
{"id":"claim-lit-b4-007","kind":"claim","name":"Reported AUROC for CUPID Data-aug-Avg","description":"","status":"source_checked","facets":{"areas":["rna-transcriptomes"],"tasks":["non-coding RNA pairwise interaction prediction"]},"source_ids":["cupid-rna-interactions-2026"],"links":[{"relation":"subject","target_id":"lit-b4-007"}],"attributes":{"field":"attributes.printed_value","value":"0.919","source_locator":"Table 1, CUPID > Data-aug-Avg row, AUROC column","review":{"method":"independent_ai_table_review","reviewer":"Codex primary curator table review","reviewed_at":"2026-09-16T10:41:06Z","notes":"CUPID section Data-aug-Avg row; AUROC not AUPRC. Transcription verified; experimental claims not independently reproduced."}}}
{"id":"claim-lit-b4-008","kind":"claim","name":"Reported AUROC for ProteinBERT LLM-encoding model","description":"","status":"source_checked","facets":{"areas":["rna-transcriptomes"],"tasks":["mRNA-protein interaction prediction"]},"source_ids":["mrna-protein-diversity-2026"],"links":[{"relation":"subject","target_id":"lit-b4-008"}],"attributes":{"field":"attributes.printed_value","value":"71.5","source_locator":"Table 2, RBP-aware test set row, auROC (%) column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:57.257Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness."}}}
{"id":"claim-lit-b4-009","kind":"claim","name":"Reported AUROC for ESM2 650M","description":"","status":"source_checked","facets":{"areas":["proteins-complexes"],"tasks":["human-versus-viral protein classification"]},"source_ids":["viral-immune-mimicry-2025"],"links":[{"relation":"subject","target_id":"lit-b4-009"}],"attributes":{"field":"attributes.printed_value","value":"99.67","source_locator":"Table 1, ESM2 650M row, AUC (%) column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:57.274Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness."}}}
{"id":"claim-lit-b4-010","kind":"claim","name":"Reported AUROC for ProtT5 embeddings + ensemble classifier","description":"","status":"source_checked","facets":{"areas":["proteins-complexes"],"tasks":["protein-protein binding-site prediction"]},"source_ids":["protein-binding-sites-2023"],"links":[{"relation":"subject","target_id":"lit-b4-010"}],"attributes":{"field":"attributes.printed_value","value":"0.810","source_locator":"Table 2, Dset_448 section, ProtT5 row, AUROC column","review":{"method":"independent_ai_table_review","reviewer":"Codex primary curator table review","reviewed_at":"2026-09-16T10:41:06Z","notes":"Dset_448 block; ProtT5 AUROC, downstream ensemble retained in protocol. Transcription verified; experimental claims not independently reproduced."}}}
{"id":"claim-lit-b4-011","kind":"claim","name":"Reported AUROC for CLAPE-SMB with ESM-2","description":"","status":"source_checked","facets":{"areas":["proteins-complexes"],"tasks":["protein-small molecule binding-site prediction"]},"source_ids":["clape-smb-2024"],"links":[{"relation":"subject","target_id":"lit-b4-011"}],"attributes":{"field":"attributes.printed_value","value":"0.917","source_locator":"Table 5, ESM-2 / SJC row, AUROC column","review":{"method":"independent_ai_table_review","reviewer":"Codex primary curator table review","reviewed_at":"2026-09-16T10:41:06Z","notes":"ESM-2 on SJC AUROC. Transcription verified; experimental claims not independently reproduced."}}}
{"id":"claim-lit-b4-012","kind":"claim","name":"Reported AUPRC for Vaxign-DL + ESM","description":"","status":"source_checked","facets":{"areas":["proteins-complexes"],"tasks":["vaccine-antigen candidate prediction"]},"source_ids":["vaxign-esm-2024"],"links":[{"relation":"subject","target_id":"lit-b4-012"}],"attributes":{"field":"attributes.printed_value","value":"0.92","source_locator":"Table 2, 4 Layers row, AUPRC column","review":{"method":"independent_ai_table_review","reviewer":"Codex primary curator table review","reviewed_at":"2026-09-16T10:41:06Z","notes":"Source spells 4 Layerss; AUPRC0.92±0.013. Transcription verified; experimental claims not independently reproduced."}}}
{"id":"claim-lit-b4-013","kind":"claim","name":"Reported AUROC for scGPT + residual geometry","description":"","status":"source_checked","facets":{"areas":["cells-tissues"],"tasks":["gene-regulatory signal prediction"]},"source_ids":["single-cell-residual-geometry-2026"],"links":[{"relation":"subject","target_id":"lit-b4-013"}],"attributes":{"field":"attributes.printed_value","value":"0.677","source_locator":"Table 4, Immune row, scGPT > +geom AUROC column","review":{"method":"independent_ai_table_review","reviewer":"Codex primary curator table review","reviewed_at":"2026-09-16T10:41:06Z","notes":"Immune row, scGPT +geom (second numeric column), not Geneformer or delta. Transcription verified; experimental claims not independently reproduced."}}}
{"id":"claim-lit-b4-014","kind":"claim","name":"Reported F1 for GREmLN","description":"","status":"source_checked","facets":{"areas":["cells-tissues"],"tasks":["cell-type annotation"]},"source_ids":["gremln-2026"],"links":[{"relation":"subject","target_id":"lit-b4-014"}],"attributes":{"field":"attributes.printed_value","value":"0.937","source_locator":"Table 2, Cell type annotation(zero-shot), Non-immune cells, F1 row, GREmLN column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:57.502Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness."}}}
{"id":"claim-lit-b4-015","kind":"claim","name":"Reported F1 for Cell-DINO ViT-L","description":"","status":"source_checked","facets":{"areas":["cells-tissues"],"tasks":["protein localization classification"]},"source_ids":["cell-dino-2025"],"links":[{"relation":"subject","target_id":"lit-b4-015"}],"attributes":{"field":"attributes.printed_value","value":"65.5","source_locator":"Table 2, HPA-FoV section, Cell-DINO row, PL column","review":{"method":"independent_ai_table_review","reviewer":"Codex primary curator table review","reviewed_at":"2026-09-16T10:41:06Z","notes":"HPA-FoV Cell-DINO PL column (protein localisation), not CL. Transcription verified; experimental claims not independently reproduced."}}}
{"id":"claim-lit-b4-016","kind":"claim","name":"Reported precision at 50% recall for scGen","description":"","status":"source_checked","facets":{"areas":["cells-tissues"],"tasks":["differentially expressed gene identification"]},"source_ids":["insilico-perturbation-auprc-2025"],"links":[{"relation":"subject","target_id":"lit-b4-016"}],"attributes":{"field":"attributes.printed_value","value":"0.91","source_locator":"Table 3, CD14+Mono section, scGen row, Precision at 50% Recall column","review":{"method":"independent_ai_table_review","reviewer":"Codex primary curator table review","reviewed_at":"2026-09-16T10:41:06Z","notes":"CD14+Mono scGen; precision at50%recall. Transcription verified; experimental claims not independently reproduced."}}}
{"id":"claim-lit-b4-017","kind":"claim","name":"Reported F1 for TCINet + HTRS","description":"","status":"source_checked","facets":{"areas":["microbes-communities"],"tasks":["pathogen detection"]},"source_ids":["metagenomic-pathogens-2025"],"links":[{"relation":"subject","target_id":"lit-b4-017"}],"attributes":{"field":"attributes.printed_value","value":"0.84","source_locator":"Table 3, MetaHIT dataset section, TCINet + HTRS (Ours) row, F1-score column","review":{"method":"independent_ai_table_review","reviewer":"Codex primary curator table review","reviewed_at":"2026-09-16T10:41:06Z","notes":"MetaHIT block TCINet+HTRS F1-score. Transcription verified; experimental claims not independently reproduced."}}}
{"id":"claim-lit-b4-018","kind":"claim","name":"Reported accuracy for DETIRE","description":"","status":"source_checked","facets":{"areas":["microbes-communities"],"tasks":["viral sequence detection"]},"source_ids":["detire-viral-metagenomes-2023"],"links":[{"relation":"subject","target_id":"lit-b4-018"}],"attributes":{"field":"attributes.printed_value","value":"0.8772","source_locator":"Table 1, Accuracy row, DETIRE column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:58.392Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness."}}}
{"id":"claim-lit-b4-019","kind":"claim","name":"Reported accuracy for PC-mer + LR","description":"","status":"source_checked","facets":{"areas":["microbes-communities"],"tasks":["metagenomic genus classification"]},"source_ids":["pc-mer-2024"],"links":[{"relation":"subject","target_id":"lit-b4-019"}],"attributes":{"field":"attributes.printed_value","value":"96.95","source_locator":"Table 3, AMP section, PC-mer + LR k=8 row, Accuracy (%) column","review":{"method":"independent_ai_table_review","reviewer":"Codex primary curator table review","reviewed_at":"2026-09-16T10:41:06Z","notes":"AMP block PC-mer+LR section, k=8, first numeric value after k is Accuracy. Transcription verified; experimental claims not independently reproduced."}}}
{"id":"claim-lit-b4-020","kind":"claim","name":"Reported accuracy for MDL4Microbiome","description":"","status":"source_checked","facets":{"areas":["microbes-communities"],"tasks":["microbiome disease-state classification"]},"source_ids":["mdl4microbiome-2022"],"links":[{"relation":"subject","target_id":"lit-b4-020"}],"attributes":{"field":"attributes.printed_value","value":"0.97","source_locator":"Table 3, CRC row, MDL4Microbiome column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:58.492Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness."}}}
{"id":"claim-lit-b4-021","kind":"claim","name":"Reported Pearson correlation for binding-affinity meta-model","description":"","status":"source_checked","facets":{"areas":["molecular-interactions"],"tasks":["protein-ligand binding affinity prediction"]},"source_ids":["ligand-affinity-meta-model-2024"],"links":[{"relation":"subject","target_id":"lit-b4-021"}],"attributes":{"field":"attributes.printed_value","value":"0.777","source_locator":"Table 4, Meta-models row, CASF-2016 Benchmark > PCC column","review":{"method":"independent_ai_table_review","reviewer":"Codex primary curator table review","reviewed_at":"2026-09-16T10:41:06Z","notes":"Meta-models CASF-2016 PCC. Confirmed XML training-set rowspan inherits preceding row, so0.777 maps to PCC. Transcription verified; experimental claims not independently reproduced."}}}
{"id":"claim-lit-b4-022","kind":"claim","name":"Reported AUROC for DeepInterAware","description":"","status":"source_checked","facets":{"areas":["molecular-interactions"],"tasks":["antigen-antibody HIV neutralization prediction"]},"source_ids":["deepinteraware-2025"],"links":[{"relation":"subject","target_id":"lit-b4-022"}],"attributes":{"field":"attributes.printed_value","value":"0.826","source_locator":"Table 2, Ab Unseen section, DeepInterAware row, AUROC column","review":{"method":"independent_ai_table_review","reviewer":"Codex primary curator table review","reviewed_at":"2026-09-16T10:41:06Z","notes":"Ab Unseen block DeepInterAware AUROC0.826±0.017, not Ag Unseen. Transcription verified; experimental claims not independently reproduced."}}}
{"id":"claim-lit-b4-023","kind":"claim","name":"Reported AUROC for TransBind","description":"","status":"source_checked","facets":{"areas":["molecular-interactions"],"tasks":["transcription-factor DNA binding-site prediction"]},"source_ids":["transbind-2026"],"links":[{"relation":"subject","target_id":"lit-b4-023"}],"attributes":{"field":"attributes.printed_value","value":"0.9508","source_locator":"Table 2, TransBind row, AUROC column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:58.585Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness."}}}
{"id":"claim-lit-b4-024","kind":"claim","name":"Reported AUROC for ESM2_AMPS","description":"","status":"source_checked","facets":{"areas":["molecular-interactions"],"tasks":["protein-protein interaction prediction"]},"source_ids":["esm2-amp-2025"],"links":[{"relation":"subject","target_id":"lit-b4-024"}],"attributes":{"field":"attributes.printed_value","value":"0.68","source_locator":"Table 4, ESM2_AMPS row, AUROC column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:58.625Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness."}}}
{"id":"clape-smb-2024","kind":"source","name":"Protein-small molecule binding site prediction based on a pre-trained protein language model with contrastive learning","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["proteins-complexes"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11542454/","version":"version of record","retrieved_at":"2026-09-15T23:37:05Z","doi":"10.1186/s13321-024-00920-2","publication_status":"peer_reviewed","year":2024,"artifact_sha256":"215919244c3dd2dfb0b55fce91c211430fd8d4aee4bb28bd03eab9f4feb73e62","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11542454/fullTextXML","artifact_retrieved_at":"2026-09-16T10:41:06Z","legacy_paper":{"id":"clape-smb-2024","title":"Protein-small molecule binding site prediction based on a pre-trained protein language model with contrastive learning","year":2024,"publication_status":"peer_reviewed","version":"version of record","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11542454/","primary_domain":"proteins-complexes","retrieved_utc":"2026-09-15T23:37:05Z","notes":"Primary full text verified via Europe PMC fullTextXML; venue: Journal of Cheminformatics; PMC ID: PMC11542454. ESM-2 feature extractor embedded in CLAPE-SMB; score belongs to combined downstream system.","doi":"10.1186/s13321-024-00920-2"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"clathrin-plm-2025","kind":"source","name":"Advancing the accuracy of clathrin protein prediction through multi-source protein language models","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["proteins-complexes"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12238356/","version":"journal full text in PMC","retrieved_at":"2026-09-15T23:33:26Z","doi":"10.1038/s41598-025-08510-4","publication_status":"peer_reviewed","year":2025,"artifact_sha256":"2edc86b25707c1b737d26117093ce8d856e79cc5d0b335f27c1c341f887f1c7e","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12238356/fullTextXML","artifact_retrieved_at":"2026-09-16T10:38:57.558194+00:00","legacy_paper":{"id":"clathrin-plm-2025","title":"Advancing the accuracy of clathrin protein prediction through multi-source protein language models","year":2025,"publication_status":"peer_reviewed","version":"journal full text in PMC","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12238356/","primary_domain":"proteins-complexes","retrieved_utc":"2026-09-15T23:33:26Z","doi":"10.1038/s41598-025-08510-4","notes":"Numeric result checked against Table 2 in primary full-text XML; journal/source: Scientific Reports."},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"cobra-rna-binding-2026","kind":"source","name":"CoBRA: compound binding site prediction using RNA language model","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["rna-transcriptomes"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12790621/","version":"journal full text in PMC","retrieved_at":"2026-09-15T23:33:26Z","doi":"10.1093/bib/bbaf713","publication_status":"peer_reviewed","year":2026,"artifact_sha256":"8c6a6f00f5fa5f62acf301a66e9e6fa9ef11c7a05ad9b7447d2ade2ce8eba793","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12790621/fullTextXML","artifact_retrieved_at":"2026-09-16T10:38:57.558197+00:00","legacy_paper":{"id":"cobra-rna-binding-2026","title":"CoBRA: compound binding site prediction using RNA language model","year":2026,"publication_status":"peer_reviewed","version":"journal full text in PMC","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12790621/","primary_domain":"rna-transcriptomes","retrieved_utc":"2026-09-15T23:33:26Z","doi":"10.1093/bib/bbaf713","notes":"Numeric result checked against Table 2 in primary full-text XML; journal/source: Briefings in Bioinformatics."},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"codonbert-vaccines-2024","kind":"source","name":"CodonBERT large language model for mRNA vaccines","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["rna-transcriptomes"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11368176/","version":"journal full text in PMC","retrieved_at":"2026-09-15T23:33:26Z","doi":"10.1101/gr.278870.123","publication_status":"peer_reviewed","year":2024,"artifact_sha256":"2968073753e6d44feff9c08b131edf23145e95b171434539dddf77bb92847033","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11368176/fullTextXML","artifact_retrieved_at":"2026-09-16T10:38:57.558201+00:00","legacy_paper":{"id":"codonbert-vaccines-2024","title":"CodonBERT large language model for mRNA vaccines","year":2024,"publication_status":"peer_reviewed","version":"journal full text in PMC","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11368176/","primary_domain":"rna-transcriptomes","retrieved_utc":"2026-09-15T23:33:26Z","doi":"10.1101/gr.278870.123","notes":"Numeric result checked against Table 2. in primary full-text XML; journal/source: Genome Research."},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"cupid-rna-interactions-2026","kind":"source","name":"Computational understanding of non-coding RNA pairwise interactions","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["rna-transcriptomes"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12957212/","version":"PMC archival version PMC12957212.1","retrieved_at":"2026-09-15T23:37:05Z","doi":"10.3389/frai.2026.1749205","publication_status":"peer_reviewed","year":2026,"artifact_sha256":"0e6719410b390ee9c4858bb9321042851100fb74df3aa109bf2af2b8aaff7ac1","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12957212/fullTextXML","artifact_retrieved_at":"2026-09-16T10:41:06Z","legacy_paper":{"id":"cupid-rna-interactions-2026","title":"Computational understanding of non-coding RNA pairwise interactions","year":2026,"publication_status":"peer_reviewed","version":"PMC archival version PMC12957212.1","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12957212/","primary_domain":"rna-transcriptomes","retrieved_utc":"2026-09-15T23:37:05Z","notes":"Primary full text verified via Europe PMC fullTextXML; venue: Frontiers in Artificial Intelligence; PMC ID: PMC12957212. RNA-RNA pairwise interaction predictor; not a foundation model.","doi":"10.3389/frai.2026.1749205"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"cyaprombert-2022","kind":"source","name":"TSSNote-CyaPromBERT: Development of an integrated platform for highly accurate promoter prediction and visualization of Synechococcus sp. and Synechocystis sp. through a state-of-the-art natural language processing model BERT","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["microbes-communities"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC9745317/","version":"version of record","retrieved_at":"2026-09-15T23:25:00Z","doi":"10.3389/fgene.2022.1067562","publication_status":"peer_reviewed","year":2022,"artifact_sha256":"74278ccd77b2bc00a3f4434546545e8bdec8b0652a0e5d1862ec0f91decccd8d","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC9745317/fullTextXML","artifact_retrieved_at":"2026-09-16T10:41:16.544033+00:00","legacy_paper":{"id":"cyaprombert-2022","title":"TSSNote-CyaPromBERT: Development of an integrated platform for highly accurate promoter prediction and visualization of Synechococcus sp. and Synechocystis sp. through a state-of-the-art natural language processing model BERT","year":2022,"publication_status":"peer_reviewed","version":"version of record","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC9745317/","primary_domain":"microbes-communities","retrieved_utc":"2026-09-15T23:25:00Z","notes":"Primary full text via Europe PMC XML; venue: Frontiers in Genetics; PMC ID: PMC9745317.","doi":"10.3389/fgene.2022.1067562"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"dart-eval-regulatory-2024","kind":"source","name":"DART-Eval: A Comprehensive DNA Language Model Evaluation Benchmark on Regulatory DNA","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["dna-genomes"]},"source_ids":[],"links":[],"attributes":{"url":"https://proceedings.neurips.cc/paper_files/paper/2024/file/71998bfc3217ffe1cca1ee084dfadadd-Paper-Datasets_and_Benchmarks_Track.pdf","version":"NeurIPS 2024 Datasets and Benchmarks Track proceedings","retrieved_at":"2026-09-15T23:33:26Z","doi":"10.52202/079017-1981","publication_status":"peer_reviewed","year":2024,"artifact_sha256":"e5aee5b1f7cc6fd961b1d2a131d02cf243b79e091d5e418fbabee7fde9b39b22","artifact_url":"https://proceedings.neurips.cc/paper_files/paper/2024/file/71998bfc3217ffe1cca1ee084dfadadd-Paper-Datasets_and_Benchmarks_Track.pdf","artifact_retrieved_at":"2026-09-16T10:38:57.558203+00:00","legacy_paper":{"id":"dart-eval-regulatory-2024","title":"DART-Eval: A Comprehensive DNA Language Model Evaluation Benchmark on Regulatory DNA","year":2024,"publication_status":"peer_reviewed","version":"NeurIPS 2024 Datasets and Benchmarks Track proceedings","source_url":"https://proceedings.neurips.cc/paper_files/paper/2024/file/71998bfc3217ffe1cca1ee084dfadadd-Paper-Datasets_and_Benchmarks_Track.pdf","primary_domain":"dna-genomes","retrieved_utc":"2026-09-15T23:33:26Z","notes":"Proceedings Table 3, DNABERT-2 Zero-Shot Accuracy 0.876 checked directly; the PMC/arXiv manuscript carries the same printed row.","doi":"10.52202/079017-1981"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"debfold-2024","kind":"source","name":"DEBFold: Computational Identification of RNA Secondary Structures for Sequences across Structural Families Using Deep Learning","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["rna-transcriptomes"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11094721/","version":"PMC11094721.1","retrieved_at":"2026-09-15T23:25:00Z","doi":"10.1021/acs.jcim.4c00458","publication_status":"peer_reviewed","year":2024,"artifact_sha256":"e8f960eafb7f00edfdd81d4fb75c6de838e9b872b7e18875fc7a5bff2a2f72b3","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11094721/fullTextXML","artifact_retrieved_at":"2026-09-16T10:41:16.509290+00:00","legacy_paper":{"id":"debfold-2024","title":"DEBFold: Computational Identification of RNA Secondary Structures for Sequences across Structural Families Using Deep Learning","year":2024,"publication_status":"peer_reviewed","version":"PMC11094721.1","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11094721/","primary_domain":"rna-transcriptomes","retrieved_utc":"2026-09-15T23:25:00Z","notes":"Primary full text via Europe PMC XML; venue: Journal of Chemical Information and Modeling; PMC ID: PMC11094721.","doi":"10.1021/acs.jcim.4c00458"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"deelig-2021","kind":"source","name":"DEELIG: A Deep Learning Approach to Predict Protein-Ligand Binding Affinity","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["molecular-interactions"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC8274096/","version":"PMC archival version PMC8274096.1","retrieved_at":"2026-09-15T23:29:32Z","doi":"10.1177/11779322211030364","publication_status":"peer_reviewed","year":2021,"artifact_sha256":"5a7620c18d0622561004e1e25b5cfaf7399e93df3547eeefdd4cf6d300bb8aba","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC8274096/fullTextXML","artifact_retrieved_at":"2026-09-16T10:33:55.586Z","legacy_paper":{"id":"deelig-2021","title":"DEELIG: A Deep Learning Approach to Predict Protein-Ligand Binding Affinity","year":2021,"publication_status":"peer_reviewed","version":"PMC archival version PMC8274096.1","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC8274096/","primary_domain":"molecular-interactions","retrieved_utc":"2026-09-15T23:29:32Z","notes":"Primary full text verified using Europe PMC XML; venue: Bioinformatics and Biology Insights; PMC ID: PMC8274096.","doi":"10.1177/11779322211030364"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"deepinteraware-2025","kind":"source","name":"DeepInterAware: Deep Interaction Interface‐Aware Network for Improving Antigen‐Antibody Interaction Prediction from Sequence Data","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["molecular-interactions"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11967782/","version":"version of record","retrieved_at":"2026-09-15T23:37:05Z","doi":"10.1002/advs.202412533","publication_status":"peer_reviewed","year":2025,"artifact_sha256":"25d3561934965f754d8712ec02b2052e9a3979e433b88ecebd5db14e930f17a1","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11967782/fullTextXML","artifact_retrieved_at":"2026-09-16T10:41:06Z","legacy_paper":{"id":"deepinteraware-2025","title":"DeepInterAware: Deep Interaction Interface‐Aware Network for Improving Antigen‐Antibody Interaction Prediction from Sequence Data","year":2025,"publication_status":"peer_reviewed","version":"version of record","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11967782/","primary_domain":"molecular-interactions","retrieved_utc":"2026-09-15T23:37:05Z","notes":"Primary full text verified via Europe PMC fullTextXML; venue: Advanced Science; PMC ID: PMC11967782. Neutralization prediction, not generic binding affinity; uncertainty printed in source table.","doi":"10.1002/advs.202412533"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"detire-viral-metagenomes-2023","kind":"source","name":"DETIRE: a hybrid deep learning model for identifying viral sequences from metagenomes","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["microbes-communities"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC10313334/","version":"PMC archival version PMC10313334.1","retrieved_at":"2026-09-15T23:37:05Z","doi":"10.3389/fmicb.2023.1169791","publication_status":"peer_reviewed","year":2023,"artifact_sha256":"9ff7d32758620f7b0b0628425f62abff103ca2e33269ce3763383584bcebfc3c","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC10313334/fullTextXML","artifact_retrieved_at":"2026-09-16T10:33:58.392Z","legacy_paper":{"id":"detire-viral-metagenomes-2023","title":"DETIRE: a hybrid deep learning model for identifying viral sequences from metagenomes","year":2023,"publication_status":"peer_reviewed","version":"PMC archival version PMC10313334.1","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC10313334/","primary_domain":"microbes-communities","retrieved_utc":"2026-09-15T23:37:05Z","notes":"Primary full text verified via Europe PMC fullTextXML; venue: Frontiers in Microbiology; PMC ID: PMC10313334. Task-specific viral classifier, included as a microbial metagenomics benchmark.","doi":"10.3389/fmicb.2023.1169791"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"attributes":{"applicability":"proposed","baseline_type":"null-control","missing_metadata":{"implementation_version":"unextracted","protocol_applicability":"unextracted"},"requirements":"Matched control cells, normalization and evaluation gene set.","scope_note":"Source documents method or task context; proposed applicability requires protocol review."},"description":"Candidate reference for task-specific evaluation; no score is implied.","facets":{"areas":["single-cell"]},"id":"discovery-baseline-cell-perturbation-no-change","kind":"baseline","links":[{"relation":"applicable_to","target_id":"discovery-benchmark-perturbench"}],"name":"Cell perturbation no-change","source_ids":["src-discovery-altoslabs-perturbench"],"status":"discovered"}
{"attributes":{"applicability":"proposed","baseline_type":"biological-procedure","missing_metadata":{"implementation_version":"unextracted","protocol_applicability":"unextracted"},"requirements":"Receptor preparation, search box, exhaustiveness and conformers.","scope_note":"Source documents method or task context; proposed applicability requires protocol review."},"description":"Candidate reference for task-specific evaluation; no score is implied.","facets":{"areas":["molecular-interactions"]},"id":"discovery-baseline-classical-molecular-docking","kind":"baseline","links":[{"relation":"applicable_to","target_id":"discovery-benchmark-plinder"},{"relation":"model","target_id":"discovery-model-autodock-vina"}],"name":"Classical molecular docking","source_ids":["src-discovery-ccsb-scripps-autodock-vina"],"status":"discovered"}
{"attributes":{"applicability":"proposed","baseline_type":"null-control","missing_metadata":{"implementation_version":"unextracted","protocol_applicability":"unextracted"},"requirements":"Matched sequence lengths and dinucleotide-preserving shuffle; fix seeds.","scope_note":"Source documents method or task context; proposed applicability requires protocol review."},"description":"Candidate reference for task-specific evaluation; no score is implied.","facets":{"areas":["genomics"]},"id":"discovery-baseline-dinucleotide-shuffled-sequence-control","kind":"baseline","links":[{"relation":"applicable_to","target_id":"discovery-benchmark-dart-eval"}],"name":"Dinucleotide-shuffled sequence control","source_ids":["src-discovery-kundajelab-dart-eval"],"status":"discovered"}
{"attributes":{"applicability":"proposed","baseline_type":"biological-procedure","missing_metadata":{"implementation_version":"unextracted","protocol_applicability":"unextracted"},"requirements":"Pinned reference genomes, taxonomy and confidence setting.","scope_note":"Source documents method or task context; proposed applicability requires protocol review."},"description":"Candidate reference for task-specific evaluation; no score is implied.","facets":{"areas":["microbiome"]},"id":"discovery-baseline-exact-sequence-taxonomic-classification","kind":"baseline","links":[{"relation":"applicable_to","target_id":"discovery-benchmark-cami-taxonomic-binning"},{"relation":"model","target_id":"discovery-model-kraken-2"}],"name":"Exact-sequence taxonomic classification","source_ids":["src-discovery-derrickwood-kraken2"],"status":"discovered"}
{"attributes":{"applicability":"proposed","baseline_type":"experimental-reference","missing_metadata":{"implementation_version":"unextracted","protocol_applicability":"unextracted"},"requirements":"Independent biological replicates under matching conditions; not a universal ceiling.","scope_note":"Source documents method or task context; proposed applicability requires protocol review."},"description":"Candidate reference for task-specific evaluation; no score is implied.","facets":{"areas":["single-cell"]},"id":"discovery-baseline-experimental-replicate-agreement","kind":"baseline","links":[{"relation":"applicable_to","target_id":"discovery-benchmark-scperteval"}],"name":"Experimental replicate agreement","source_ids":["src-discovery-virtual-cell-research-community-scperteval"],"status":"discovered"}
{"attributes":{"applicability":"proposed","baseline_type":"mechanistic","missing_metadata":{"implementation_version":"unextracted","protocol_applicability":"unextracted"},"requirements":"Stoichiometric reconstruction, growth medium, bounds and objective.","scope_note":"Source documents method or task context; proposed applicability requires protocol review."},"description":"Candidate reference for task-specific evaluation; no score is implied.","facets":{"areas":["mechanistic-biology"]},"id":"discovery-baseline-flux-balance-prediction","kind":"baseline","links":[{"relation":"model","target_id":"discovery-model-cobrapy"}],"name":"Flux-balance prediction","source_ids":["src-discovery-opencobra-cobrapy"],"status":"discovered"}
{"attributes":{"applicability":"proposed","baseline_type":"simple-statistical","missing_metadata":{"implementation_version":"unextracted","protocol_applicability":"unextracted"},"requirements":"Training-only motif features and leakage-aware glycan split.","scope_note":"Source documents method or task context; proposed applicability requires protocol review."},"description":"Candidate reference for task-specific evaluation; no score is implied.","facets":{"areas":["glycomics"]},"id":"discovery-baseline-glycan-motif-feature-classifier","kind":"baseline","links":[{"relation":"applicable_to","target_id":"discovery-benchmark-glycanml"}],"name":"Glycan motif feature classifier","source_ids":["src-discovery-bojarlab-glycowork"],"status":"discovered"}
{"attributes":{"applicability":"proposed","baseline_type":"simple-statistical","missing_metadata":{"implementation_version":"unextracted","protocol_applicability":"unextracted"},"requirements":"Training-only feature fitting; choose k and penalty within training folds.","scope_note":"Source documents method or task context; proposed applicability requires protocol review."},"description":"Candidate reference for task-specific evaluation; no score is implied.","facets":{"areas":["genomics"]},"id":"discovery-baseline-k-mer-ridge-regression","kind":"baseline","links":[{"relation":"applicable_to","target_id":"discovery-benchmark-genomic-benchmarks"}],"name":"k-mer ridge regression","source_ids":["src-discovery-ml-bioinfo-ceitec-genomic-benchmarks"],"status":"discovered"}
{"attributes":{"applicability":"proposed","baseline_type":"biological-procedure","missing_metadata":{"implementation_version":"unextracted","protocol_applicability":"unextracted"},"requirements":"Adduct, ion mode, library version, mass tolerance and annotation resolution.","scope_note":"Source documents method or task context; proposed applicability requires protocol review."},"description":"Candidate reference for task-specific evaluation; no score is implied.","facets":{"areas":["lipidomics"]},"id":"discovery-baseline-lipid-fragmentation-library-match","kind":"baseline","links":[{"relation":"model","target_id":"discovery-model-lipidblast"}],"name":"Lipid fragmentation library match","source_ids":["src-discovery-lipidblast"],"status":"discovered"}
{"attributes":{"applicability":"proposed","baseline_type":"biological-procedure","missing_metadata":{"implementation_version":"unextracted","protocol_applicability":"unextracted"},"requirements":"Pinned marker database and taxonomic rank.","scope_note":"Source documents method or task context; proposed applicability requires protocol review."},"description":"Candidate reference for task-specific evaluation; no score is implied.","facets":{"areas":["microbiome"]},"id":"discovery-baseline-marker-based-microbial-profiling","kind":"baseline","links":[{"relation":"applicable_to","target_id":"discovery-benchmark-cami-taxonomic-profiling"},{"relation":"model","target_id":"discovery-model-metaphlan"}],"name":"Marker-based microbial profiling","source_ids":["src-discovery-biobakery-metaphlan"],"status":"discovered"}
{"attributes":{"applicability":"proposed","baseline_type":"biological-procedure","missing_metadata":{"implementation_version":"unextracted","protocol_applicability":"unextracted"},"requirements":"Peak filtering, precursor tolerance, library and candidate set.","scope_note":"Source documents method or task context; proposed applicability requires protocol review."},"description":"Candidate reference for task-specific evaluation; no score is implied.","facets":{"areas":["metabolomics"]},"id":"discovery-baseline-mass-spectral-cosine-matching","kind":"baseline","links":[{"relation":"applicable_to","target_id":"discovery-benchmark-massspecgym-molecule-retrieval"},{"relation":"model","target_id":"discovery-model-matchms"}],"name":"Mass spectral cosine matching","source_ids":["src-discovery-matchms-matchms"],"status":"discovered"}
{"attributes":{"applicability":"proposed","baseline_type":"biological-procedure","missing_metadata":{"implementation_version":"unextracted","protocol_applicability":"unextracted"},"requirements":"Pinned motif library, background frequencies and strand convention.","scope_note":"Source documents method or task context; proposed applicability requires protocol review."},"description":"Candidate reference for task-specific evaluation; no score is implied.","facets":{"areas":["genomics"]},"id":"discovery-baseline-motif-scanning","kind":"baseline","links":[{"relation":"applicable_to","target_id":"discovery-benchmark-dart-eval"},{"relation":"model","target_id":"discovery-model-fimo"}],"name":"Motif scanning","source_ids":["src-discovery-meme"],"status":"discovered"}
{"attributes":{"applicability":"source_supported","baseline_type":"simple-statistical","missing_metadata":{"implementation_version":"unextracted","protocol_applicability":"unextracted"},"requirements":"Task-specific training split and predictor head.","scope_note":"One Hot is an explicitly reported comparator in official TAPE task tables. This record does not imply the same baseline protocol suits every protein task."},"description":"Candidate reference for task-specific evaluation; no score is implied.","facets":{"areas":["protein-function"]},"id":"discovery-baseline-one-hot-protein-encoding","kind":"baseline","links":[{"relation":"applicable_to","target_id":"discovery-benchmark-tape"}],"name":"One-hot protein encoding","source_ids":["src-discovery-songlab-cal-tape"],"status":"discovered"}
{"attributes":{"applicability":"proposed","baseline_type":"biological-procedure","missing_metadata":{"implementation_version":"unextracted","protocol_applicability":"unextracted"},"requirements":"Pinned sequence database, MSA construction and score threshold.","scope_note":"Source documents method or task context; proposed applicability requires protocol review."},"description":"Candidate reference for task-specific evaluation; no score is implied.","facets":{"areas":["protein-function"]},"id":"discovery-baseline-profile-hmm-sequence-search","kind":"baseline","links":[{"relation":"applicable_to","target_id":"discovery-benchmark-tape-remote-homology-detection"},{"relation":"model","target_id":"discovery-model-hh-suite"}],"name":"Profile-HMM sequence search","source_ids":["src-discovery-soedinglab-hh-suite"],"status":"discovered"}
{"attributes":{"applicability":"proposed","baseline_type":"biological-procedure","missing_metadata":{"implementation_version":"unextracted","protocol_applicability":"unextracted"},"requirements":"Training reference set and homology leakage controls.","scope_note":"Source documents method or task context; proposed applicability requires protocol review."},"description":"Candidate reference for task-specific evaluation; no score is implied.","facets":{"areas":["protein-function"]},"id":"discovery-baseline-protein-homology-transfer","kind":"baseline","links":[{"relation":"applicable_to","target_id":"discovery-benchmark-cafa"},{"relation":"model","target_id":"discovery-model-mmseqs2"}],"name":"Protein homology transfer","source_ids":["src-discovery-soedinglab-mmseqs2"],"status":"discovered"}
{"attributes":{"applicability":"proposed","baseline_type":"learned-specialist","missing_metadata":{"implementation_version":"unextracted","protocol_applicability":"unextracted"},"requirements":"Pinned backbone, checkpoint and sampling temperature.","scope_note":"Source documents method or task context; proposed applicability requires protocol review."},"description":"Candidate reference for task-specific evaluation; no score is implied.","facets":{"areas":["protein-structure"]},"id":"discovery-baseline-protein-sequence-recovery-specialist","kind":"baseline","links":[{"relation":"applicable_to","target_id":"discovery-benchmark-proteinbench"},{"relation":"model","target_id":"discovery-model-proteinmpnn"}],"name":"Protein sequence recovery specialist","source_ids":["src-discovery-dauparas-proteinmpnn"],"status":"discovered"}
{"attributes":{"applicability":"proposed","baseline_type":"null-control","missing_metadata":{"implementation_version":"unextracted","protocol_applicability":"unextracted"},"requirements":"Matched candidate edge universe, edge density and seed.","scope_note":"Source documents method or task context; proposed applicability requires protocol review."},"description":"Candidate reference for task-specific evaluation; no score is implied.","facets":{"areas":["biological-networks"]},"id":"discovery-baseline-random-regulatory-network","kind":"baseline","links":[{"relation":"applicable_to","target_id":"discovery-benchmark-beeline"}],"name":"Random regulatory network","source_ids":["src-discovery-murali-group-beeline"],"status":"discovered"}
{"attributes":{"applicability":"proposed","baseline_type":"biological-procedure","missing_metadata":{"implementation_version":"unextracted","protocol_applicability":"unextracted"},"requirements":"Temperature, thermodynamic parameter set and pseudoknot policy.","scope_note":"Source documents method or task context; proposed applicability requires protocol review."},"description":"Candidate reference for task-specific evaluation; no score is implied.","facets":{"areas":["rna"]},"id":"discovery-baseline-rna-minimum-free-energy-folding","kind":"baseline","links":[{"relation":"applicable_to","target_id":"discovery-benchmark-beacon"},{"relation":"model","target_id":"discovery-model-viennarna-rnafold"}],"name":"RNA minimum-free-energy folding","source_ids":["src-discovery-viennarna-viennarna"],"status":"discovered"}
{"attributes":{"applicability":"proposed","baseline_type":"simple-statistical","missing_metadata":{"implementation_version":"unextracted","protocol_applicability":"unextracted"},"requirements":"Training-only GC/codon composition features and held-out split.","scope_note":"Source documents method or task context; proposed applicability requires protocol review."},"description":"Candidate reference for task-specific evaluation; no score is implied.","facets":{"areas":["rna"]},"id":"discovery-baseline-rna-sequence-composition-regression","kind":"baseline","links":[{"relation":"applicable_to","target_id":"discovery-benchmark-mrnabench"}],"name":"RNA sequence-composition regression","source_ids":["src-discovery-morrislab-mrnabench"],"status":"discovered"}
{"attributes":{"applicability":"proposed","baseline_type":"simple-statistical","missing_metadata":{"implementation_version":"unextracted","protocol_applicability":"unextracted"},"requirements":"Frozen features and donor-disjoint folds; training-only regularization.","scope_note":"Source documents method or task context; proposed applicability requires protocol review."},"description":"Candidate reference for task-specific evaluation; no score is implied.","facets":{"areas":["spatial-omics"]},"id":"discovery-baseline-spatial-expression-ridge-regression","kind":"baseline","links":[{"relation":"applicable_to","target_id":"discovery-benchmark-hest-benchmark"}],"name":"Spatial expression ridge regression","source_ids":["src-discovery-mahmoodlab-hest"],"status":"discovered"}
{"attributes":{"applicability":"proposed","baseline_type":"learned-specialist","missing_metadata":{"implementation_version":"unextracted","protocol_applicability":"unextracted"},"requirements":"Genome assembly, transcript context and model release.","scope_note":"Source documents method or task context; proposed applicability requires protocol review."},"description":"Candidate reference for task-specific evaluation; no score is implied.","facets":{"areas":["genomics"]},"id":"discovery-baseline-specialist-splicing-predictor","kind":"baseline","links":[{"relation":"model","target_id":"discovery-model-spliceai"}],"name":"Specialist splicing predictor","source_ids":["src-discovery-illumina-spliceai"],"status":"discovered"}
{"attributes":{"applicability":"proposed","baseline_type":"simple-statistical","missing_metadata":{"implementation_version":"unextracted","protocol_applicability":"unextracted"},"requirements":"Training-only expression mean with explicit perturbation averaging.","scope_note":"Source documents method or task context; proposed applicability requires protocol review."},"description":"Candidate reference for task-specific evaluation; no score is implied.","facets":{"areas":["single-cell"]},"id":"discovery-baseline-training-perturbation-mean","kind":"baseline","links":[{"relation":"applicable_to","target_id":"discovery-benchmark-perturbench"}],"name":"Training perturbation mean","source_ids":["src-discovery-altoslabs-perturbench"],"status":"discovered"}
{"attributes":{"applicability":"proposed","baseline_type":"learned-specialist","missing_metadata":{"implementation_version":"unextracted","protocol_applicability":"unextracted"},"requirements":"Expression normalization, regulator list and training cells.","scope_note":"Source documents method or task context; proposed applicability requires protocol review."},"description":"Candidate reference for task-specific evaluation; no score is implied.","facets":{"areas":["biological-networks"]},"id":"discovery-baseline-tree-ensemble-regulatory-inference","kind":"baseline","links":[{"relation":"applicable_to","target_id":"discovery-benchmark-beeline"},{"relation":"model","target_id":"discovery-model-genie3"}],"name":"Tree-ensemble regulatory inference","source_ids":["src-discovery-aertslab-genie3"],"status":"discovered"}
{"attributes":{"applicability":"proposed","baseline_type":"null-control","missing_metadata":{"implementation_version":"unextracted","protocol_applicability":"unextracted"},"requirements":"Same cells, preprocessing and metrics as integrated embeddings.","scope_note":"Source documents method or task context; proposed applicability requires protocol review."},"description":"Candidate reference for task-specific evaluation; no score is implied.","facets":{"areas":["single-cell"]},"id":"discovery-baseline-unintegrated-expression-reference","kind":"baseline","links":[{"relation":"applicable_to","target_id":"discovery-benchmark-scib"}],"name":"Unintegrated expression reference","source_ids":["src-discovery-theislab-scib"],"status":"discovered"}
{"attributes":{"entity_level":"evaluator","scope_note":"Specialist molecular or omics evaluation; protocol details require review before numerical comparison.","task":"Genome reconstruction and taxonomic assignment evaluation","version":null,"historical_missing_metadata":{"dataset_release":"unextracted","metric_implementation":"unextracted","split_manifest":"unextracted","version":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"AMBER assesses metagenomic genome bins and taxonomic assignments against gold-standard assignments.","summary_source_ids":["src-discovery-cami-challenge-amber"],"summary_source_locator":"Pinned README: introduction; Metrics computed per bin/per sample; input formats","sections":[{"title":"Evaluation methodology","body":"User-supplied bins plus sample-matched gold-standard assignments; example CAMI datasets are linked. The evaluator scores supplied outputs; predictor train/test partitions belong to the dataset/run being evaluated. Bin purity/completeness; sample accuracy, contamination, adjusted Rand index, binned fraction and recovered-genome counts; UniFrac for taxonomic binning. Multiple programs or parameter settings can be compared using the same reference. Uncertainty across samples, datasets or training runs must be defined by the evaluation study; this evaluator entry does not fix one experiment.","source_ids":["src-discovery-cami-challenge-amber"],"source_locator":"Pinned README: introduction; Metrics computed per bin/per sample; input formats"}],"facts":[{"label":"Datasets","value":"User-supplied bins plus sample-matched gold-standard assignments; example CAMI datasets are linked.","status":"source_checked","source_ids":["src-discovery-cami-challenge-amber"],"source_locator":"Pinned README: introduction; Metrics computed per bin/per sample; input formats"},{"label":"Splits","value":"The evaluator scores supplied outputs; predictor train/test partitions belong to the dataset/run being evaluated.","status":"inapplicable","source_ids":["src-discovery-cami-challenge-amber"],"source_locator":"Pinned README: introduction; Metrics computed per bin/per sample; input formats"},{"label":"Metrics","value":"Bin purity/completeness; sample accuracy, contamination, adjusted Rand index, binned fraction and recovered-genome counts; UniFrac for taxonomic binning.","status":"source_checked","source_ids":["src-discovery-cami-challenge-amber"],"source_locator":"Pinned README: introduction; Metrics computed per bin/per sample; input formats"},{"label":"Baselines","value":"Multiple programs or parameter settings can be compared using the same reference.","status":"source_checked","source_ids":["src-discovery-cami-challenge-amber"],"source_locator":"Pinned README: introduction; Metrics computed per bin/per sample; input formats"},{"label":"Leakage controls","value":"AMBER compares submitted bin assignments with a user-provided gold standard. It does not construct model-training partitions or certify reference-database independence; those controls belong to the evaluated study.","status":"inapplicable","source_ids":["evidence-discovery-final-amber"],"source_locator":"Implementation and benchmarking: gold-standard mapping, input formats and metrics"},{"label":"Uncertainty","value":"Uncertainty across samples, datasets or training runs must be defined by the evaluation study; this evaluator entry does not fix one experiment.","status":"inapplicable","source_ids":["src-discovery-cami-challenge-amber"],"source_locator":"Pinned README: introduction; Metrics computed per bin/per sample; input formats"},{"label":"Entity type","value":"Evaluator for genome and taxonomic binning.","status":"source_checked","source_ids":["src-discovery-cami-challenge-amber"],"source_locator":"Pinned README: introduction; Metrics computed per bin/per sample; input formats"},{"label":"Organisms","value":"The evaluator accepts any community with compatible gold-standard assignments; organism scope belongs to the input dataset.","status":"inapplicable","source_ids":["src-discovery-cami-challenge-amber"],"source_locator":"Pinned README: introduction; Metrics computed per bin/per sample; input formats"},{"label":"Assays","value":"Metagenomic sequence bins and reference assignments.","status":"source_checked","source_ids":["src-discovery-cami-challenge-amber"],"source_locator":"Pinned README: introduction; Metrics computed per bin/per sample; input formats"},{"label":"Allowed inputs","value":"Predicted sequence-to-bin assignments and sample-matched gold standards.","status":"source_checked","source_ids":["src-discovery-cami-challenge-amber"],"source_locator":"Pinned README: introduction; Metrics computed per bin/per sample; input formats"},{"label":"Adaptation","value":"AMBER scores submitted assignments; it does not prescribe predictor training.","status":"inapplicable","source_ids":["src-discovery-cami-challenge-amber"],"source_locator":"Pinned README: introduction; Metrics computed per bin/per sample; input formats"}],"strengths":[{"text":"Reports purity and completeness separately, revealing over-splitting versus contamination.","source_ids":["src-discovery-cami-challenge-amber"],"source_locator":"Pinned README: introduction; Metrics computed per bin/per sample; input formats"}],"limitations":[{"text":"Results depend on the supplied gold-standard assignments and taxonomy version. AMBER evaluation alone does not establish independence of the predictor from those references.","source_ids":["evidence-discovery-final-amber"],"source_locator":"Implementation and benchmarking: gold-standard mapping, input formats and metrics"}],"diagram":{"title":"Evaluation procedure","steps":["Allowed inputs: Predicted sequence-to-bin assignments and sample-matched gold standards.","Splits: The evaluator scores supplied outputs; predictor train/test partitions belong to the dataset/run being evaluated.","Metrics: Bin purity/completeness; sample accuracy, contamination, adjusted Rand index, binned fraction and recovered-genome counts; UniFrac for taxonomic binning."],"caption":"Conceptual procedure. Task variants and protocol versions retain their separate scoring conditions.","source_ids":["src-discovery-cami-challenge-amber"],"source_locator":"Pinned README: introduction; Metrics computed per bin/per sample; input formats"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction."}}},"description":"Genome reconstruction and taxonomic assignment evaluation","facets":{"areas":["microbiome"]},"id":"discovery-benchmark-amber","kind":"benchmark","links":[],"name":"AMBER","source_ids":["src-discovery-cami-challenge-amber"],"status":"discovered"}
{"attributes":{"entity_level":"suite","scope_note":"Specialist molecular or omics evaluation; protocol details require review before numerical comparison.","task":"Three-dimensional molecular learning tasks","version":null,"historical_missing_metadata":{"dataset_release":"unextracted","metric_implementation":"unextracted","split_manifest":"unextracted","version":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"ATOM3D provides molecular-structure datasets and utilities for task-specific evaluation.","summary_source_ids":["src-discovery-drorlab-atom3d"],"summary_source_locator":"Pinned README: Overview; dataset access; supported formats and splitting/filtering utilities","sections":[{"title":"Evaluation methodology","body":"ATOM3D evaluates predictions from molecular structures using eight separate tasks. It supplies task-specific reference labels and partitions, ranging from random small molecules to held-out protein families and future structure-prediction targets. Each task has its own metric and representation-matched baseline; the suite is not a single universal structure score.","source_ids":["evidence-discovery-final-atom3d"],"source_locator":"Sections 3.1–3.8, 4–5; Appendix D–F; Table 8"}],"facts":[{"label":"Datasets","value":"Three-dimensional molecular data with associated labels/metadata; ligand-binding affinity is one documented example.","status":"source_checked","source_ids":["src-discovery-drorlab-atom3d"],"source_locator":"Pinned README: Overview; dataset access; supported formats and splitting/filtering utilities"},{"label":"Splits","value":"Task-specific: random molecules for SMP; 30% protein sequence identity for PIP, MSP and the strict LBA split; CATH topology groups for RES; protein targets for LEP; competition years for PSR and RSR. LBA also provides a less restrictive 60% identity split.","status":"source_checked","source_ids":["evidence-discovery-final-atom3d"],"source_locator":"Sections 3.1–3.8, 4–5; Appendix D–F; Table 8"},{"label":"Metrics","value":"SMP: MAE; PIP/MSP/LEP: classification AUROC; RES: accuracy; LBA: RMSE and correlations; PSR/RSR: correlations of predicted structure quality with reference GDT_TS/RMSD. Metrics and aggregation belong to each task.","status":"source_checked","source_ids":["evidence-discovery-final-atom3d"],"source_locator":"Sections 3.1–3.8, 4–5; Appendix D–F; Table 8"},{"label":"Baselines","value":"3D CNNs, graph networks and equivariant networks are compared with task-specific 1D/2D methods; structure-ranking tasks also use established 3D methods. Appendix F documents the per-task comparators.","status":"source_checked","source_ids":["evidence-discovery-final-atom3d"],"source_locator":"Sections 3.1–3.8, 4–5; Appendix D–F; Table 8"},{"label":"Leakage controls","value":"Protein-sequence, topology, target and temporal partitions reduce task-specific overlap; PIP prunes DIPS proteins against DB5. SMP uses a random molecular split, so it is not a scaffold-held-out test.","status":"source_checked","source_ids":["evidence-discovery-final-atom3d"],"source_locator":"Sections 3.1–3.8, 4–5; Appendix D–F; Table 8"},{"label":"Uncertainty","value":"Table 8 reports standard deviations over three replicates. The paper distinguishes this run variation from the choice of molecular dataset and split.","status":"source_checked","source_ids":["evidence-discovery-final-atom3d"],"source_locator":"Sections 3.1–3.8, 4–5; Appendix D–F; Table 8"},{"label":"Entity type","value":"Molecular dataset and evaluation suite.","status":"source_checked","source_ids":["src-discovery-drorlab-atom3d"],"source_locator":"Pinned README: Overview; dataset access; supported formats and splitting/filtering utilities"},{"label":"Organisms","value":"No single organism defines this suite of molecular structure datasets.","status":"inapplicable","source_ids":["src-discovery-drorlab-atom3d"],"source_locator":"Pinned README: Overview; dataset access; supported formats and splitting/filtering utilities"},{"label":"Assays","value":"Task-specific structural and molecular-property labels.","status":"source_checked","source_ids":["src-discovery-drorlab-atom3d"],"source_locator":"Pinned README: Overview; dataset access; supported formats and splitting/filtering utilities"},{"label":"Allowed inputs","value":"Three-dimensional molecular coordinates and task labels; supported formats include PDB, SDF and XYZ.","status":"source_checked","source_ids":["src-discovery-drorlab-atom3d"],"source_locator":"Pinned README: Overview; dataset access; supported formats and splitting/filtering utilities"},{"label":"Adaptation","value":"Supervised task evaluation; task-specific training configurations are linked separately.","status":"source_checked","source_ids":["src-discovery-drorlab-atom3d"],"source_locator":"Pinned README: Overview; dataset access; supported formats and splitting/filtering utilities"}],"strengths":[{"text":"Common data loaders allow different 3D representations to use the same dataset definitions.","source_ids":["src-discovery-drorlab-atom3d"],"source_locator":"Pinned README: Overview; dataset access; supported formats and splitting/filtering utilities"}],"limitations":[{"text":"Comparisons must retain the task, split and available structural information. In particular, the two LBA identity thresholds test different generalization regimes.","source_ids":["evidence-discovery-final-atom3d"],"source_locator":"Sections 3.1–3.8, 4–5; Appendix D–F; Table 8"}],"diagram":{"title":"Evaluation procedure","steps":["Allowed inputs: Three-dimensional molecular coordinates and task labels; supported formats include PDB, SDF and XYZ.","Splits: Task-specific: random molecules for SMP; 30% protein sequence identity for PIP, MSP and the strict LBA split; CATH topology groups for RES; protein targets for LEP; competition years for PSR and RSR. LBA also provides a less restrictive 60% identity split.","Metrics: SMP: MAE; PIP/MSP/LEP: classification AUROC; RES: accuracy; LBA: RMSE and correlations; PSR/RSR: correlations of predicted structure quality with reference GDT_TS/RMSD. Metrics and aggregation belong to each task."],"caption":"Conceptual procedure. Task variants and protocol versions retain their separate scoring conditions.","source_ids":["src-discovery-drorlab-atom3d","evidence-discovery-final-atom3d"],"source_locator":"Pinned README: Overview; dataset access; supported formats and splitting/filtering utilities; Sections 3.1–3.8, 4–5; Appendix D–F; Table 8"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction."}}},"description":"Three-dimensional molecular learning tasks","facets":{"areas":["molecular-interactions"]},"id":"discovery-benchmark-atom3d","kind":"benchmark","links":[],"name":"ATOM3D","source_ids":["src-discovery-drorlab-atom3d"],"status":"discovered"}
{"attributes":{"entity_level":"suite","scope_note":"Specialist molecular or omics evaluation; protocol details require review before numerical comparison.","task":"RNA structure, function and engineering tasks","version":null,"historical_missing_metadata":{"dataset_release":"unextracted","metric_implementation":"unextracted","split_manifest":"unextracted","version":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"BEACON compares RNA representations across structural and functional downstream tasks.","summary_source_ids":["src-discovery-terry-r123-rnabenchmark"],"summary_source_locator":"Pinned README: Dataset; task list; Models and Model settings","sections":[{"title":"Evaluation methodology","body":"BEACON tests RNA structure, function and engineering with 13 separately labelled datasets. Models predict nucleotide-level labels, pairwise structural maps or sequence-level properties. Its supplied task partitions and metrics must be preserved, and results are reported over three training seeds.","source_ids":["evidence-discovery-final-beacon"],"source_locator":"Sections 3.1–3.3, 4 and 5.1; Table 1; Appendix A"}],"facts":[{"label":"Datasets","value":"Tasks include secondary structure, contact/distance maps, RNA-family classification, modification and expression-related outcomes.","status":"source_checked","source_ids":["src-discovery-terry-r123-rnabenchmark"],"source_locator":"Pinned README: Dataset; task list; Models and Model settings"},{"label":"Splits","value":"Table 1 publishes separate training, validation and test sizes for all 13 tasks; these reuse different source datasets and are not one shared RNA partition. Structure-map tasks share their 188/23/80 partition, while other tasks use their own supplied folds.","status":"source_checked","source_ids":["evidence-discovery-final-beacon"],"source_locator":"Sections 3.1–3.3, 4 and 5.1; Table 1; Appendix A"},{"label":"Metrics","value":"F1 for secondary structure; top-L precision for contacts; R² for distance maps, imputation, APA, ribosome loading and switches; top-k accuracy for splice sites; accuracy for ncRNA class; AUC for modification; MCRMSE for degradation; weighted Spearman correlation for CRISPR tasks.","status":"source_checked","source_ids":["evidence-discovery-final-beacon"],"source_locator":"Sections 3.1–3.3, 4 and 5.1; Table 1; Appendix A"},{"label":"Baselines","value":"RNA-FM, RNABERT, RNA-MSM, SpliceBERT, UTR-LM, UTRBERT and BEACON variants are listed.","status":"source_checked","source_ids":["src-discovery-terry-r123-rnabenchmark"],"source_locator":"Pinned README: Dataset; task list; Models and Model settings"},{"label":"Leakage controls","value":"The paper specifies source datasets and per-task partitions, but does not establish one suite-wide homology or RNA-family exclusion rule. A train/test size table alone does not demonstrate independence from model pretraining.","status":"unreported","source_ids":["evidence-discovery-final-beacon"],"source_locator":"Sections 3.1–3.3, 4 and 5.1; Table 1; Appendix A"},{"label":"Uncertainty","value":"Experiments are repeated with three random seeds; Section 5.1 reports their mean and sample standard deviation.","status":"source_checked","source_ids":["evidence-discovery-final-beacon"],"source_locator":"Sections 3.1–3.3, 4 and 5.1; Table 1; Appendix A"},{"label":"Entity type","value":"RNA model benchmark suite (BEACON).","status":"source_checked","source_ids":["src-discovery-terry-r123-rnabenchmark"],"source_locator":"Pinned README: Dataset; task list; Models and Model settings"},{"label":"Organisms","value":"Task dependent: human HEK293 icSHAPE data and human splice/UTR assays coexist with RNA structure collections and synthetic constructs. The programmable-switch dataset includes sequences from viral genomes and human transcription factors; the suite is not a single-organism assay.","status":"source_checked","source_ids":["evidence-discovery-final-beacon"],"source_locator":"Sections 3.1–3.3, 4 and 5.1; Table 1; Appendix A"},{"label":"Assays","value":"Secondary structure, contact/distance, family, modification and expression-related labels.","status":"source_checked","source_ids":["src-discovery-terry-r123-rnabenchmark"],"source_locator":"Pinned README: Dataset; task list; Models and Model settings"},{"label":"Allowed inputs","value":"RNA sequences; task-specific targets are supplied in the benchmark datasets.","status":"source_checked","source_ids":["src-discovery-terry-r123-rnabenchmark"],"source_locator":"Pinned README: Dataset; task list; Models and Model settings"},{"label":"Adaptation","value":"Task-specific supervised evaluation using the supplied training scripts/configurations.","status":"source_checked","source_ids":["src-discovery-terry-r123-rnabenchmark"],"source_locator":"Pinned README: Dataset; task list; Models and Model settings"}],"strengths":[{"text":"Multiple RNA endpoints test distinct representation properties rather than one aggregate biological claim.","source_ids":["src-discovery-terry-r123-rnabenchmark"],"source_locator":"Pinned README: Dataset; task list; Models and Model settings"}],"limitations":[{"text":"BEACON combines heterogeneous source assays. Its published task partitions do not establish a common pretraining-overlap audit or a single family-held-out generalization test.","source_ids":["evidence-discovery-final-beacon"],"source_locator":"Sections 3.1–3.3, 4 and 5.1; Table 1; Appendix A"}],"diagram":{"title":"Evaluation procedure","steps":["Allowed inputs: RNA sequences; task-specific targets are supplied in the benchmark datasets.","Splits: Table 1 publishes separate training, validation and test sizes for all 13 tasks; these reuse different source datasets and are not one shared RNA partition. Structure-map tasks share their 188/23/80 partition, while other tasks use their own supplied folds.","Metrics: F1 for secondary structure; top-L precision for contacts; R² for distance maps, imputation, APA, ribosome loading and switches; top-k accuracy for splice sites; accuracy for ncRNA class; AUC for modification; MCRMSE for degradation; weighted Spearman correlation for CRISPR tasks."],"caption":"Conceptual procedure. Task variants and protocol versions retain their separate scoring conditions.","source_ids":["src-discovery-terry-r123-rnabenchmark","evidence-discovery-final-beacon"],"source_locator":"Pinned README: Dataset; task list; Models and Model settings; Sections 3.1–3.3, 4 and 5.1; Table 1; Appendix A"},"coverage":"limited","gaps":["Leakage controls: The paper specifies source datasets and per-task partitions, but does not establish one suite-wide homology or RNA-family exclusion rule. A train/test size table alone does not demonstrate independence from model pretraining."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction."}}},"description":"RNA structure, function and engineering tasks","facets":{"areas":["rna"]},"id":"discovery-benchmark-beacon","kind":"benchmark","links":[],"name":"BEACON","source_ids":["src-discovery-terry-r123-rnabenchmark"],"status":"discovered"}
{"attributes":{"entity_level":"suite","scope_note":"Specialist molecular or omics evaluation; protocol details require review before numerical comparison.","task":"Gene regulatory network inference","version":null,"historical_missing_metadata":{"dataset_release":"unextracted","metric_implementation":"unextracted","split_manifest":"unextracted","version":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"BEELINE compares inferred gene-regulatory edge rankings with reference networks.","summary_source_ids":["src-discovery-murali-group-beeline"],"summary_source_locator":"Pinned README: Overview; Running algorithms; Evaluate results metric table; Plot results","sections":[{"title":"Evaluation methodology","body":"BEELINE evaluates gene regulatory network inference from single-cell expression. It runs inference algorithms on synthetic, curated-model and experimental datasets, then compares ranked predicted edges with known or constructed reference networks. Pseudotime requirements and reference-network reliability are part of each evaluation.","source_ids":["evidence-discovery-final-beeline"],"source_locator":"Methods: algorithm execution, simulated/curated datasets and experimental datasets"}],"facts":[{"label":"Datasets","value":"Single-cell expression inputs and ground-truth regulatory networks, configured per dataset.","status":"source_checked","source_ids":["src-discovery-murali-group-beeline"],"source_locator":"Pinned README: Overview; Running algorithms; Evaluate results metric table; Plot results"},{"label":"Splits","value":"BEELINE infers a network from each expression dataset and scores predicted edges against a reference network. Simulated replicates and experimental contexts are separate benchmark cases; the benchmark is not a single supervised train/validation/test classification split.","status":"source_checked","source_ids":["evidence-discovery-final-beeline"],"source_locator":"Methods: algorithm execution, simulated/curated datasets and experimental datasets"},{"label":"Metrics","value":"AUROC, AUPRC, early-precision ratio, signed early precision, rank correlation and top-edge overlap; runtime/network diagnostics are separate outputs.","status":"source_checked","source_ids":["src-discovery-murali-group-beeline"],"source_locator":"Pinned README: Overview; Running algorithms; Evaluate results metric table; Plot results"},{"label":"Baselines","value":"Containerized inference methods share ranked-edge outputs for a common evaluator.","status":"source_checked","source_ids":["src-discovery-murali-group-beeline"],"source_locator":"Pinned README: Overview; Running algorithms; Evaluate results metric table; Plot results"},{"label":"Leakage controls","value":"Reference networks are used to score inferred edges, while inputs are expression and, for some methods, pseudotime. Algorithm assumptions, simulated ground truth and experimental reference construction are explicit; no single held-out-gene training protocol applies to all methods.","status":"source_checked","source_ids":["evidence-discovery-final-beeline"],"source_locator":"Methods: algorithm execution, simulated/curated datasets and experimental datasets"},{"label":"Uncertainty","value":"Plotting supports multiple-run distributions, but a particular repeated-run design is not fixed by the README.","status":"source_checked","source_ids":["src-discovery-murali-group-beeline"],"source_locator":"Pinned README: Overview; Running algorithms; Evaluate results metric table; Plot results"},{"label":"Entity type","value":"Gene-regulatory network inference evaluation pipeline.","status":"source_checked","source_ids":["src-discovery-murali-group-beeline"],"source_locator":"Pinned README: Overview; Running algorithms; Evaluate results metric table; Plot results"},{"label":"Organisms","value":"Human transcription-factor metadata is supported; the chosen single-cell dataset defines organism scope.","status":"source_checked","source_ids":["src-discovery-murali-group-beeline"],"source_locator":"Pinned README: Overview; Running algorithms; Evaluate results metric table; Plot results"},{"label":"Assays","value":"Single-cell expression with a reference regulatory network.","status":"source_checked","source_ids":["src-discovery-murali-group-beeline"],"source_locator":"Pinned README: Overview; Running algorithms; Evaluate results metric table; Plot results"},{"label":"Allowed inputs","value":"Expression matrix, gene metadata and ground-truth edges for evaluation.","status":"source_checked","source_ids":["src-discovery-murali-group-beeline"],"source_locator":"Pinned README: Overview; Running algorithms; Evaluate results metric table; Plot results"},{"label":"Adaptation","value":"Inference algorithms operate on the supplied expression data; labeled reference edges are used for assessment.","status":"source_checked","source_ids":["src-discovery-murali-group-beeline"],"source_locator":"Pinned README: Overview; Running algorithms; Evaluate results metric table; Plot results"}],"strengths":[{"text":"Ranked-edge evaluation exposes precision–recall behavior rather than treating a thresholded graph as ground truth.","source_ids":["src-discovery-murali-group-beeline"],"source_locator":"Pinned README: Overview; Running algorithms; Evaluate results metric table; Plot results"}],"limitations":[{"text":"An inferred association is not automatically a causal regulatory edge. Simulated ground truth, curated biological models and experimentally assembled references support different conclusions.","source_ids":["evidence-discovery-final-beeline"],"source_locator":"Methods: algorithm execution, simulated/curated datasets and experimental datasets"}],"diagram":{"title":"Evaluation procedure","steps":["Allowed inputs: Expression matrix, gene metadata and ground-truth edges for evaluation.","Splits: BEELINE infers a network from each expression dataset and scores predicted edges against a reference network. Simulated replicates and experimental contexts are separate benchmark cases; the benchmark is not a single supervised train/validation/test classification split.","Metrics: AUROC, AUPRC, early-precision ratio, signed early precision, rank correlation and top-edge overlap; runtime/network diagnostics are separate outputs."],"caption":"Conceptual procedure. Task variants and protocol versions retain their separate scoring conditions.","source_ids":["src-discovery-murali-group-beeline","evidence-discovery-final-beeline"],"source_locator":"Pinned README: Overview; Running algorithms; Evaluate results metric table; Plot results; Methods: algorithm execution, simulated/curated datasets and experimental datasets"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction."}}},"description":"Gene regulatory network inference","facets":{"areas":["biological-networks"]},"id":"discovery-benchmark-beeline","kind":"benchmark","links":[],"name":"BEELINE","source_ids":["src-discovery-murali-group-beeline"],"status":"discovered"}
{"attributes":{"entity_level":"suite","scope_note":"Specialist molecular or omics evaluation; protocol details require review before numerical comparison.","task":"DNA representations on biological downstream tasks","version":null,"historical_missing_metadata":{"dataset_release":"unextracted","metric_implementation":"unextracted","split_manifest":"unextracted","version":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"BEND evaluates DNA representations using task-specific genomic annotations and explicit split membership.","summary_source_ids":["src-discovery-frederikkemarin-bend"],"summary_source_locator":"Pinned README: Data format; dataset use; Citation Guidelines","sections":[{"title":"Evaluation methodology","body":"BEND asks whether DNA representations support realistic human-genome annotation across short and long contexts. Frozen embeddings feed a small CNN for supervised tasks, while variant-effect prediction compares reference and alternate embeddings without fitting a task predictor. Whole-chromosome or sequence-identity partitions and specialist baselines make the tested capability explicit.","source_ids":["evidence-discovery-final-bend"],"source_locator":"Sections 3–4, Table 1, Table 3, Appendix A.1 and A.6"}],"facts":[{"label":"Datasets","value":"Genomic-coordinate tables paired with labels, including HDF5 labels for complex outputs; reference genome supplies input sequences.","status":"source_checked","source_ids":["src-discovery-frederikkemarin-bend"],"source_locator":"Pinned README: Data format; dataset use; Citation Guidelines"},{"label":"Splits","value":"BED task files contain an explicit split column, and paired label files share the same row index.","status":"source_checked","source_ids":["src-discovery-frederikkemarin-bend"],"source_locator":"Pinned README: Data format; dataset use; Citation Guidelines"},{"label":"Metrics","value":"Gene finding uses multiclass MCC; enhancer annotation uses AUPRC; chromatin accessibility, histone modification and methylation use label-wise AUROC; variant-effect tasks use AUROC.","status":"source_checked","source_ids":["evidence-discovery-final-bend"],"source_locator":"Sections 3–4, Table 1, Table 3, Appendix A.1 and A.6"},{"label":"Baselines","value":"A one-hot two-layer CNN is matched to frozen-embedding probes. Task specialists include AUGUSTUS, Enformer, Basset and DeepSEA, and the study includes supervised ResNet and simple pretrained language-model controls.","status":"source_checked","source_ids":["evidence-discovery-final-bend"],"source_locator":"Sections 3–4, Table 1, Table 3, Appendix A.1 and A.6"},{"label":"Leakage controls","value":"Supervised tasks hold out chromosomes, except gene finding, whose cross-partition pairs share no more than 80% identity of the mature protein. Enhancer evaluation uses ten chromosome-based folds. Variant effects are zero-shot tests; these partitions do not by themselves remove unsupervised genome-pretraining exposure.","status":"source_checked","source_ids":["evidence-discovery-final-bend"],"source_locator":"Appendix A.1.1 gene-finding split: mature-protein identity; other supervised task and enhancer split descriptions in Appendix A.1"},{"label":"Uncertainty","value":"Enhancer evaluation uses ten-fold cross-validation because the dataset is small. Table 3 does not provide a uniform repeated-seed confidence-interval protocol for all task/model scores; its Enformer ± entry must not be generalized to every row.","status":"unreported","source_ids":["evidence-discovery-final-bend"],"source_locator":"Sections 3–4, Table 1, Table 3, Appendix A.1 and A.6"},{"label":"Entity type","value":"Human genomic task benchmark suite.","status":"source_checked","source_ids":["src-discovery-frederikkemarin-bend"],"source_locator":"Pinned README: Data format; dataset use; Citation Guidelines"},{"label":"Organisms","value":"Human genomic evaluation; the README distinguishes models trained on other organisms from the evaluated set.","status":"source_checked","source_ids":["src-discovery-frederikkemarin-bend"],"source_locator":"Pinned README: Data format; dataset use; Citation Guidelines"},{"label":"Assays","value":"Task-dependent regulatory and annotation datasets, including ENCODE-derived resources.","status":"source_checked","source_ids":["src-discovery-frederikkemarin-bend"],"source_locator":"Pinned README: Data format; dataset use; Citation Guidelines"},{"label":"Allowed inputs","value":"DNA sequences extracted from genomic coordinates and a reference genome; complex labels align by index to HDF5 records.","status":"source_checked","source_ids":["src-discovery-frederikkemarin-bend"],"source_locator":"Pinned README: Data format; dataset use; Citation Guidelines"},{"label":"Adaptation","value":"Downstream supervised models use precomputed embeddings; unsupervised variant-effect scoring is a separate evaluation route.","status":"source_checked","source_ids":["src-discovery-frederikkemarin-bend"],"source_locator":"Pinned README: Data format; dataset use; Citation Guidelines"}],"strengths":[{"text":"Explicit split columns and aligned label indices make task membership inspectable.","source_ids":["src-discovery-frederikkemarin-bend"],"source_locator":"Pinned README: Data format; dataset use; Citation Guidelines"}],"limitations":[{"text":"BEND measures different biological tasks with different metrics. Its zero-shot variant tests, frozen probes and literature specialist results require separate interpretation; pretraining on the reference genome is not equivalent to supervised label leakage.","source_ids":["evidence-discovery-final-bend"],"source_locator":"Sections 3–4, Table 1, Table 3, Appendix A.1 and A.6"}],"diagram":{"title":"Evaluation procedure","steps":["Allowed inputs: DNA sequences extracted from genomic coordinates and a reference genome; complex labels align by index to HDF5 records.","Splits: BED task files contain an explicit split column, and paired label files share the same row index.","Metrics: Gene finding uses multiclass MCC; enhancer annotation uses AUPRC; chromatin accessibility, histone modification and methylation use label-wise AUROC; variant-effect tasks use AUROC."],"caption":"Conceptual procedure. Task variants and protocol versions retain their separate scoring conditions.","source_ids":["src-discovery-frederikkemarin-bend","evidence-discovery-final-bend"],"source_locator":"Pinned README: Data format; dataset use; Citation Guidelines; Sections 3–4, Table 1, Table 3, Appendix A.1 and A.6"},"coverage":"limited","gaps":["Uncertainty: Enhancer evaluation uses ten-fold cross-validation because the dataset is small. Table 3 does not provide a uniform repeated-seed confidence-interval protocol for all task/model scores; its Enformer ± entry must not be generalized to every row."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction. Independent automated spot review corrected interval terminology and sequence-identity scope against the original figure captions and methods."}}},"description":"DNA representations on biological downstream tasks","facets":{"areas":["genomics"]},"id":"discovery-benchmark-bend","kind":"benchmark","links":[],"name":"BEND","source_ids":["src-discovery-frederikkemarin-bend"],"status":"discovered"}
{"attributes":{"entity_level":"challenge","scope_note":"Specialist molecular or omics evaluation; protocol details require review before numerical comparison.","task":"Protein function prediction challenge","version":null,"historical_missing_metadata":{"dataset_release":"unextracted","metric_implementation":"unextracted","split_manifest":"unextracted","version":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"CAFA evaluates prospective protein-function predictions against annotations that become available after prediction submission.","summary_source_ids":["evidence-benchmark-cafa-20260916"],"summary_source_locator":"Official CAFA page: The problem; The solution; challenge timeline","sections":[{"title":"Evaluation methodology","body":"CAFA is a time-delayed protein-function challenge: predictions are submitted before new experimental annotations become available. Evaluation then compares predictions with those newly added labels, separately by ontology and evaluation mode. CAFA3 provides explicit frequency and homology-transfer baselines and protein-level bootstrap intervals.","source_ids":["evidence-discovery-final-cafa3"],"source_locator":"CAFA3 Methods: Protein-centric evaluation; Figures 3–4"}],"facts":[{"label":"Datasets","value":"Protein sequences supplied for a challenge; proteins gaining experimental annotations after the deadline become evaluation targets.","status":"source_checked","source_ids":["evidence-benchmark-cafa-20260916"],"source_locator":"Official CAFA page: The problem; The solution; challenge timeline"},{"label":"Splits","value":"Prospective temporal assessment separates prediction submission from subsequent annotation growth.","status":"source_checked","source_ids":["evidence-benchmark-cafa-20260916"],"source_locator":"Official CAFA page: The problem; The solution; challenge timeline"},{"label":"Metrics","value":"CAFA3 protein-centric evaluation uses Fmax and semantic distance Smin, with prediction coverage reported. Term-centric assays also use AUROC. Ontology, knowledge category and full/partial evaluation mode must accompany the score.","status":"source_checked","source_ids":["evidence-discovery-final-cafa3"],"source_locator":"Methods: Protein-centric and term-centric evaluation; Figures 3–4"},{"label":"Baselines","value":"In CAFA3 protein-centric evaluation, Naïve predicts each term’s training-set frequency and BLAST transfers annotations using the highest matching sequence identity. Term-centric experiments also include expression-based comparators where available.","status":"source_checked","source_ids":["evidence-discovery-final-cafa3"],"source_locator":"CAFA3 Methods: Protein-centric evaluation; Figures 3–4"},{"label":"Leakage controls","value":"Newly acquired experimental annotations are used for later assessment; exact knowledge-cutoff controls depend on the challenge edition.","status":"source_checked","source_ids":["evidence-benchmark-cafa-20260916"],"source_locator":"Official CAFA page: The problem; The solution; challenge timeline"},{"label":"Uncertainty","value":"CAFA3 Figures 3–4 estimate 95% confidence intervals with 10,000 bootstrap samples of benchmark proteins. This is a version-specific evaluation, not a universal rule for every CAFA round.","status":"source_checked","source_ids":["evidence-discovery-final-cafa3"],"source_locator":"CAFA3 Methods: Protein-centric evaluation; Figures 3–4"},{"label":"Entity type","value":"Prospective protein-function prediction challenge.","status":"source_checked","source_ids":["evidence-benchmark-cafa-20260916"],"source_locator":"Official CAFA page: The problem; The solution; challenge timeline"},{"label":"Organisms","value":"Protein targets spanning challenge-selected taxa.","status":"source_checked","source_ids":["evidence-benchmark-cafa-20260916"],"source_locator":"Official CAFA page: The problem; The solution; challenge timeline"},{"label":"Assays","value":"Experimental functional annotations acquired after the prediction deadline.","status":"source_checked","source_ids":["evidence-benchmark-cafa-20260916"],"source_locator":"Official CAFA page: The problem; The solution; challenge timeline"},{"label":"Allowed inputs","value":"Challenge target protein sequences and permitted pre-deadline knowledge.","status":"source_checked","source_ids":["evidence-benchmark-cafa-20260916"],"source_locator":"Official CAFA page: The problem; The solution; challenge timeline"},{"label":"Adaptation","value":"Methods submit predictions before the target proteins gain evaluation annotations.","status":"source_checked","source_ids":["evidence-benchmark-cafa-20260916"],"source_locator":"Official CAFA page: The problem; The solution; challenge timeline"}],"strengths":[{"text":"Time-delayed annotation supplies a prospective assessment boundary.","source_ids":["evidence-benchmark-cafa-20260916"],"source_locator":"Official CAFA page: The problem; The solution; challenge timeline"}],"limitations":[{"text":"Annotation incompleteness, prediction coverage and the challenge’s ontology version affect interpretation. New CAFA rounds may revise the eligible proteins, metrics and uncertainty procedure.","source_ids":["evidence-discovery-final-cafa3"],"source_locator":"CAFA3 Methods: Protein-centric evaluation; Figures 3–4"}],"diagram":{"title":"Evaluation procedure","steps":["Allowed inputs: Challenge target protein sequences and permitted pre-deadline knowledge.","Splits: Prospective temporal assessment separates prediction submission from subsequent annotation growth.","Metrics: CAFA3 protein-centric evaluation uses Fmax and semantic distance Smin, with prediction coverage reported. Term-centric assays also use AUROC. Ontology, knowledge category and full/partial evaluation mode must accompany the score."],"caption":"Conceptual procedure. Task variants and protocol versions retain their separate scoring conditions.","source_ids":["evidence-benchmark-cafa-20260916","evidence-discovery-final-cafa3"],"source_locator":"Official CAFA page: The problem; The solution; challenge timeline; Methods: Protein-centric and term-centric evaluation; Figures 3–4"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction."}}},"description":"Protein function prediction challenge","facets":{"areas":["protein-function"]},"id":"discovery-benchmark-cafa","kind":"benchmark","links":[],"name":"CAFA","source_ids":["src-discovery-cafa"],"status":"discovered"}
{"attributes":{"entity_level":"challenge","scope_note":"Specialist molecular or omics evaluation; protocol details require review before numerical comparison.","task":"Metagenomic assembly, binning and profiling assessment","version":null,"historical_missing_metadata":{"dataset_release":"unextracted","metric_implementation":"unextracted","split_manifest":"unextracted","version":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"CAMI is a community benchmark program for metagenomic computational methods.","summary_source_ids":["evidence-benchmark-cami-snapshot"],"summary_source_locator":"Official CAMI homepage: initiative description; challenges; dataset correction notices; toolkit citations","sections":[{"title":"Evaluation methodology","body":"CAMI is a series of blinded metagenomic software challenges. In CAMI II, participants received simulated short and long reads from defined communities and submitted assemblies, genome bins, taxonomic assignments or abundance profiles. Reference truth was used only for scoring, and software versions, input read types and community conditions were kept distinct.","source_ids":["evidence-discovery-final-cami2"],"source_locator":"Methods: Challenge datasets, Challenge organization, Evaluation metrics; Figures 2–4; Table 1"}],"facts":[{"label":"Datasets","value":"Challenge-specific metagenomic datasets, including a longitudinal human-gut collection in CAMI III.","status":"source_checked","source_ids":["evidence-benchmark-cami-snapshot"],"source_locator":"Official CAMI homepage: initiative description; challenges; dataset correction notices; toolkit citations"},{"label":"Splits","value":"CAMI II supplied public-genome practice datasets with ground truth before its blinded challenge. Challenge datasets were marine, strain-madness and plant-associated communities; these are challenge conditions, not a standard supervised train/validation/test partition.","status":"source_checked","source_ids":["evidence-discovery-final-cami2"],"source_locator":"Methods: Challenge datasets, Challenge organization, Evaluation metrics; Figures 2–4; Table 1"},{"label":"Metrics","value":"Assembly: genome fraction, NGA50, mismatches, misassemblies and strain precision/recall. Genome binning: purity, completeness, ARI and binned fraction. Taxonomic binning: purity, completeness, F1 and accuracy. Profiling: identification, abundance and diversity metrics, including L1, Bray–Curtis and weighted UniFrac.","status":"source_checked","source_ids":["evidence-discovery-final-cami2"],"source_locator":"Methods: Evaluation metrics; Table 1"},{"label":"Baselines","value":"Submitted programs are compared under the same data condition. Gold-standard assemblies and MEGAHIT assemblies separate binning performance from upstream assembly error; published method identities and versions are listed in Table 1.","status":"source_checked","source_ids":["evidence-discovery-final-cami2"],"source_locator":"Methods: Challenge datasets, Challenge organization, Evaluation metrics; Figures 2–4; Table 1"},{"label":"Leakage controls","value":"Challenge genome data and metadata were kept confidential until the challenge ended. Public reference collections dated 8 January 2019 were supplied for reference-based methods. CAMI II also includes public genomes, so novelty is stratified rather than assumed for every organism.","status":"source_checked","source_ids":["evidence-discovery-final-cami2"],"source_locator":"Methods: Challenge datasets, Challenge organization, Evaluation metrics; Figures 2–4; Table 1"},{"label":"Uncertainty","value":"Uncertainty is task-specific: taxonomic binning Figure 3 uses standard errors across bins; taxonomic profiling Figure 4 reports means across samples with standard deviations. These are not a common seed-based interval for every CAMI metric.","status":"source_checked","source_ids":["evidence-discovery-final-cami2"],"source_locator":"Figure 3 and Figure 4 captions: standard error across taxonomic bins versus standard deviation across samples"},{"label":"Entity type","value":"Metagenomic community challenge series.","status":"source_checked","source_ids":["evidence-benchmark-cami-snapshot"],"source_locator":"Official CAMI homepage: initiative description; challenges; dataset correction notices; toolkit citations"},{"label":"Organisms","value":"Microbial communities; CAMI III includes longitudinal human-gut samples.","status":"source_checked","source_ids":["evidence-benchmark-cami-snapshot"],"source_locator":"Official CAMI homepage: initiative description; challenges; dataset correction notices; toolkit citations"},{"label":"Assays","value":"Challenge-specific metagenomic sequence data and reference composition.","status":"source_checked","source_ids":["evidence-benchmark-cami-snapshot"],"source_locator":"Official CAMI homepage: initiative description; challenges; dataset correction notices; toolkit citations"},{"label":"Allowed inputs","value":"Released sequence data and track-specific reference resources.","status":"source_checked","source_ids":["evidence-benchmark-cami-snapshot"],"source_locator":"Official CAMI homepage: initiative description; challenges; dataset correction notices; toolkit citations"},{"label":"Adaptation","value":"Methods process challenge inputs; a challenge edition and track determine resource rules.","status":"source_checked","source_ids":["evidence-benchmark-cami-snapshot"],"source_locator":"Official CAMI homepage: initiative description; challenges; dataset correction notices; toolkit citations"}],"strengths":[{"text":"Multiple tracks distinguish assembly, binning and abundance estimation.","source_ids":["evidence-benchmark-cami-snapshot"],"source_locator":"Official CAMI homepage: initiative description; challenges; dataset correction notices; toolkit citations"}],"limitations":[{"text":"This profile documents CAMI II as a concrete protocol example. Other CAMI rounds may use different genomes, reference databases and metrics; strain diversity and input assembly quality materially change the task.","source_ids":["evidence-discovery-final-cami2"],"source_locator":"Methods: Challenge datasets, Challenge organization, Evaluation metrics; Figures 2–4; Table 1"}],"diagram":{"title":"Evaluation procedure","steps":["Allowed inputs: Released sequence data and track-specific reference resources.","Splits: CAMI II supplied public-genome practice datasets with ground truth before its blinded challenge. Challenge datasets were marine, strain-madness and plant-associated communities; these are challenge conditions, not a standard supervised train/validation/test partition.","Metrics: Assembly: genome fraction, NGA50, mismatches, misassemblies and strain precision/recall. Genome binning: purity, completeness, ARI and binned fraction. Taxonomic binning: purity, completeness, F1 and accuracy. Profiling: identification, abundance and diversity metrics, including L1, Bray–Curtis and weighted UniFrac."],"caption":"Conceptual procedure. Task variants and protocol versions retain their separate scoring conditions.","source_ids":["evidence-benchmark-cami-snapshot","evidence-discovery-final-cami2"],"source_locator":"Official CAMI homepage: initiative description; challenges; dataset correction notices; toolkit citations; Methods: Challenge datasets, Challenge organization, Evaluation metrics; Figures 2–4; Table 1; Methods: Evaluation metrics; Table 1"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction. Independent automated spot review corrected interval terminology and sequence-identity scope against the original figure captions and methods."}}},"description":"Metagenomic assembly, binning and profiling assessment","facets":{"areas":["microbiome"]},"id":"discovery-benchmark-cami","kind":"benchmark","links":[],"name":"CAMI","source_ids":["src-discovery-cami"],"status":"discovered"}
{"attributes":{"entity_level":"task","scope_note":"Specialist molecular or omics evaluation; protocol details require review before numerical comparison.","task":"genome binning","version":null,"historical_missing_metadata":{"dataset_release":"unextracted","metric_implementation":"unextracted","split_manifest":"unextracted","version":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"The CAMI genome-binning task can be understood through its documented assessment tool; this guide does not identify a challenge-specific run.","summary_source_ids":["src-discovery-cami-challenge-amber"],"summary_source_locator":"Pinned README: introduction; Metrics computed per bin/per sample; input formats","sections":[{"title":"Evaluation methodology","body":"CAMI is a series of blinded metagenomic software challenges. In CAMI II, participants received simulated short and long reads from defined communities and submitted assemblies, genome bins, taxonomic assignments or abundance profiles. Reference truth was used only for scoring, and software versions, input read types and community conditions were kept distinct.","source_ids":["evidence-discovery-final-cami2"],"source_locator":"Methods: Challenge datasets, Challenge organization, Evaluation metrics; Figures 2–4; Table 1"}],"facts":[{"label":"Datasets","value":"A selected CAMI challenge dataset and matching gold standard are required; this entry does not fix the edition.","status":"source_checked","source_ids":["src-discovery-cami-challenge-amber"],"source_locator":"Pinned README: introduction; Metrics computed per bin/per sample; input formats"},{"label":"Splits","value":"CAMI II supplied public-genome practice datasets with ground truth before its blinded challenge. Challenge datasets were marine, strain-madness and plant-associated communities; these are challenge conditions, not a standard supervised train/validation/test partition.","status":"source_checked","source_ids":["evidence-discovery-final-cami2"],"source_locator":"Methods: Challenge datasets, Challenge organization, Evaluation metrics; Figures 2–4; Table 1"},{"label":"Metrics","value":"Bin purity/completeness; sample accuracy, contamination, adjusted Rand index, binned fraction and recovered-genome counts; UniFrac for taxonomic binning.","status":"source_checked","source_ids":["src-discovery-cami-challenge-amber"],"source_locator":"Pinned README: introduction; Metrics computed per bin/per sample; input formats"},{"label":"Baselines","value":"Submitted programs are compared under the same data condition. Gold-standard assemblies and MEGAHIT assemblies separate binning performance from upstream assembly error; published method identities and versions are listed in Table 1.","status":"source_checked","source_ids":["evidence-discovery-final-cami2"],"source_locator":"Methods: Challenge datasets, Challenge organization, Evaluation metrics; Figures 2–4; Table 1"},{"label":"Leakage controls","value":"Challenge genome data and metadata were kept confidential until the challenge ended. Public reference collections dated 8 January 2019 were supplied for reference-based methods. CAMI II also includes public genomes, so novelty is stratified rather than assumed for every organism.","status":"source_checked","source_ids":["evidence-discovery-final-cami2"],"source_locator":"Methods: Challenge datasets, Challenge organization, Evaluation metrics; Figures 2–4; Table 1"},{"label":"Uncertainty","value":"Uncertainty is task-specific: taxonomic binning Figure 3 uses standard errors across bins; taxonomic profiling Figure 4 reports means across samples with standard deviations. These are not a common seed-based interval for every CAMI metric.","status":"source_checked","source_ids":["evidence-discovery-final-cami2"],"source_locator":"Figure 3 and Figure 4 captions: standard error across taxonomic bins versus standard deviation across samples"},{"label":"Entity type","value":"Constituent benchmark task: CAMI genome binning","status":"source_checked","source_ids":["src-discovery-cami-challenge-amber"],"source_locator":"Pinned README: introduction; Metrics computed per bin/per sample; input formats"},{"label":"Organisms","value":"AMBER accepts community gold standards; the selected CAMI dataset supplies organism membership.","status":"inapplicable","source_ids":["src-discovery-cami-challenge-amber"],"source_locator":"Pinned README: introduction; Metrics computed per bin/per sample; input formats"},{"label":"Assays","value":"Metagenomic sequence assignments and their gold-standard bins/taxa.","status":"source_checked","source_ids":["src-discovery-cami-challenge-amber"],"source_locator":"Pinned README: introduction; Metrics computed per bin/per sample; input formats"},{"label":"Allowed inputs","value":"Predicted sequence-to-bin or sequence-to-taxon assignments and the matched gold standard.","status":"source_checked","source_ids":["src-discovery-cami-challenge-amber"],"source_locator":"Pinned README: introduction; Metrics computed per bin/per sample; input formats"},{"label":"Adaptation","value":"AMBER assesses assignments; predictor-training conditions are outside the evaluator.","status":"inapplicable","source_ids":["src-discovery-cami-challenge-amber"],"source_locator":"Pinned README: introduction; Metrics computed per bin/per sample; input formats"}],"strengths":[{"text":"Purity and completeness expose different binning errors.","source_ids":["src-discovery-cami-challenge-amber"],"source_locator":"Pinned README: introduction; Metrics computed per bin/per sample; input formats"}],"limitations":[{"text":"This profile documents CAMI II as a concrete protocol example. Other CAMI rounds may use different genomes, reference databases and metrics; strain diversity and input assembly quality materially change the task.","source_ids":["evidence-discovery-final-cami2"],"source_locator":"Methods: Challenge datasets, Challenge organization, Evaluation metrics; Figures 2–4; Table 1"}],"diagram":{"title":"Evaluation procedure","steps":["Allowed inputs: Predicted sequence-to-bin or sequence-to-taxon assignments and the matched gold standard.","Splits: CAMI II supplied public-genome practice datasets with ground truth before its blinded challenge. Challenge datasets were marine, strain-madness and plant-associated communities; these are challenge conditions, not a standard supervised train/validation/test partition.","Metrics: Bin purity/completeness; sample accuracy, contamination, adjusted Rand index, binned fraction and recovered-genome counts; UniFrac for taxonomic binning."],"caption":"Conceptual procedure. Task variants and protocol versions retain their separate scoring conditions.","source_ids":["src-discovery-cami-challenge-amber","evidence-discovery-final-cami2"],"source_locator":"Pinned README: introduction; Metrics computed per bin/per sample; input formats; Methods: Challenge datasets, Challenge organization, Evaluation metrics; Figures 2–4; Table 1"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction. Independent automated spot review corrected interval terminology and sequence-identity scope against the original figure captions and methods."}}},"description":"genome binning","facets":{"areas":["microbiome"]},"id":"discovery-benchmark-cami-genome-binning","kind":"benchmark","links":[{"relation":"parent","target_id":"discovery-benchmark-cami"},{"relation":"part_of","target_id":"discovery-benchmark-cami"}],"name":"CAMI genome binning","source_ids":["src-discovery-cami"],"status":"discovered"}
{"attributes":{"entity_level":"task","scope_note":"Specialist molecular or omics evaluation; protocol details require review before numerical comparison.","task":"metagenome assembly","version":null,"historical_missing_metadata":{"dataset_release":"unextracted","metric_implementation":"unextracted","split_manifest":"unextracted","version":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"The CAMI assembly task concerns reconstructing metagenomic sequence assemblies; the homepage links MetaQUAST evaluation resources.","summary_source_ids":["evidence-benchmark-cami-snapshot"],"summary_source_locator":"Official CAMI homepage: initiative description; challenges; dataset correction notices; toolkit citations","sections":[{"title":"Evaluation methodology","body":"CAMI is a series of blinded metagenomic software challenges. In CAMI II, participants received simulated short and long reads from defined communities and submitted assemblies, genome bins, taxonomic assignments or abundance profiles. Reference truth was used only for scoring, and software versions, input read types and community conditions were kept distinct.","source_ids":["evidence-discovery-final-cami2"],"source_locator":"Methods: Challenge datasets, Challenge organization, Evaluation metrics; Figures 2–4; Table 1"}],"facts":[{"label":"Datasets","value":"CAMI II marine, strain-madness and plant-associated simulated metagenomes provide short-read, long-read and hybrid conditions. Genome truth and gold-standard assemblies are available after the blinded challenge; preserve dataset and sequencing condition.","status":"source_checked","source_ids":["evidence-discovery-final-cami2"],"source_locator":"Methods: Challenge datasets and Challenge organization"},{"label":"Splits","value":"CAMI II supplied public-genome practice datasets with ground truth before its blinded challenge. Challenge datasets were marine, strain-madness and plant-associated communities; these are challenge conditions, not a standard supervised train/validation/test partition.","status":"source_checked","source_ids":["evidence-discovery-final-cami2"],"source_locator":"Methods: Challenge datasets, Challenge organization, Evaluation metrics; Figures 2–4; Table 1"},{"label":"Metrics","value":"MetaQUAST evaluates genome fraction, mismatches per 100 kb, duplication ratio, NGA50 and misassemblies; strain precision and recall additionally measure high-quality strain reconstruction. Undefined per-genome NGA50 is set to zero before the reported genome average.","status":"source_checked","source_ids":["evidence-discovery-final-cami2"],"source_locator":"Methods: Assembly metrics; Figure 1"},{"label":"Baselines","value":"Submitted programs are compared under the same data condition. Gold-standard assemblies and MEGAHIT assemblies separate binning performance from upstream assembly error; published method identities and versions are listed in Table 1.","status":"source_checked","source_ids":["evidence-discovery-final-cami2"],"source_locator":"Methods: Challenge datasets, Challenge organization, Evaluation metrics; Figures 2–4; Table 1"},{"label":"Leakage controls","value":"Challenge genome data and metadata were kept confidential until the challenge ended. Public reference collections dated 8 January 2019 were supplied for reference-based methods. CAMI II also includes public genomes, so novelty is stratified rather than assumed for every organism.","status":"source_checked","source_ids":["evidence-discovery-final-cami2"],"source_locator":"Methods: Challenge datasets, Challenge organization, Evaluation metrics; Figures 2–4; Table 1"},{"label":"Uncertainty","value":"Figure 1 and the assembly-metrics Methods define descriptive per-genome and dataset summaries, but no universal bootstrap or repeated-seed interval for all assembly scores.","status":"unreported","source_ids":["evidence-discovery-final-cami2"],"source_locator":"Methods: Assembly metrics; Figure 1"},{"label":"Entity type","value":"Constituent benchmark task: CAMI metagenome assembly","status":"source_checked","source_ids":["evidence-benchmark-cami-snapshot"],"source_locator":"Official CAMI homepage: initiative description; challenges; dataset correction notices; toolkit citations"},{"label":"Organisms","value":"Microbial communities; CAMI III includes longitudinal human-gut samples.","status":"source_checked","source_ids":["evidence-benchmark-cami-snapshot"],"source_locator":"Official CAMI homepage: initiative description; challenges; dataset correction notices; toolkit citations"},{"label":"Assays","value":"Challenge-specific metagenomic sequence data and reference composition.","status":"source_checked","source_ids":["evidence-benchmark-cami-snapshot"],"source_locator":"Official CAMI homepage: initiative description; challenges; dataset correction notices; toolkit citations"},{"label":"Allowed inputs","value":"Released sequence data and track-specific reference resources.","status":"source_checked","source_ids":["evidence-benchmark-cami-snapshot"],"source_locator":"Official CAMI homepage: initiative description; challenges; dataset correction notices; toolkit citations"},{"label":"Adaptation","value":"Methods process challenge inputs; a challenge edition and track determine resource rules.","status":"source_checked","source_ids":["evidence-benchmark-cami-snapshot"],"source_locator":"Official CAMI homepage: initiative description; challenges; dataset correction notices; toolkit citations"}],"strengths":[{"text":"Multiple tracks distinguish assembly, binning and abundance estimation.","source_ids":["evidence-benchmark-cami-snapshot"],"source_locator":"Official CAMI homepage: initiative description; challenges; dataset correction notices; toolkit citations"}],"limitations":[{"text":"This profile documents CAMI II as a concrete protocol example. Other CAMI rounds may use different genomes, reference databases and metrics; strain diversity and input assembly quality materially change the task.","source_ids":["evidence-discovery-final-cami2"],"source_locator":"Methods: Challenge datasets, Challenge organization, Evaluation metrics; Figures 2–4; Table 1"}],"diagram":{"title":"Evaluation procedure","steps":["Allowed inputs: Released sequence data and track-specific reference resources.","Splits: CAMI II supplied public-genome practice datasets with ground truth before its blinded challenge. Challenge datasets were marine, strain-madness and plant-associated communities; these are challenge conditions, not a standard supervised train/validation/test partition.","Metrics: MetaQUAST evaluates genome fraction, mismatches per 100 kb, duplication ratio, NGA50 and misassemblies; strain precision and recall additionally measure high-quality strain reconstruction. Undefined per-genome NGA50 is set to zero before the reported genome average."],"caption":"Conceptual procedure. Task variants and protocol versions retain their separate scoring conditions.","source_ids":["evidence-benchmark-cami-snapshot","evidence-discovery-final-cami2"],"source_locator":"Official CAMI homepage: initiative description; challenges; dataset correction notices; toolkit citations; Methods: Challenge datasets, Challenge organization, Evaluation metrics; Figures 2–4; Table 1; Methods: Assembly metrics; Figure 1"},"coverage":"limited","gaps":["Uncertainty: Figure 1 and the assembly-metrics Methods define descriptive per-genome and dataset summaries, but no universal bootstrap or repeated-seed interval for all assembly scores."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction."}}},"description":"metagenome assembly","facets":{"areas":["microbiome"]},"id":"discovery-benchmark-cami-metagenome-assembly","kind":"benchmark","links":[{"relation":"parent","target_id":"discovery-benchmark-cami"},{"relation":"part_of","target_id":"discovery-benchmark-cami"}],"name":"CAMI metagenome assembly","source_ids":["src-discovery-cami"],"status":"discovered"}
{"attributes":{"entity_level":"task","scope_note":"Specialist molecular or omics evaluation; protocol details require review before numerical comparison.","task":"taxonomic binning","version":null,"historical_missing_metadata":{"dataset_release":"unextracted","metric_implementation":"unextracted","split_manifest":"unextracted","version":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"The CAMI taxonomic-binning task can be understood through its documented assessment tool; this guide does not identify a challenge-specific run.","summary_source_ids":["src-discovery-cami-challenge-amber"],"summary_source_locator":"Pinned README: introduction; Metrics computed per bin/per sample; input formats","sections":[{"title":"Evaluation methodology","body":"CAMI is a series of blinded metagenomic software challenges. In CAMI II, participants received simulated short and long reads from defined communities and submitted assemblies, genome bins, taxonomic assignments or abundance profiles. Reference truth was used only for scoring, and software versions, input read types and community conditions were kept distinct.","source_ids":["evidence-discovery-final-cami2"],"source_locator":"Methods: Challenge datasets, Challenge organization, Evaluation metrics; Figures 2–4; Table 1"}],"facts":[{"label":"Datasets","value":"A selected CAMI challenge dataset and matching gold standard are required; this entry does not fix the edition.","status":"source_checked","source_ids":["src-discovery-cami-challenge-amber"],"source_locator":"Pinned README: introduction; Metrics computed per bin/per sample; input formats"},{"label":"Splits","value":"CAMI II supplied public-genome practice datasets with ground truth before its blinded challenge. Challenge datasets were marine, strain-madness and plant-associated communities; these are challenge conditions, not a standard supervised train/validation/test partition.","status":"source_checked","source_ids":["evidence-discovery-final-cami2"],"source_locator":"Methods: Challenge datasets, Challenge organization, Evaluation metrics; Figures 2–4; Table 1"},{"label":"Metrics","value":"Bin purity/completeness; sample accuracy, contamination, adjusted Rand index, binned fraction and recovered-genome counts; UniFrac for taxonomic binning.","status":"source_checked","source_ids":["src-discovery-cami-challenge-amber"],"source_locator":"Pinned README: introduction; Metrics computed per bin/per sample; input formats"},{"label":"Baselines","value":"Submitted programs are compared under the same data condition. Gold-standard assemblies and MEGAHIT assemblies separate binning performance from upstream assembly error; published method identities and versions are listed in Table 1.","status":"source_checked","source_ids":["evidence-discovery-final-cami2"],"source_locator":"Methods: Challenge datasets, Challenge organization, Evaluation metrics; Figures 2–4; Table 1"},{"label":"Leakage controls","value":"Challenge genome data and metadata were kept confidential until the challenge ended. Public reference collections dated 8 January 2019 were supplied for reference-based methods. CAMI II also includes public genomes, so novelty is stratified rather than assumed for every organism.","status":"source_checked","source_ids":["evidence-discovery-final-cami2"],"source_locator":"Methods: Challenge datasets, Challenge organization, Evaluation metrics; Figures 2–4; Table 1"},{"label":"Uncertainty","value":"Uncertainty is task-specific: taxonomic binning Figure 3 uses standard errors across bins; taxonomic profiling Figure 4 reports means across samples with standard deviations. These are not a common seed-based interval for every CAMI metric.","status":"source_checked","source_ids":["evidence-discovery-final-cami2"],"source_locator":"Figure 3 and Figure 4 captions: standard error across taxonomic bins versus standard deviation across samples"},{"label":"Entity type","value":"Constituent benchmark task: CAMI taxonomic binning","status":"source_checked","source_ids":["src-discovery-cami-challenge-amber"],"source_locator":"Pinned README: introduction; Metrics computed per bin/per sample; input formats"},{"label":"Organisms","value":"AMBER accepts community gold standards; the selected CAMI dataset supplies organism membership.","status":"inapplicable","source_ids":["src-discovery-cami-challenge-amber"],"source_locator":"Pinned README: introduction; Metrics computed per bin/per sample; input formats"},{"label":"Assays","value":"Metagenomic sequence assignments and their gold-standard bins/taxa.","status":"source_checked","source_ids":["src-discovery-cami-challenge-amber"],"source_locator":"Pinned README: introduction; Metrics computed per bin/per sample; input formats"},{"label":"Allowed inputs","value":"Predicted sequence-to-bin or sequence-to-taxon assignments and the matched gold standard.","status":"source_checked","source_ids":["src-discovery-cami-challenge-amber"],"source_locator":"Pinned README: introduction; Metrics computed per bin/per sample; input formats"},{"label":"Adaptation","value":"AMBER assesses assignments; predictor-training conditions are outside the evaluator.","status":"inapplicable","source_ids":["src-discovery-cami-challenge-amber"],"source_locator":"Pinned README: introduction; Metrics computed per bin/per sample; input formats"}],"strengths":[{"text":"Purity and completeness expose different binning errors.","source_ids":["src-discovery-cami-challenge-amber"],"source_locator":"Pinned README: introduction; Metrics computed per bin/per sample; input formats"}],"limitations":[{"text":"This profile documents CAMI II as a concrete protocol example. Other CAMI rounds may use different genomes, reference databases and metrics; strain diversity and input assembly quality materially change the task.","source_ids":["evidence-discovery-final-cami2"],"source_locator":"Methods: Challenge datasets, Challenge organization, Evaluation metrics; Figures 2–4; Table 1"}],"diagram":{"title":"Evaluation procedure","steps":["Allowed inputs: Predicted sequence-to-bin or sequence-to-taxon assignments and the matched gold standard.","Splits: CAMI II supplied public-genome practice datasets with ground truth before its blinded challenge. Challenge datasets were marine, strain-madness and plant-associated communities; these are challenge conditions, not a standard supervised train/validation/test partition.","Metrics: Bin purity/completeness; sample accuracy, contamination, adjusted Rand index, binned fraction and recovered-genome counts; UniFrac for taxonomic binning."],"caption":"Conceptual procedure. Task variants and protocol versions retain their separate scoring conditions.","source_ids":["src-discovery-cami-challenge-amber","evidence-discovery-final-cami2"],"source_locator":"Pinned README: introduction; Metrics computed per bin/per sample; input formats; Methods: Challenge datasets, Challenge organization, Evaluation metrics; Figures 2–4; Table 1"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction. Independent automated spot review corrected interval terminology and sequence-identity scope against the original figure captions and methods."}}},"description":"taxonomic binning","facets":{"areas":["microbiome"]},"id":"discovery-benchmark-cami-taxonomic-binning","kind":"benchmark","links":[{"relation":"parent","target_id":"discovery-benchmark-cami"},{"relation":"part_of","target_id":"discovery-benchmark-cami"}],"name":"CAMI taxonomic binning","source_ids":["src-discovery-cami"],"status":"discovered"}
{"attributes":{"entity_level":"task","scope_note":"Specialist molecular or omics evaluation; protocol details require review before numerical comparison.","task":"taxonomic profiling","version":null,"historical_missing_metadata":{"dataset_release":"unextracted","metric_implementation":"unextracted","split_manifest":"unextracted","version":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"The CAMI taxonomic-profiling task can be understood through its documented assessment tool; this guide does not identify a challenge-specific run.","summary_source_ids":["src-discovery-cami-challenge-opal"],"summary_source_locator":"Pinned README: introduction; Computed metrics; input format; example evaluations","sections":[{"title":"Evaluation methodology","body":"CAMI is a series of blinded metagenomic software challenges. In CAMI II, participants received simulated short and long reads from defined communities and submitted assemblies, genome bins, taxonomic assignments or abundance profiles. Reference truth was used only for scoring, and software versions, input read types and community conditions were kept distinct.","source_ids":["evidence-discovery-final-cami2"],"source_locator":"Methods: Challenge datasets, Challenge organization, Evaluation metrics; Figures 2–4; Table 1"}],"facts":[{"label":"Datasets","value":"A selected CAMI challenge dataset and matching gold standard are required; this entry does not fix the edition.","status":"source_checked","source_ids":["src-discovery-cami-challenge-opal"],"source_locator":"Pinned README: introduction; Computed metrics; input format; example evaluations"},{"label":"Splits","value":"CAMI II supplied public-genome practice datasets with ground truth before its blinded challenge. Challenge datasets were marine, strain-madness and plant-associated communities; these are challenge conditions, not a standard supervised train/validation/test partition.","status":"source_checked","source_ids":["evidence-discovery-final-cami2"],"source_locator":"Methods: Challenge datasets, Challenge organization, Evaluation metrics; Figures 2–4; Table 1"},{"label":"Metrics","value":"Precision, recall, F1, Jaccard, L1 error, UniFrac, Bray–Curtis and diversity measures.","status":"source_checked","source_ids":["src-discovery-cami-challenge-opal"],"source_locator":"Pinned README: introduction; Computed metrics; input format; example evaluations"},{"label":"Baselines","value":"Submitted programs are compared under the same data condition. Gold-standard assemblies and MEGAHIT assemblies separate binning performance from upstream assembly error; published method identities and versions are listed in Table 1.","status":"source_checked","source_ids":["evidence-discovery-final-cami2"],"source_locator":"Methods: Challenge datasets, Challenge organization, Evaluation metrics; Figures 2–4; Table 1"},{"label":"Leakage controls","value":"Challenge genome data and metadata were kept confidential until the challenge ended. Public reference collections dated 8 January 2019 were supplied for reference-based methods. CAMI II also includes public genomes, so novelty is stratified rather than assumed for every organism.","status":"source_checked","source_ids":["evidence-discovery-final-cami2"],"source_locator":"Methods: Challenge datasets, Challenge organization, Evaluation metrics; Figures 2–4; Table 1"},{"label":"Uncertainty","value":"Uncertainty is task-specific: taxonomic binning Figure 3 uses standard errors across bins; taxonomic profiling Figure 4 reports means across samples with standard deviations. These are not a common seed-based interval for every CAMI metric.","status":"source_checked","source_ids":["evidence-discovery-final-cami2"],"source_locator":"Figure 3 and Figure 4 captions: standard error across taxonomic bins versus standard deviation across samples"},{"label":"Entity type","value":"Constituent benchmark task: CAMI taxonomic profiling","status":"source_checked","source_ids":["src-discovery-cami-challenge-opal"],"source_locator":"Pinned README: introduction; Computed metrics; input format; example evaluations"},{"label":"Organisms","value":"Taxa are supplied by the selected reference community.","status":"inapplicable","source_ids":["src-discovery-cami-challenge-opal"],"source_locator":"Pinned README: introduction; Computed metrics; input format; example evaluations"},{"label":"Assays","value":"Challenge-specific metagenomic sequence data and reference composition.","status":"source_checked","source_ids":["src-discovery-cami-challenge-opal"],"source_locator":"Pinned README: introduction; Computed metrics; input format; example evaluations"},{"label":"Allowed inputs","value":"Predicted and reference abundance profiles.","status":"source_checked","source_ids":["src-discovery-cami-challenge-opal"],"source_locator":"Pinned README: introduction; Computed metrics; input format; example evaluations"},{"label":"Adaptation","value":"OPAL scores profiles; it does not train the submitting method.","status":"inapplicable","source_ids":["src-discovery-cami-challenge-opal"],"source_locator":"Pinned README: introduction; Computed metrics; input format; example evaluations"}],"strengths":[{"text":"Abundance agreement and taxon detection are separate evaluation dimensions.","source_ids":["src-discovery-cami-challenge-opal"],"source_locator":"Pinned README: introduction; Computed metrics; input format; example evaluations"}],"limitations":[{"text":"This profile documents CAMI II as a concrete protocol example. Other CAMI rounds may use different genomes, reference databases and metrics; strain diversity and input assembly quality materially change the task.","source_ids":["evidence-discovery-final-cami2"],"source_locator":"Methods: Challenge datasets, Challenge organization, Evaluation metrics; Figures 2–4; Table 1"}],"diagram":{"title":"Evaluation procedure","steps":["Allowed inputs: Predicted and reference abundance profiles.","Splits: CAMI II supplied public-genome practice datasets with ground truth before its blinded challenge. Challenge datasets were marine, strain-madness and plant-associated communities; these are challenge conditions, not a standard supervised train/validation/test partition.","Metrics: Precision, recall, F1, Jaccard, L1 error, UniFrac, Bray–Curtis and diversity measures."],"caption":"Conceptual procedure. Task variants and protocol versions retain their separate scoring conditions.","source_ids":["src-discovery-cami-challenge-opal","evidence-discovery-final-cami2"],"source_locator":"Pinned README: introduction; Computed metrics; input format; example evaluations; Methods: Challenge datasets, Challenge organization, Evaluation metrics; Figures 2–4; Table 1"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction. Independent automated spot review corrected interval terminology and sequence-identity scope against the original figure captions and methods."}}},"description":"taxonomic profiling","facets":{"areas":["microbiome"]},"id":"discovery-benchmark-cami-taxonomic-profiling","kind":"benchmark","links":[{"relation":"parent","target_id":"discovery-benchmark-cami"},{"relation":"part_of","target_id":"discovery-benchmark-cami"}],"name":"CAMI taxonomic profiling","source_ids":["src-discovery-cami"],"status":"discovered"}
{"attributes":{"entity_level":"challenge","scope_note":"Specialist molecular or omics evaluation; protocol details require review before numerical comparison.","task":"Protein interaction docking assessment","version":null,"historical_missing_metadata":{"dataset_release":"unextracted","metric_implementation":"unextracted","split_manifest":"unextracted","version":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"CAPRI assesses blind predictions of protein-complex structures supplied before experimental publication.","summary_source_ids":["evidence-benchmark-capri-snapshot"],"summary_source_locator":"Official CAPRI homepage: introductory description","sections":[{"title":"Evaluation methodology","body":"CAPRI is a blinded interaction-structure challenge. Predictors submit complex models, while scoring tracks select promising models from candidate sets. Assessors compare submitted interfaces with withheld experimental structures; quality classes and DockQ-based summaries retain the target and round context.","source_ids":["evidence-discovery-final-capri"],"source_locator":"Assessment page: joint CASP-CAPRI Round 57 and Round 58"}],"facts":[{"label":"Datasets","value":"Challenge-round protein-complex targets supplied by experimental contributors.","status":"source_checked","source_ids":["evidence-benchmark-capri-snapshot"],"source_locator":"Official CAPRI homepage: introductory description"},{"label":"Splits","value":"Target structures are unreleased to predictors during the blind prediction phase.","status":"source_checked","source_ids":["evidence-benchmark-capri-snapshot"],"source_locator":"Official CAPRI homepage: introductory description"},{"label":"Metrics","value":"Interface contact recovery (Fnat), interface RMSD and ligand RMSD define classical CAPRI quality classes. DockQ combines these complementary structural measures into a continuous score.","status":"source_checked","source_ids":["evidence-discovery-final-dockq"],"source_locator":"Introduction and Methods: CAPRI criteria and DockQ"},{"label":"Baselines","value":"Round 57 assessment compares submissions with ColabFold and an AlphaFold 3 submission. Round 58 supplies MassiveFold candidate models and tests improvement over that model pool; other rounds have different reference conditions.","status":"source_checked","source_ids":["evidence-discovery-final-capri"],"source_locator":"Assessment page: joint CASP-CAPRI Round 57 and Round 58"},{"label":"Leakage controls","value":"The prospective blind-target design separates submissions from public experimental structures.","status":"source_checked","source_ids":["evidence-benchmark-capri-snapshot"],"source_locator":"Official CAPRI homepage: introductory description"},{"label":"Uncertainty","value":"The inspected assessment page provides round-specific rankings and quality classes but does not define one common bootstrap interval for all CAPRI targets and rounds.","status":"unreported","source_ids":["evidence-discovery-final-capri"],"source_locator":"Assessment page: joint CASP-CAPRI Round 57 and Round 58"},{"label":"Entity type","value":"Blind protein-complex prediction challenge.","status":"source_checked","source_ids":["evidence-benchmark-capri-snapshot"],"source_locator":"Official CAPRI homepage: introductory description"},{"label":"Organisms","value":"Target-dependent protein complexes; this is not a single-species benchmark.","status":"source_checked","source_ids":["evidence-benchmark-capri-snapshot"],"source_locator":"Official CAPRI homepage: introductory description"},{"label":"Assays","value":"Experimentally determined complex structures withheld for assessment.","status":"source_checked","source_ids":["evidence-benchmark-capri-snapshot"],"source_locator":"Official CAPRI homepage: introductory description"},{"label":"Allowed inputs","value":"Target information distributed for each challenge round.","status":"source_checked","source_ids":["evidence-benchmark-capri-snapshot"],"source_locator":"Official CAPRI homepage: introductory description"},{"label":"Adaptation","value":"Prediction submissions are assessed against subsequently available experimental structures.","status":"source_checked","source_ids":["evidence-benchmark-capri-snapshot"],"source_locator":"Official CAPRI homepage: introductory description"}],"strengths":[{"text":"Blind rounds separate predictor submissions from the assessment structures.","source_ids":["evidence-benchmark-capri-snapshot"],"source_locator":"Official CAPRI homepage: introductory description"}],"limitations":[{"text":"A model-selection round with supplied structures is different from docking from sequence or component structures. Reference inputs, candidate pools and assessment-unit definitions must stay visible.","source_ids":["evidence-discovery-final-capri"],"source_locator":"Assessment page: joint CASP-CAPRI Round 57 and Round 58"}],"diagram":{"title":"Evaluation procedure","steps":["Allowed inputs: Target information distributed for each challenge round.","Splits: Target structures are unreleased to predictors during the blind prediction phase.","Metrics: Interface contact recovery (Fnat), interface RMSD and ligand RMSD define classical CAPRI quality classes. DockQ combines these complementary structural measures into a continuous score."],"caption":"Conceptual procedure. Task variants and protocol versions retain their separate scoring conditions.","source_ids":["evidence-benchmark-capri-snapshot","evidence-discovery-final-dockq"],"source_locator":"Official CAPRI homepage: introductory description; Introduction and Methods: CAPRI criteria and DockQ"},"coverage":"limited","gaps":["Uncertainty: The inspected assessment page provides round-specific rankings and quality classes but does not define one common bootstrap interval for all CAPRI targets and rounds."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction."}}},"description":"Protein interaction docking assessment","facets":{"areas":["protein-structure"]},"id":"discovery-benchmark-capri","kind":"benchmark","links":[],"name":"CAPRI","source_ids":["src-discovery-capri"],"status":"discovered"}
{"attributes":{"entity_level":"challenge","scope_note":"Specialist molecular or omics evaluation; protocol details require review before numerical comparison.","task":"Community protein structure assessment","version":null,"historical_missing_metadata":{"dataset_release":"unextracted","metric_implementation":"unextracted","split_manifest":"unextracted","version":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"CASP assesses structure-prediction methods through blind predictions and category-specific evaluation.","summary_source_ids":["evidence-benchmark-casp-snapshot"],"summary_source_locator":"Protein Structure Prediction Center homepage: Welcome; assessment and archived-data description","sections":[{"title":"Evaluation methodology","body":"CASP evaluates predictions against experimentally determined structures withheld at submission time. Its independent assessors define evaluation units and category-specific scores. In CASP16 monomer assessment, controlled MSA inputs and automated ColabFold references help distinguish pipeline contributions from differences in information supplied.","source_ids":["evidence-discovery-final-casp16"],"source_locator":"CASP16 monomer assessment: Evaluation of Model 6, model sampling and head-to-head comparisons; Methods"}],"facts":[{"label":"Datasets","value":"Experiment-specific targets, submissions and numerical assessment files are archived by the Prediction Center.","status":"source_checked","source_ids":["evidence-benchmark-casp-snapshot"],"source_locator":"Protein Structure Prediction Center homepage: Welcome; assessment and archived-data description"},{"label":"Splits","value":"Blind prediction tasks are organized by assessment edition and category.","status":"source_checked","source_ids":["evidence-benchmark-casp-snapshot"],"source_locator":"Protein Structure Prediction Center homepage: Welcome; assessment and archived-data description"},{"label":"Metrics","value":"CASP16 monomer assessment combines standardized GDT_HA, QSE, reLLG_const, SphGr, CAD_AA, GDC_SC, AL0_P, lDDT and MolProbity measures. These scores apply to its defined evaluation units; other CASP categories use different protocols.","status":"source_checked","source_ids":["evidence-discovery-final-casp16"],"source_locator":"Monomer assessment: scoring formula and Methods"},{"label":"Baselines","value":"CASP16 monomer assessment uses ColabFold as an automated reference, includes MassiveFold sampling, and compares models constrained to the same ColabFold MSAs to separate input improvements from prediction-network changes.","status":"source_checked","source_ids":["evidence-discovery-final-casp16"],"source_locator":"CASP16 monomer assessment: Evaluation of Model 6, model sampling and head-to-head comparisons; Methods"},{"label":"Leakage controls","value":"The blind-target setting is explicit; exact training-cutoff enforcement and template restrictions depend on the category.","status":"source_checked","source_ids":["evidence-benchmark-casp-snapshot"],"source_locator":"Protein Structure Prediction Center homepage: Welcome; assessment and archived-data description"},{"label":"Uncertainty","value":"CASP16 monomer head-to-head comparisons use 1,000 bootstrap samples of evaluation units. This protocol is category- and round-specific; it is not a confidence interval for every CASP result.","status":"source_checked","source_ids":["evidence-discovery-final-casp16"],"source_locator":"CASP16 monomer assessment: Evaluation of Model 6, model sampling and head-to-head comparisons; Methods"},{"label":"Entity type","value":"Community protein-structure prediction experiment.","status":"source_checked","source_ids":["evidence-benchmark-casp-snapshot"],"source_locator":"Protein Structure Prediction Center homepage: Welcome; assessment and archived-data description"},{"label":"Organisms","value":"Target-dependent proteins and complexes.","status":"source_checked","source_ids":["evidence-benchmark-casp-snapshot"],"source_locator":"Protein Structure Prediction Center homepage: Welcome; assessment and archived-data description"},{"label":"Assays","value":"Experiment-specific experimentally determined structures.","status":"source_checked","source_ids":["evidence-benchmark-casp-snapshot"],"source_locator":"Protein Structure Prediction Center homepage: Welcome; assessment and archived-data description"},{"label":"Allowed inputs","value":"Sequence and target information issued for the selected experiment.","status":"source_checked","source_ids":["evidence-benchmark-casp-snapshot"],"source_locator":"Protein Structure Prediction Center homepage: Welcome; assessment and archived-data description"},{"label":"Adaptation","value":"Prediction and assessment phases are separated; rules vary by target category and experiment.","status":"source_checked","source_ids":["evidence-benchmark-casp-snapshot"],"source_locator":"Protein Structure Prediction Center homepage: Welcome; assessment and archived-data description"}],"strengths":[{"text":"Archived targets, submissions and assessment files support round-specific comparisons.","source_ids":["evidence-benchmark-casp-snapshot"],"source_locator":"Protein Structure Prediction Center homepage: Welcome; assessment and archived-data description"}],"limitations":[{"text":"CASP rounds and categories differ in targets, allowed information and scoring. Domain-level, whole-complex and nucleic-acid results cannot be pooled as interchangeable observations.","source_ids":["evidence-discovery-final-casp16"],"source_locator":"CASP16 monomer assessment: Evaluation of Model 6, model sampling and head-to-head comparisons; Methods"}],"diagram":{"title":"Evaluation procedure","steps":["Allowed inputs: Sequence and target information issued for the selected experiment.","Splits: Blind prediction tasks are organized by assessment edition and category.","Metrics: CASP16 monomer assessment combines standardized GDT_HA, QSE, reLLG_const, SphGr, CAD_AA, GDC_SC, AL0_P, lDDT and MolProbity measures. These scores apply to its defined evaluation units; other CASP categories use different protocols."],"caption":"Conceptual procedure. Task variants and protocol versions retain their separate scoring conditions.","source_ids":["evidence-benchmark-casp-snapshot","evidence-discovery-final-casp16"],"source_locator":"Protein Structure Prediction Center homepage: Welcome; assessment and archived-data description; Monomer assessment: scoring formula and Methods"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction."}}},"description":"Community protein structure assessment","facets":{"areas":["protein-structure"]},"id":"discovery-benchmark-casp","kind":"benchmark","links":[],"name":"CASP","source_ids":["src-discovery-casp"],"status":"discovered"}
{"attributes":{"entity_level":"suite","scope_note":"Specialist molecular or omics evaluation; protocol details require review before numerical comparison.","task":"Human regulatory DNA representation evaluation","version":null,"historical_missing_metadata":{"dataset_release":"unextracted","metric_implementation":"unextracted","split_manifest":"unextracted","version":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"DART-Eval measures human regulatory-DNA representations under zero-shot, probing and fine-tuning regimes.","summary_source_ids":["src-discovery-kundajelab-dart-eval"],"summary_source_locator":"Pinned README: Overview; Data download; task-organized outputs; probing/fine-tuning sections","sections":[{"title":"Evaluation methodology","body":"DART-Eval separates regulatory sequence discrimination, motif sensitivity, cell-type specificity, quantitative accessibility and variant effects. It uses matched controls and consistent chromosome partitions where models are fitted. A result therefore needs its precise task and adaptation setting, rather than a single regulatory-intelligence score.","source_ids":["evidence-discovery-final-dart"],"source_locator":"Appendix: training/test splits, clustering and supervised evaluation; reproducibility checklist"}],"facts":[{"label":"Datasets","value":"Task-specific HDF5 inputs/outputs with raw data and evaluated model outputs organized in a Synapse project.","status":"source_checked","source_ids":["src-discovery-kundajelab-dart-eval"],"source_locator":"Pinned README: Overview; Data download; task-organized outputs; probing/fine-tuning sections"},{"label":"Splits","value":"Training uses chromosomes other than the held-out sets; validation is chromosomes 6 and 21; test is chromosomes 5, 10, 14, 18, 20 and 22. Fitted checkpoints are chosen using validation loss.","status":"source_checked","source_ids":["evidence-discovery-final-dart"],"source_locator":"Appendix: common train/validation/test split"},{"label":"Metrics","value":"Regulatory and variant classification use AUROC/AUPRC; quantitative accessibility uses Pearson and Spearman correlations on peaks alone and peaks plus background. Cell-type clustering uses adjusted mutual information. Motif sensitivity is a separate paired-sequence evaluation.","status":"source_checked","source_ids":["evidence-discovery-final-dart"],"source_locator":"Appendix: training/test splits, clustering and supervised evaluation; reproducibility checklist"},{"label":"Baselines","value":"Probing-head-like ab initio models are documented alongside pretrained-model probing and fine-tuning.","status":"source_checked","source_ids":["src-discovery-kundajelab-dart-eval"],"source_locator":"Pinned README: Overview; Data download; task-organized outputs; probing/fine-tuning sections"},{"label":"Leakage controls","value":"The common genomic partition holds out chromosomes 5, 10, 14, 18, 20 and 22 for test, with chromosomes 6 and 21 for validation. Lowest validation loss selects checkpoints. Dinucleotide-shuffled negatives preserve composition; this control does not eliminate all pretrained-genome exposure.","status":"source_checked","source_ids":["evidence-discovery-final-dart"],"source_locator":"Appendix: training/test splits, clustering and supervised evaluation; reproducibility checklist"},{"label":"Uncertainty","value":"Clustering is repeated 100 times and reports a 95% interval across clustering runs. Motif-sensitivity intervals are provided in the linked artifacts; the paper does not report repeated-training intervals for every other task.","status":"source_checked","source_ids":["evidence-discovery-final-dart"],"source_locator":"Appendix: training/test splits, clustering and supervised evaluation; reproducibility checklist"},{"label":"Entity type","value":"DNA regulatory evaluation suite.","status":"source_checked","source_ids":["src-discovery-kundajelab-dart-eval"],"source_locator":"Pinned README: Overview; Data download; task-organized outputs; probing/fine-tuning sections"},{"label":"Organisms","value":"Human regulatory datasets.","status":"source_checked","source_ids":["src-discovery-kundajelab-dart-eval"],"source_locator":"Pinned README: Overview; Data download; task-organized outputs; probing/fine-tuning sections"},{"label":"Assays","value":"Regulatory-element, chromatin-accessibility and variant-effect measurements.","status":"source_checked","source_ids":["src-discovery-kundajelab-dart-eval"],"source_locator":"Pinned README: Overview; Data download; task-organized outputs; probing/fine-tuning sections"},{"label":"Allowed inputs","value":"Task-specific DNA sequences and HDF5 targets; raw data and model outputs are linked through Synapse.","status":"source_checked","source_ids":["src-discovery-kundajelab-dart-eval"],"source_locator":"Pinned README: Overview; Data download; task-organized outputs; probing/fine-tuning sections"},{"label":"Adaptation","value":"Zero-shot, probing, fine-tuning and ab-initio comparison regimes are distinct.","status":"source_checked","source_ids":["src-discovery-kundajelab-dart-eval"],"source_locator":"Pinned README: Overview; Data download; task-organized outputs; probing/fine-tuning sections"}],"strengths":[{"text":"Tasks separate generic regulatory recognition from cell-specific activity and variant effects.","source_ids":["src-discovery-kundajelab-dart-eval"],"source_locator":"Pinned README: Overview; Data download; task-organized outputs; probing/fine-tuning sections"}],"limitations":[{"text":"The assessed tasks use local regulatory contexts. A chromosome-held-out supervised test does not establish that those sequences were absent from unsupervised pretraining.","source_ids":["evidence-discovery-final-dart"],"source_locator":"Appendix: training/test splits, clustering and supervised evaluation; reproducibility checklist"}],"diagram":{"title":"Evaluation procedure","steps":["Allowed inputs: Task-specific DNA sequences and HDF5 targets; raw data and model outputs are linked through Synapse.","Splits: Training uses chromosomes other than the held-out sets; validation is chromosomes 6 and 21; test is chromosomes 5, 10, 14, 18, 20 and 22. Fitted checkpoints are chosen using validation loss.","Metrics: Regulatory and variant classification use AUROC/AUPRC; quantitative accessibility uses Pearson and Spearman correlations on peaks alone and peaks plus background. Cell-type clustering uses adjusted mutual information. Motif sensitivity is a separate paired-sequence evaluation."],"caption":"Conceptual procedure. Task variants and protocol versions retain their separate scoring conditions.","source_ids":["src-discovery-kundajelab-dart-eval","evidence-discovery-final-dart"],"source_locator":"Pinned README: Overview; Data download; task-organized outputs; probing/fine-tuning sections; Appendix: common train/validation/test split; Appendix: training/test splits, clustering and supervised evaluation; reproducibility checklist"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction."}}},"description":"Human regulatory DNA representation evaluation","facets":{"areas":["genomics"]},"id":"discovery-benchmark-dart-eval","kind":"benchmark","links":[],"name":"DART-Eval","source_ids":["src-discovery-kundajelab-dart-eval"],"status":"discovered"}
{"attributes":{"entity_level":"suite","scope_note":"Specialist molecular or omics evaluation; protocol details require review before numerical comparison.","task":"Generalisation in protein fitness landscapes","version":null,"historical_missing_metadata":{"dataset_release":"unextracted","metric_implementation":"unextracted","split_manifest":"unextracted","version":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"FLIP evaluates protein-sequence representations using multiple deliberately defined train/test splits.","summary_source_ids":["src-discovery-j-snackkb-flip"],"summary_source_locator":"Pinned README: repository organization; Splits","sections":[{"title":"Evaluation methodology","body":"FLIP evaluates sequence-to-fitness models under protein-engineering distribution shifts. It separates landscapes from their partitions: the same assay can test random interpolation, higher mutation counts, higher fitness or transfer across sequence groups. Comparisons are meaningful within the same landscape and split.","source_ids":["evidence-discovery-final-flip"],"source_locator":"Sections 3–5; Tables 2 and 4–7"}],"facts":[{"label":"Datasets","value":"Protein sequence/property collections distributed as raw data, processed splits and FASTA resources.","status":"source_checked","source_ids":["src-discovery-j-snackkb-flip"],"source_locator":"Pinned README: repository organization; Splits"},{"label":"Splits","value":"The splits directory documents biological/statistical split logic; multiple splits may exist for one dataset.","status":"source_checked","source_ids":["src-discovery-j-snackkb-flip"],"source_locator":"Pinned README: repository organization; Splits"},{"label":"Metrics","value":"Spearman rank correlation between predicted and measured fitness on each landscape/split.","status":"source_checked","source_ids":["evidence-discovery-final-flip"],"source_locator":"Sections 3–5; Tables 2 and 4–7"},{"label":"Baselines","value":"A baselines directory provides reference implementations.","status":"source_checked","source_ids":["src-discovery-j-snackkb-flip"],"source_locator":"Pinned README: repository organization; Splits"},{"label":"Leakage controls","value":"FLIP contrasts random partitions with mutation-number, fitness and sequence-family or diversity partitions. Its purpose is to expose distribution shifts; a random partition is an easier control, not an equivalent test.","status":"source_checked","source_ids":["evidence-discovery-final-flip"],"source_locator":"Sections 3–5; Tables 2 and 4–7"},{"label":"Uncertainty","value":"The inspected baseline tables report point correlations. Their Methods do not define a common repeated-seed or bootstrap confidence interval for all landscape/split results.","status":"unreported","source_ids":["evidence-discovery-final-flip"],"source_locator":"Sections 3–5; Tables 2 and 4–7"},{"label":"Entity type","value":"Protein sequence learning benchmark with multiple split regimes.","status":"source_checked","source_ids":["src-discovery-j-snackkb-flip"],"source_locator":"Pinned README: repository organization; Splits"},{"label":"Organisms","value":"GB1 binding variants, adeno-associated virus capsid variants and Meltome proteins across the tree of life; the landscapes have different organism and assay scopes.","status":"source_checked","source_ids":["evidence-discovery-final-flip"],"source_locator":"Sections 3–5; Tables 2 and 4–7"},{"label":"Assays","value":"Measured protein properties from the selected source datasets.","status":"source_checked","source_ids":["src-discovery-j-snackkb-flip"],"source_locator":"Pinned README: repository organization; Splits"},{"label":"Allowed inputs","value":"Protein sequences with property labels.","status":"source_checked","source_ids":["src-discovery-j-snackkb-flip"],"source_locator":"Pinned README: repository organization; Splits"},{"label":"Adaptation","value":"Supervised learning on the chosen training partition; different splits test different generalization conditions.","status":"source_checked","source_ids":["src-discovery-j-snackkb-flip"],"source_locator":"Pinned README: repository organization; Splits"}],"strengths":[{"text":"Split semaphores explicitly mark active, cautionary and obsolete comparison settings.","source_ids":["src-discovery-j-snackkb-flip"],"source_locator":"Pinned README: repository organization; Splits"}],"limitations":[{"text":"Random and extrapolative splits answer different questions. Closely related mutants are intentional in engineering tasks, so independence cannot be reduced to a universal sequence-identity threshold.","source_ids":["evidence-discovery-final-flip"],"source_locator":"Sections 3–5; Tables 2 and 4–7"}],"diagram":{"title":"Evaluation procedure","steps":["Allowed inputs: Protein sequences with property labels.","Splits: The splits directory documents biological/statistical split logic; multiple splits may exist for one dataset.","Metrics: Spearman rank correlation between predicted and measured fitness on each landscape/split."],"caption":"Conceptual procedure. Task variants and protocol versions retain their separate scoring conditions.","source_ids":["src-discovery-j-snackkb-flip","evidence-discovery-final-flip"],"source_locator":"Pinned README: repository organization; Splits; Sections 3–5; Tables 2 and 4–7"},"coverage":"limited","gaps":["Uncertainty: The inspected baseline tables report point correlations. Their Methods do not define a common repeated-seed or bootstrap confidence interval for all landscape/split results."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction."}}},"description":"Generalisation in protein fitness landscapes","facets":{"areas":["protein-function"]},"id":"discovery-benchmark-flip","kind":"benchmark","links":[],"name":"FLIP","source_ids":["src-discovery-j-snackkb-flip"],"status":"discovered"}
{"attributes":{"entity_level":"suite","scope_note":"Specialist molecular or omics evaluation; protocol details require review before numerical comparison.","task":"Expanded protein fitness landscapes","version":null,"historical_missing_metadata":{"dataset_release":"unextracted","metric_implementation":"unextracted","split_manifest":"unextracted","version":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"FLIP2 evaluates protein fitness prediction under deliberately shifted training and test distributions.","summary_source_ids":["evidence-benchmark-flip2-snapshot"],"summary_source_locator":"FLIP2 official website: Overview; benchmark features; datasets and split categories","sections":[{"title":"Evaluation methodology","body":"FLIP2 extends protein fitness testing to enzyme activity, spectral properties, hydrophobic-core effects and protein interactions. Its partitions hold out mutation positions, mutation numbers, fitness ranges or parent sequences. Simple one-hot ridge models and pretrained versus randomly initialized networks make the baseline comparison interpretable.","source_ids":["evidence-discovery-final-flip2"],"source_locator":"Sections 2–4; Table 1; supplementary result tables"}],"facts":[{"label":"Datasets","value":"Datasets span enzymatic activity, protein interactions and other measured protein properties.","status":"source_checked","source_ids":["evidence-benchmark-flip2-snapshot"],"source_locator":"FLIP2 official website: Overview; benchmark features; datasets and split categories"},{"label":"Splits","value":"Split categories separate mutation counts, positions, mutation identities, fitness ranges or reference proteins.","status":"source_checked","source_ids":["evidence-benchmark-flip2-snapshot"],"source_locator":"FLIP2 official website: Overview; benchmark features; datasets and split categories"},{"label":"Metrics","value":"Spearman rank correlation is the primary metric, with NDCG additionally measuring prioritization of high-fitness variants.","status":"source_checked","source_ids":["evidence-discovery-final-flip2"],"source_locator":"Sections 2–4; Table 1; supplementary result tables"},{"label":"Baselines","value":"Zero-shot protein models, ridge regression and fine-tuned models.","status":"source_checked","source_ids":["evidence-benchmark-flip2-snapshot"],"source_locator":"FLIP2 official website: Overview; benchmark features; datasets and split categories"},{"label":"Leakage controls","value":"Distribution-shift splits are explicit; exact homology/pretraining overlap depends on the dataset.","status":"source_checked","source_ids":["evidence-benchmark-flip2-snapshot"],"source_locator":"FLIP2 official website: Overview; benchmark features; datasets and split categories"},{"label":"Uncertainty","value":"Fine-tuned language models are run five times with different random seeds and validation-based early stopping; reported metrics are averaged. The zero-shot and ridge evaluations follow separate procedures.","status":"source_checked","source_ids":["evidence-discovery-final-flip2"],"source_locator":"Sections 2–4; Table 1; supplementary result tables"},{"label":"Entity type","value":"Protein fitness benchmark collection.","status":"source_checked","source_ids":["evidence-benchmark-flip2-snapshot"],"source_locator":"FLIP2 official website: Overview; benchmark features; datasets and split categories"},{"label":"Organisms","value":"Mixed natural and designed protein landscapes: amylase, imine reductase, NucB, TrpB, hydrophobic-core variants, microbial rhodopsins and a PDZ3–peptide interaction system. These are not a single-species benchmark.","status":"source_checked","source_ids":["evidence-discovery-final-flip2"],"source_locator":"Sections 2–4; Table 1; supplementary result tables"},{"label":"Assays","value":"Enzymatic activity, protein interactions and other measured properties.","status":"source_checked","source_ids":["evidence-benchmark-flip2-snapshot"],"source_locator":"FLIP2 official website: Overview; benchmark features; datasets and split categories"},{"label":"Allowed inputs","value":"Protein sequence representations and dataset-specific labels.","status":"source_checked","source_ids":["evidence-benchmark-flip2-snapshot"],"source_locator":"FLIP2 official website: Overview; benchmark features; datasets and split categories"},{"label":"Adaptation","value":"Supervised prediction under task-specific data splits.","status":"source_checked","source_ids":["evidence-benchmark-flip2-snapshot"],"source_locator":"FLIP2 official website: Overview; benchmark features; datasets and split categories"}],"strengths":[{"text":"Explicit task diversity avoids treating fitness as a single interchangeable measurement.","source_ids":["evidence-benchmark-flip2-snapshot"],"source_locator":"FLIP2 official website: Overview; benchmark features; datasets and split categories"}],"limitations":[{"text":"The same number of training examples does not make random and engineering-oriented splits equivalent. Per-landscape ranking and top-fitness retrieval should be retained alongside any aggregate.","source_ids":["evidence-discovery-final-flip2"],"source_locator":"Sections 2–4; Table 1; supplementary result tables"}],"diagram":{"title":"Evaluation procedure","steps":["Allowed inputs: Protein sequence representations and dataset-specific labels.","Splits: Split categories separate mutation counts, positions, mutation identities, fitness ranges or reference proteins.","Metrics: Spearman rank correlation is the primary metric, with NDCG additionally measuring prioritization of high-fitness variants."],"caption":"Conceptual procedure. Task variants and protocol versions retain their separate scoring conditions.","source_ids":["evidence-benchmark-flip2-snapshot","evidence-discovery-final-flip2"],"source_locator":"FLIP2 official website: Overview; benchmark features; datasets and split categories; Sections 2–4; Table 1; supplementary result tables"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction."}}},"description":"Expanded protein fitness landscapes","facets":{"areas":["protein-function"]},"id":"discovery-benchmark-flip2","kind":"benchmark","links":[],"name":"FLIP2","source_ids":["src-discovery-flip2"],"status":"discovered"}
{"attributes":{"entity_level":"suite","scope_note":"Specialist molecular or omics evaluation; protocol details require review before numerical comparison.","task":"Frozen genomic representations with linear probes","version":null,"historical_missing_metadata":{"dataset_release":"unextracted","metric_implementation":"unextracted","split_manifest":"unextracted","version":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"GENEB compares frozen genomic representations across classification tasks and label-budget regimes.","summary_source_ids":["src-discovery-darlednik-geneb"],"summary_source_locator":"Pinned README: Overview; data regimes; leaderboard aggregation definition","sections":[{"title":"Evaluation methodology","body":"GENEB compares frozen genomic model representations under standardized probe and sample-budget settings. It keeps preprocessing, partitions and random seeds consistent across models and summarizes several biological task categories. Organism and dataset coverage are uneven, so category-specific results remain necessary.","source_ids":["evidence-discovery-final-geneb"],"source_locator":"Evaluation protocol; benchmark construction appendix; Limitations"}],"facts":[{"label":"Datasets","value":"DNA classification tasks grouped into functional categories such as promoters, enhancers, methylation and splice sites.","status":"source_checked","source_ids":["src-discovery-darlednik-geneb"],"source_locator":"Pinned README: Overview; data regimes; leaderboard aggregation definition"},{"label":"Splits","value":"Full-data, ten-shot and one-shot settings use a common frozen-embedding protocol.","status":"source_checked","source_ids":["src-discovery-darlednik-geneb"],"source_locator":"Pinned README: Overview; data regimes; leaderboard aggregation definition"},{"label":"Metrics","value":"MCC; macro aggregation averages category scores, while the labelled micro aggregate averages task scores.","status":"source_checked","source_ids":["src-discovery-darlednik-geneb"],"source_locator":"Pinned README: Overview; data regimes; leaderboard aggregation definition"},{"label":"Baselines","value":"A broad set of genomic foundation models evaluated with the same representation protocol.","status":"source_checked","source_ids":["src-discovery-darlednik-geneb"],"source_locator":"Pinned README: Overview; data regimes; leaderboard aggregation definition"},{"label":"Leakage controls","value":"The protocol fixes preprocessing, partitions and seeds across models, but the inspected paper does not establish a benchmark-wide homology exclusion or pretraining-contamination audit. Shared evaluation settings are not proof that training corpora exclude test sequences.","status":"unreported","source_ids":["evidence-discovery-final-geneb"],"source_locator":"Evaluation protocol; benchmark construction appendix; Limitations"},{"label":"Uncertainty","value":"The stated protocol averages over five fixed random seeds for probing in the 1-shot, 10-shot and full-data regimes. Averaging seeds is not itself a confidence interval.","status":"source_checked","source_ids":["evidence-discovery-final-geneb"],"source_locator":"Evaluation protocol; benchmark construction appendix; Limitations"},{"label":"Entity type","value":"DNA-model classification benchmark suite.","status":"source_checked","source_ids":["src-discovery-darlednik-geneb"],"source_locator":"Pinned README: Overview; data regimes; leaderboard aggregation definition"},{"label":"Organisms","value":"The collection combines tasks from human and other well-studied organisms, including mouse and plant categories. Its limitations explicitly note this organism bias; individual dataset provenance remains the appropriate species definition.","status":"source_checked","source_ids":["evidence-discovery-final-geneb"],"source_locator":"Evaluation protocol; benchmark construction appendix; Limitations"},{"label":"Assays","value":"Promoter, enhancer, methylation and splice-site classification labels.","status":"source_checked","source_ids":["src-discovery-darlednik-geneb"],"source_locator":"Pinned README: Overview; data regimes; leaderboard aggregation definition"},{"label":"Allowed inputs","value":"DNA sequences and classification labels.","status":"source_checked","source_ids":["src-discovery-darlednik-geneb"],"source_locator":"Pinned README: Overview; data regimes; leaderboard aggregation definition"},{"label":"Adaptation","value":"Benchmark-specific supervised evaluation of pretrained DNA models.","status":"source_checked","source_ids":["src-discovery-darlednik-geneb"],"source_locator":"Pinned README: Overview; data regimes; leaderboard aggregation definition"}],"strengths":[{"text":"Task and functional-group aggregation distinguish data-rich groups from balanced coverage.","source_ids":["src-discovery-darlednik-geneb"],"source_locator":"Pinned README: Overview; data regimes; leaderboard aggregation definition"}],"limitations":[{"text":"The paper acknowledges concentration on human and well-studied organisms. A consistent probe protocol cannot establish an unseen-genome boundary for every pretrained model.","source_ids":["evidence-discovery-final-geneb"],"source_locator":"Evaluation protocol; benchmark construction appendix; Limitations"}],"diagram":{"title":"Evaluation procedure","steps":["Allowed inputs: DNA sequences and classification labels.","Splits: Full-data, ten-shot and one-shot settings use a common frozen-embedding protocol.","Metrics: MCC; macro aggregation averages category scores, while the labelled micro aggregate averages task scores."],"caption":"Conceptual procedure. Task variants and protocol versions retain their separate scoring conditions.","source_ids":["src-discovery-darlednik-geneb"],"source_locator":"Pinned README: Overview; data regimes; leaderboard aggregation definition"},"coverage":"limited","gaps":["Leakage controls: The protocol fixes preprocessing, partitions and seeds across models, but the inspected paper does not establish a benchmark-wide homology exclusion or pretraining-contamination audit. Shared evaluation settings are not proof that training corpora exclude test sequences."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction."}}},"description":"Frozen genomic representations with linear probes","facets":{"areas":["genomics"]},"id":"discovery-benchmark-geneb","kind":"benchmark","links":[],"name":"GENEB","source_ids":["src-discovery-darlednik-geneb"],"status":"discovered"}
{"attributes":{"entity_level":"suite","scope_note":"Specialist molecular or omics evaluation; protocol details require review before numerical comparison.","task":"Genomic sequence classification","version":null,"historical_missing_metadata":{"dataset_release":"unextracted","metric_implementation":"unextracted","split_manifest":"unextracted","version":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Genomic Benchmarks packages genomic sequence-classification datasets with explicit versions and supplied train/test folders.","summary_source_ids":["src-discovery-ml-bioinfo-ceitec-genomic-benchmarks"],"summary_source_locator":"Pinned README: repository purpose; info and download_dataset examples","sections":[{"title":"Evaluation methodology","body":"Genomic Benchmarks packages sequence-classification datasets together with versioned genomic coordinates, construction notebooks and a small CNN baseline. Each dataset supplies its own train and test subsets. Dataset-specific negative sampling and split provenance matter as much as model architecture for interpreting performance.","source_ids":["evidence-discovery-final-genomic-benchmarks"],"source_locator":"Methods: Reproducibility and Baseline model; Table 2"}],"facts":[{"label":"Datasets","value":"Named genomic classification datasets; the README illustrates a non-TATA human-promoter task.","status":"source_checked","source_ids":["src-discovery-ml-bioinfo-ceitec-genomic-benchmarks"],"source_locator":"Pinned README: repository purpose; info and download_dataset examples"},{"label":"Splits","value":"The download API delivers prepartitioned train/test data organized by class.","status":"source_checked","source_ids":["src-discovery-ml-bioinfo-ceitec-genomic-benchmarks"],"source_locator":"Pinned README: repository purpose; info and download_dataset examples"},{"label":"Metrics","value":"The original paper reports classification accuracy and F1 for its CNN baseline on each dataset.","status":"source_checked","source_ids":["evidence-discovery-final-genomic-benchmarks"],"source_locator":"Methods: Reproducibility and Baseline model; Table 2"},{"label":"Baselines","value":"The repository provides neural-network training helpers and links experiment reports.","status":"source_checked","source_ids":["src-discovery-ml-bioinfo-ceitec-genomic-benchmarks"],"source_locator":"Pinned README: repository purpose; info and download_dataset examples"},{"label":"Leakage controls","value":"Dataset-construction notebooks use fixed seeds. Generated negative regions match positive lengths and are rejected if they overlap positives; this does not establish a suite-wide chromosome-held-out or homology-filtered partition.","status":"source_checked","source_ids":["evidence-discovery-final-genomic-benchmarks"],"source_locator":"Methods: Reproducibility and Baseline model; Table 2"},{"label":"Uncertainty","value":"Table 2 gives point estimates for PyTorch and TensorFlow baseline implementations. The inspected Methods do not specify repeated-training or bootstrap uncertainty for those values.","status":"unreported","source_ids":["evidence-discovery-final-genomic-benchmarks"],"source_locator":"Methods: Reproducibility and Baseline model; Table 2"},{"label":"Entity type","value":"Repository of genomic sequence classification datasets.","status":"source_checked","source_ids":["src-discovery-ml-bioinfo-ceitec-genomic-benchmarks"],"source_locator":"Pinned README: repository purpose; info and download_dataset examples"},{"label":"Organisms","value":"Dataset-specific organisms; a human non-TATA promoter dataset is documented in the README.","status":"source_checked","source_ids":["src-discovery-ml-bioinfo-ceitec-genomic-benchmarks"],"source_locator":"Pinned README: repository purpose; info and download_dataset examples"},{"label":"Assays","value":"Curated genomic classification labels from linked source datasets.","status":"source_checked","source_ids":["src-discovery-ml-bioinfo-ceitec-genomic-benchmarks"],"source_locator":"Pinned README: repository purpose; info and download_dataset examples"},{"label":"Allowed inputs","value":"Sequences and categorical labels through the dataset loader.","status":"source_checked","source_ids":["src-discovery-ml-bioinfo-ceitec-genomic-benchmarks"],"source_locator":"Pinned README: repository purpose; info and download_dataset examples"},{"label":"Adaptation","value":"Supervised train/test classification with a documented CNN example.","status":"source_checked","source_ids":["src-discovery-ml-bioinfo-ceitec-genomic-benchmarks"],"source_locator":"Pinned README: repository purpose; info and download_dataset examples"}],"strengths":[{"text":"Common download/load conventions expose which named dataset a classifier used.","source_ids":["src-discovery-ml-bioinfo-ceitec-genomic-benchmarks"],"source_locator":"Pinned README: repository purpose; info and download_dataset examples"}],"limitations":[{"text":"Non-overlapping positive and negative genomic intervals do not guarantee independence between related sequences across the train/test boundary. Baseline implementation differences and data versions should remain explicit.","source_ids":["evidence-discovery-final-genomic-benchmarks"],"source_locator":"Methods: Reproducibility and Baseline model; Table 2"}],"diagram":{"title":"Evaluation procedure","steps":["Allowed inputs: Sequences and categorical labels through the dataset loader.","Splits: The download API delivers prepartitioned train/test data organized by class.","Metrics: The original paper reports classification accuracy and F1 for its CNN baseline on each dataset."],"caption":"Conceptual procedure. Task variants and protocol versions retain their separate scoring conditions.","source_ids":["src-discovery-ml-bioinfo-ceitec-genomic-benchmarks","evidence-discovery-final-genomic-benchmarks"],"source_locator":"Pinned README: repository purpose; info and download_dataset examples; Methods: Reproducibility and Baseline model; Table 2"},"coverage":"limited","gaps":["Uncertainty: Table 2 gives point estimates for PyTorch and TensorFlow baseline implementations. The inspected Methods do not specify repeated-training or bootstrap uncertainty for those values."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction."}}},"description":"Genomic sequence classification","facets":{"areas":["genomics"]},"id":"discovery-benchmark-genomic-benchmarks","kind":"benchmark","links":[],"name":"Genomic Benchmarks","source_ids":["src-discovery-ml-bioinfo-ceitec-genomic-benchmarks"],"status":"discovered"}
{"attributes":{"entity_level":"suite","scope_note":"Specialist molecular or omics evaluation; protocol details require review before numerical comparison.","task":"Glycan properties, taxonomy and molecular interactions","version":null,"historical_missing_metadata":{"dataset_release":"unextracted","metric_implementation":"unextracted","split_manifest":"unextracted","version":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"GlycanML evaluates glycan learning across multiple classification and interaction tasks.","summary_source_ids":["src-discovery-glycanml-glycanml","evidence-benchmark-glycanml-bert-species-bert-yaml","evidence-benchmark-glycanml-bert-immunogenicity-bert-yaml","evidence-benchmark-glycanml-bert-link-bert-yaml","evidence-benchmark-glycanml-bert-interaction-bert-yaml"],"summary_source_locator":"Pinned README: Introduction; Experiment configurations; leaderboard; pinned single-task BERT configurations for species, link, immunogenicity and interaction","sections":[{"title":"Evaluation methodology","body":"GlycanML evaluates taxonomy, immunogenicity, glycosylation type and protein–glycan interaction. Glycans are encoded as IUPAC sequences or graphs. Structural motif clusters define the first three task partitions, whereas interaction prediction holds out protein sequence clusters and predicts a transformed fluorescence binding signal.","source_ids":["evidence-discovery-final-glycanml"],"source_locator":"Sections 3.1–3.4"}],"facts":[{"label":"Datasets","value":"Glycan taxonomy, immunogenicity, glycosylation-type and protein–glycan interaction datasets.","status":"source_checked","source_ids":["src-discovery-glycanml-glycanml","evidence-benchmark-glycanml-bert-species-bert-yaml","evidence-benchmark-glycanml-bert-immunogenicity-bert-yaml","evidence-benchmark-glycanml-bert-link-bert-yaml","evidence-benchmark-glycanml-bert-interaction-bert-yaml"],"source_locator":"Pinned README: Introduction; Experiment configurations; leaderboard; pinned single-task BERT configurations for species, link, immunogenicity and interaction"},{"label":"Splits","value":"Taxonomy, immunogenicity and glycosylation tasks use motif-cluster partitions. Protein–glycan interaction uses MMseqs2 protein clusters with threshold 0.5. Both allocate clusters 8:1:1; the two notions of held-out data are different.","status":"source_checked","source_ids":["evidence-discovery-final-glycanml"],"source_locator":"Sections 3.1–3.4"},{"label":"Metrics","value":"Checked single-task configurations use accuracy/MCC for taxonomy and glycosylation type, AUROC/AUPRC for immunogenicity, and MAE/RMSE/Spearman for protein–glycan interaction regression.","status":"source_checked","source_ids":["src-discovery-glycanml-glycanml","evidence-benchmark-glycanml-bert-species-bert-yaml","evidence-benchmark-glycanml-bert-immunogenicity-bert-yaml","evidence-benchmark-glycanml-bert-link-bert-yaml","evidence-benchmark-glycanml-bert-interaction-bert-yaml"],"source_locator":"Pinned README: Introduction; Experiment configurations; leaderboard; pinned single-task BERT configurations for species, link, immunogenicity and interaction"},{"label":"Baselines","value":"Sequence-model CNN, ResNet, LSTM and BERT configurations; graph-model GCN, RGCN, GAT, GIN, CompGCN and MPNN configurations.","status":"source_checked","source_ids":["src-discovery-glycanml-glycanml","evidence-benchmark-glycanml-bert-species-bert-yaml","evidence-benchmark-glycanml-bert-immunogenicity-bert-yaml","evidence-benchmark-glycanml-bert-link-bert-yaml","evidence-benchmark-glycanml-bert-interaction-bert-yaml"],"source_locator":"Pinned README: Introduction; Experiment configurations; leaderboard; pinned single-task BERT configurations for species, link, immunogenicity and interaction"},{"label":"Leakage controls","value":"Motif-based cluster separation tests transfer to structurally different glycans. This is a glycan-structure control, not a claim that all organisms or source studies are held out.","status":"source_checked","source_ids":["evidence-discovery-final-glycanml"],"source_locator":"Sections 3.1–3.3; Table 1"},{"label":"Uncertainty","value":"Every experiment uses seeds 0, 1 and 2; reported summaries are the mean and standard deviation over those three runs.","status":"source_checked","source_ids":["evidence-discovery-final-glycanml"],"source_locator":"Sections 3.1–3.4 and 5.1; Tables 1 and 3"},{"label":"Entity type","value":"Glycan representation benchmark suite.","status":"source_checked","source_ids":["src-discovery-glycanml-glycanml","evidence-benchmark-glycanml-bert-species-bert-yaml","evidence-benchmark-glycanml-bert-immunogenicity-bert-yaml","evidence-benchmark-glycanml-bert-link-bert-yaml","evidence-benchmark-glycanml-bert-interaction-bert-yaml"],"source_locator":"Pinned README: Introduction; Experiment configurations; leaderboard; pinned single-task BERT configurations for species, link, immunogenicity and interaction"},{"label":"Organisms","value":"Taxonomy tasks explicitly predict organism categories; species scope depends on the constituent dataset.","status":"source_checked","source_ids":["src-discovery-glycanml-glycanml","evidence-benchmark-glycanml-bert-species-bert-yaml","evidence-benchmark-glycanml-bert-immunogenicity-bert-yaml","evidence-benchmark-glycanml-bert-link-bert-yaml","evidence-benchmark-glycanml-bert-interaction-bert-yaml"],"source_locator":"Pinned README: Introduction; Experiment configurations; leaderboard; pinned single-task BERT configurations for species, link, immunogenicity and interaction"},{"label":"Assays","value":"Taxonomy, immunogenicity, glycosylation-type and protein–glycan interaction annotations.","status":"source_checked","source_ids":["src-discovery-glycanml-glycanml","evidence-benchmark-glycanml-bert-species-bert-yaml","evidence-benchmark-glycanml-bert-immunogenicity-bert-yaml","evidence-benchmark-glycanml-bert-link-bert-yaml","evidence-benchmark-glycanml-bert-interaction-bert-yaml"],"source_locator":"Pinned README: Introduction; Experiment configurations; leaderboard; pinned single-task BERT configurations for species, link, immunogenicity and interaction"},{"label":"Allowed inputs","value":"Glycan sequence or graph representations and, for interaction tasks, paired protein data.","status":"source_checked","source_ids":["src-discovery-glycanml-glycanml","evidence-benchmark-glycanml-bert-species-bert-yaml","evidence-benchmark-glycanml-bert-immunogenicity-bert-yaml","evidence-benchmark-glycanml-bert-link-bert-yaml","evidence-benchmark-glycanml-bert-interaction-bert-yaml"],"source_locator":"Pinned README: Introduction; Experiment configurations; leaderboard; pinned single-task BERT configurations for species, link, immunogenicity and interaction"},{"label":"Adaptation","value":"Separate single-task and multi-task training configurations are supplied.","status":"source_checked","source_ids":["src-discovery-glycanml-glycanml","evidence-benchmark-glycanml-bert-species-bert-yaml","evidence-benchmark-glycanml-bert-immunogenicity-bert-yaml","evidence-benchmark-glycanml-bert-link-bert-yaml","evidence-benchmark-glycanml-bert-interaction-bert-yaml"],"source_locator":"Pinned README: Introduction; Experiment configurations; leaderboard; pinned single-task BERT configurations for species, link, immunogenicity and interaction"}],"strengths":[{"text":"Shared configurations allow controlled comparisons between single-task and multi-task learning.","source_ids":["src-discovery-glycanml-glycanml","evidence-benchmark-glycanml-bert-species-bert-yaml","evidence-benchmark-glycanml-bert-immunogenicity-bert-yaml","evidence-benchmark-glycanml-bert-link-bert-yaml","evidence-benchmark-glycanml-bert-interaction-bert-yaml"],"source_locator":"Pinned README: Introduction; Experiment configurations; leaderboard; pinned single-task BERT configurations for species, link, immunogenicity and interaction"}],"limitations":[{"text":"Taxonomy, immunogenicity and glycosylation tasks hold out glycan motif clusters; interaction prediction holds out protein clusters. These boundaries do not imply that both proteins and glycans are unseen in the interaction task.","source_ids":["evidence-discovery-final-glycanml"],"source_locator":"Sections 3.1–3.4 and 5.1; Tables 1 and 3"}],"diagram":{"title":"Evaluation procedure","steps":["Allowed inputs: Glycan sequence or graph representations and, for interaction tasks, paired protein data.","Splits: Taxonomy, immunogenicity and glycosylation tasks use motif-cluster partitions. Protein–glycan interaction uses MMseqs2 protein clusters with threshold 0.5. Both allocate clusters 8:1:1; the two notions of held-out data are different.","Metrics: Checked single-task configurations use accuracy/MCC for taxonomy and glycosylation type, AUROC/AUPRC for immunogenicity, and MAE/RMSE/Spearman for protein–glycan interaction regression."],"caption":"Conceptual procedure. Task variants and protocol versions retain their separate scoring conditions.","source_ids":["src-discovery-glycanml-glycanml","evidence-benchmark-glycanml-bert-species-bert-yaml","evidence-benchmark-glycanml-bert-immunogenicity-bert-yaml","evidence-benchmark-glycanml-bert-link-bert-yaml","evidence-benchmark-glycanml-bert-interaction-bert-yaml","evidence-discovery-final-glycanml"],"source_locator":"Pinned README: Introduction; Experiment configurations; leaderboard; pinned single-task BERT configurations for species, link, immunogenicity and interaction; Sections 3.1–3.4"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction."}}},"description":"Glycan properties, taxonomy and molecular interactions","facets":{"areas":["glycomics"]},"id":"discovery-benchmark-glycanml","kind":"benchmark","links":[],"name":"GlycanML","source_ids":["src-discovery-glycanml-glycanml"],"status":"discovered"}
{"attributes":{"entity_level":"task","scope_note":"Specialist molecular or omics evaluation; protocol details require review before numerical comparison.","task":"glycosylation type prediction","version":null,"historical_missing_metadata":{"dataset_release":"unextracted","metric_implementation":"unextracted","split_manifest":"unextracted","version":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"This GlycanML task evaluates glycosylation-type classification using glycan representations.","summary_source_ids":["src-discovery-glycanml-glycanml","evidence-benchmark-glycanml-bert-link-bert-yaml","evidence-benchmark-glycan-link-dataset"],"summary_source_locator":"Pinned README: Introduction; Experiment configurations; leaderboard; pinned configs/single_task/BERT/link_BERT.yaml, module/custom_datasets/glycan_link.py (task metric, dataset class and split methods)","sections":[{"title":"Evaluation methodology","body":"N-linked, O-linked and free glycan classes. Dataset loader retains the train/validation/test assignments supplied in the downloaded CSV; this inspection does not establish how those original assignments were constructed. Accuracy and Matthews correlation coefficient in the checked three-class configuration. Sequence-model CNN, ResNet, LSTM and BERT configurations; graph-model GCN, RGCN, GAT, GIN, CompGCN and MPNN configurations.","source_ids":["src-discovery-glycanml-glycanml","evidence-benchmark-glycanml-bert-link-bert-yaml","evidence-benchmark-glycan-link-dataset"],"source_locator":"Pinned README: Introduction; Experiment configurations; leaderboard; pinned configs/single_task/BERT/link_BERT.yaml, module/custom_datasets/glycan_link.py (task metric, dataset class and split methods)"}],"facts":[{"label":"Datasets","value":"N-linked, O-linked and free glycan classes.","status":"source_checked","source_ids":["src-discovery-glycanml-glycanml","evidence-benchmark-glycanml-bert-link-bert-yaml","evidence-benchmark-glycan-link-dataset"],"source_locator":"Pinned README: Introduction; Experiment configurations; leaderboard; pinned configs/single_task/BERT/link_BERT.yaml, module/custom_datasets/glycan_link.py (task metric, dataset class and split methods)"},{"label":"Splits","value":"Glycans are represented by motif frequencies and clustered; motif groups are allocated to training, validation and test in an 8:1:1 grouping scheme. Table 1 preserves the resulting dataset-specific counts.","status":"source_checked","source_ids":["evidence-discovery-final-glycanml"],"source_locator":"Sections 3.1–3.3; Table 1"},{"label":"Metrics","value":"Accuracy and Matthews correlation coefficient in the checked three-class configuration.","status":"source_checked","source_ids":["src-discovery-glycanml-glycanml","evidence-benchmark-glycanml-bert-link-bert-yaml","evidence-benchmark-glycan-link-dataset"],"source_locator":"Pinned README: Introduction; Experiment configurations; leaderboard; pinned configs/single_task/BERT/link_BERT.yaml, module/custom_datasets/glycan_link.py (task metric, dataset class and split methods)"},{"label":"Baselines","value":"Sequence-model CNN, ResNet, LSTM and BERT configurations; graph-model GCN, RGCN, GAT, GIN, CompGCN and MPNN configurations.","status":"source_checked","source_ids":["src-discovery-glycanml-glycanml","evidence-benchmark-glycanml-bert-link-bert-yaml","evidence-benchmark-glycan-link-dataset"],"source_locator":"Pinned README: Introduction; Experiment configurations; leaderboard; pinned configs/single_task/BERT/link_BERT.yaml, module/custom_datasets/glycan_link.py (task metric, dataset class and split methods)"},{"label":"Leakage controls","value":"Motif-based cluster separation tests transfer to structurally different glycans. This is a glycan-structure control, not a claim that all organisms or source studies are held out.","status":"source_checked","source_ids":["evidence-discovery-final-glycanml"],"source_locator":"Sections 3.1–3.3; Table 1"},{"label":"Uncertainty","value":"Every experiment uses seeds 0, 1 and 2; reported summaries are the mean and standard deviation over those three runs.","status":"source_checked","source_ids":["evidence-discovery-final-glycanml"],"source_locator":"Sections 3.1–3.4 and 5.1; Tables 1 and 3"},{"label":"Entity type","value":"Constituent benchmark task: GlycanML glycosylation type prediction","status":"source_checked","source_ids":["src-discovery-glycanml-glycanml","evidence-benchmark-glycanml-bert-link-bert-yaml","evidence-benchmark-glycan-link-dataset"],"source_locator":"Pinned README: Introduction; Experiment configurations; leaderboard; pinned configs/single_task/BERT/link_BERT.yaml, module/custom_datasets/glycan_link.py (task metric, dataset class and split methods)"},{"label":"Organisms","value":"Taxonomy tasks explicitly predict organism categories; species scope depends on the constituent dataset.","status":"source_checked","source_ids":["src-discovery-glycanml-glycanml","evidence-benchmark-glycanml-bert-link-bert-yaml","evidence-benchmark-glycan-link-dataset"],"source_locator":"Pinned README: Introduction; Experiment configurations; leaderboard; pinned configs/single_task/BERT/link_BERT.yaml, module/custom_datasets/glycan_link.py (task metric, dataset class and split methods)"},{"label":"Assays","value":"Taxonomy, immunogenicity, glycosylation-type and protein–glycan interaction annotations.","status":"source_checked","source_ids":["src-discovery-glycanml-glycanml","evidence-benchmark-glycanml-bert-link-bert-yaml","evidence-benchmark-glycan-link-dataset"],"source_locator":"Pinned README: Introduction; Experiment configurations; leaderboard; pinned configs/single_task/BERT/link_BERT.yaml, module/custom_datasets/glycan_link.py (task metric, dataset class and split methods)"},{"label":"Allowed inputs","value":"Glycan sequence or graph representation with a glycosylation-type target.","status":"source_checked","source_ids":["src-discovery-glycanml-glycanml","evidence-benchmark-glycanml-bert-link-bert-yaml","evidence-benchmark-glycan-link-dataset"],"source_locator":"Pinned README: Introduction; Experiment configurations; leaderboard; pinned configs/single_task/BERT/link_BERT.yaml, module/custom_datasets/glycan_link.py (task metric, dataset class and split methods)"},{"label":"Adaptation","value":"Separate single-task and multi-task training configurations are supplied.","status":"source_checked","source_ids":["src-discovery-glycanml-glycanml","evidence-benchmark-glycanml-bert-link-bert-yaml","evidence-benchmark-glycan-link-dataset"],"source_locator":"Pinned README: Introduction; Experiment configurations; leaderboard; pinned configs/single_task/BERT/link_BERT.yaml, module/custom_datasets/glycan_link.py (task metric, dataset class and split methods)"}],"strengths":[{"text":"Shared configurations allow controlled comparisons between single-task and multi-task learning.","source_ids":["src-discovery-glycanml-glycanml","evidence-benchmark-glycanml-bert-link-bert-yaml","evidence-benchmark-glycan-link-dataset"],"source_locator":"Pinned README: Introduction; Experiment configurations; leaderboard; pinned configs/single_task/BERT/link_BERT.yaml, module/custom_datasets/glycan_link.py (task metric, dataset class and split methods)"}],"limitations":[{"text":"Taxonomy, immunogenicity and glycosylation tasks hold out glycan motif clusters; interaction prediction holds out protein clusters. These boundaries do not imply that both proteins and glycans are unseen in the interaction task.","source_ids":["evidence-discovery-final-glycanml"],"source_locator":"Sections 3.1–3.4 and 5.1; Tables 1 and 3"}],"diagram":{"title":"Evaluation procedure","steps":["Allowed inputs: Glycan sequence or graph representation with a glycosylation-type target.","Splits: Glycans are represented by motif frequencies and clustered; motif groups are allocated to training, validation and test in an 8:1:1 grouping scheme. Table 1 preserves the resulting dataset-specific counts.","Metrics: Accuracy and Matthews correlation coefficient in the checked three-class configuration."],"caption":"Conceptual procedure. Task variants and protocol versions retain their separate scoring conditions.","source_ids":["src-discovery-glycanml-glycanml","evidence-benchmark-glycanml-bert-link-bert-yaml","evidence-benchmark-glycan-link-dataset","evidence-discovery-final-glycanml"],"source_locator":"Pinned README: Introduction; Experiment configurations; leaderboard; pinned configs/single_task/BERT/link_BERT.yaml, module/custom_datasets/glycan_link.py (task metric, dataset class and split methods); Sections 3.1–3.3; Table 1"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction."}}},"description":"glycosylation type prediction","facets":{"areas":["glycomics"]},"id":"discovery-benchmark-glycanml-glycosylation-type-prediction","kind":"benchmark","links":[{"relation":"parent","target_id":"discovery-benchmark-glycanml"},{"relation":"part_of","target_id":"discovery-benchmark-glycanml"}],"name":"GlycanML glycosylation type prediction","source_ids":["src-discovery-glycanml-glycanml"],"status":"discovered"}
{"attributes":{"entity_level":"task","scope_note":"Specialist molecular or omics evaluation; protocol details require review before numerical comparison.","task":"immunogenicity prediction","version":null,"historical_missing_metadata":{"dataset_release":"unextracted","metric_implementation":"unextracted","split_manifest":"unextracted","version":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"This GlycanML task evaluates immunogenicity classification using glycan representations.","summary_source_ids":["src-discovery-glycanml-glycanml","evidence-benchmark-glycanml-bert-immunogenicity-bert-yaml","evidence-benchmark-glycan-immunogenicity-dataset"],"summary_source_locator":"Pinned README: Introduction; Experiment configurations; leaderboard; pinned configs/single_task/BERT/immunogenicity_BERT.yaml, module/custom_datasets/glycan_immunogenicity.py (task metric, dataset class and split methods)","sections":[{"title":"Evaluation methodology","body":"Binary glycan immunogenicity annotations. Dataset loader retains the train/validation/test assignments supplied in the downloaded CSV; this inspection does not establish how those original assignments were constructed. AUROC and AUPRC in the checked binary-classification configuration. Sequence-model CNN, ResNet, LSTM and BERT configurations; graph-model GCN, RGCN, GAT, GIN, CompGCN and MPNN configurations.","source_ids":["src-discovery-glycanml-glycanml","evidence-benchmark-glycanml-bert-immunogenicity-bert-yaml","evidence-benchmark-glycan-immunogenicity-dataset"],"source_locator":"Pinned README: Introduction; Experiment configurations; leaderboard; pinned configs/single_task/BERT/immunogenicity_BERT.yaml, module/custom_datasets/glycan_immunogenicity.py (task metric, dataset class and split methods)"}],"facts":[{"label":"Datasets","value":"Binary glycan immunogenicity annotations.","status":"source_checked","source_ids":["src-discovery-glycanml-glycanml","evidence-benchmark-glycanml-bert-immunogenicity-bert-yaml","evidence-benchmark-glycan-immunogenicity-dataset"],"source_locator":"Pinned README: Introduction; Experiment configurations; leaderboard; pinned configs/single_task/BERT/immunogenicity_BERT.yaml, module/custom_datasets/glycan_immunogenicity.py (task metric, dataset class and split methods)"},{"label":"Splits","value":"Glycans are represented by motif frequencies and clustered; motif groups are allocated to training, validation and test in an 8:1:1 grouping scheme. Table 1 preserves the resulting dataset-specific counts.","status":"source_checked","source_ids":["evidence-discovery-final-glycanml"],"source_locator":"Sections 3.1–3.3; Table 1"},{"label":"Metrics","value":"AUROC and AUPRC in the checked binary-classification configuration.","status":"source_checked","source_ids":["src-discovery-glycanml-glycanml","evidence-benchmark-glycanml-bert-immunogenicity-bert-yaml","evidence-benchmark-glycan-immunogenicity-dataset"],"source_locator":"Pinned README: Introduction; Experiment configurations; leaderboard; pinned configs/single_task/BERT/immunogenicity_BERT.yaml, module/custom_datasets/glycan_immunogenicity.py (task metric, dataset class and split methods)"},{"label":"Baselines","value":"Sequence-model CNN, ResNet, LSTM and BERT configurations; graph-model GCN, RGCN, GAT, GIN, CompGCN and MPNN configurations.","status":"source_checked","source_ids":["src-discovery-glycanml-glycanml","evidence-benchmark-glycanml-bert-immunogenicity-bert-yaml","evidence-benchmark-glycan-immunogenicity-dataset"],"source_locator":"Pinned README: Introduction; Experiment configurations; leaderboard; pinned configs/single_task/BERT/immunogenicity_BERT.yaml, module/custom_datasets/glycan_immunogenicity.py (task metric, dataset class and split methods)"},{"label":"Leakage controls","value":"Motif-based cluster separation tests transfer to structurally different glycans. This is a glycan-structure control, not a claim that all organisms or source studies are held out.","status":"source_checked","source_ids":["evidence-discovery-final-glycanml"],"source_locator":"Sections 3.1–3.3; Table 1"},{"label":"Uncertainty","value":"Every experiment uses seeds 0, 1 and 2; reported summaries are the mean and standard deviation over those three runs.","status":"source_checked","source_ids":["evidence-discovery-final-glycanml"],"source_locator":"Sections 3.1–3.4 and 5.1; Tables 1 and 3"},{"label":"Entity type","value":"Constituent benchmark task: GlycanML immunogenicity prediction","status":"source_checked","source_ids":["src-discovery-glycanml-glycanml","evidence-benchmark-glycanml-bert-immunogenicity-bert-yaml","evidence-benchmark-glycan-immunogenicity-dataset"],"source_locator":"Pinned README: Introduction; Experiment configurations; leaderboard; pinned configs/single_task/BERT/immunogenicity_BERT.yaml, module/custom_datasets/glycan_immunogenicity.py (task metric, dataset class and split methods)"},{"label":"Organisms","value":"Taxonomy tasks explicitly predict organism categories; species scope depends on the constituent dataset.","status":"source_checked","source_ids":["src-discovery-glycanml-glycanml","evidence-benchmark-glycanml-bert-immunogenicity-bert-yaml","evidence-benchmark-glycan-immunogenicity-dataset"],"source_locator":"Pinned README: Introduction; Experiment configurations; leaderboard; pinned configs/single_task/BERT/immunogenicity_BERT.yaml, module/custom_datasets/glycan_immunogenicity.py (task metric, dataset class and split methods)"},{"label":"Assays","value":"Taxonomy, immunogenicity, glycosylation-type and protein–glycan interaction annotations.","status":"source_checked","source_ids":["src-discovery-glycanml-glycanml","evidence-benchmark-glycanml-bert-immunogenicity-bert-yaml","evidence-benchmark-glycan-immunogenicity-dataset"],"source_locator":"Pinned README: Introduction; Experiment configurations; leaderboard; pinned configs/single_task/BERT/immunogenicity_BERT.yaml, module/custom_datasets/glycan_immunogenicity.py (task metric, dataset class and split methods)"},{"label":"Allowed inputs","value":"Glycan sequence or graph representation with an immunogenicity target.","status":"source_checked","source_ids":["src-discovery-glycanml-glycanml","evidence-benchmark-glycanml-bert-immunogenicity-bert-yaml","evidence-benchmark-glycan-immunogenicity-dataset"],"source_locator":"Pinned README: Introduction; Experiment configurations; leaderboard; pinned configs/single_task/BERT/immunogenicity_BERT.yaml, module/custom_datasets/glycan_immunogenicity.py (task metric, dataset class and split methods)"},{"label":"Adaptation","value":"Separate single-task and multi-task training configurations are supplied.","status":"source_checked","source_ids":["src-discovery-glycanml-glycanml","evidence-benchmark-glycanml-bert-immunogenicity-bert-yaml","evidence-benchmark-glycan-immunogenicity-dataset"],"source_locator":"Pinned README: Introduction; Experiment configurations; leaderboard; pinned configs/single_task/BERT/immunogenicity_BERT.yaml, module/custom_datasets/glycan_immunogenicity.py (task metric, dataset class and split methods)"}],"strengths":[{"text":"Shared configurations allow controlled comparisons between single-task and multi-task learning.","source_ids":["src-discovery-glycanml-glycanml","evidence-benchmark-glycanml-bert-immunogenicity-bert-yaml","evidence-benchmark-glycan-immunogenicity-dataset"],"source_locator":"Pinned README: Introduction; Experiment configurations; leaderboard; pinned configs/single_task/BERT/immunogenicity_BERT.yaml, module/custom_datasets/glycan_immunogenicity.py (task metric, dataset class and split methods)"}],"limitations":[{"text":"Taxonomy, immunogenicity and glycosylation tasks hold out glycan motif clusters; interaction prediction holds out protein clusters. These boundaries do not imply that both proteins and glycans are unseen in the interaction task.","source_ids":["evidence-discovery-final-glycanml"],"source_locator":"Sections 3.1–3.4 and 5.1; Tables 1 and 3"}],"diagram":{"title":"Evaluation procedure","steps":["Allowed inputs: Glycan sequence or graph representation with an immunogenicity target.","Splits: Glycans are represented by motif frequencies and clustered; motif groups are allocated to training, validation and test in an 8:1:1 grouping scheme. Table 1 preserves the resulting dataset-specific counts.","Metrics: AUROC and AUPRC in the checked binary-classification configuration."],"caption":"Conceptual procedure. Task variants and protocol versions retain their separate scoring conditions.","source_ids":["src-discovery-glycanml-glycanml","evidence-benchmark-glycanml-bert-immunogenicity-bert-yaml","evidence-benchmark-glycan-immunogenicity-dataset","evidence-discovery-final-glycanml"],"source_locator":"Pinned README: Introduction; Experiment configurations; leaderboard; pinned configs/single_task/BERT/immunogenicity_BERT.yaml, module/custom_datasets/glycan_immunogenicity.py (task metric, dataset class and split methods); Sections 3.1–3.3; Table 1"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction."}}},"description":"immunogenicity prediction","facets":{"areas":["glycomics"]},"id":"discovery-benchmark-glycanml-immunogenicity-prediction","kind":"benchmark","links":[{"relation":"parent","target_id":"discovery-benchmark-glycanml"},{"relation":"part_of","target_id":"discovery-benchmark-glycanml"}],"name":"GlycanML immunogenicity prediction","source_ids":["src-discovery-glycanml-glycanml"],"status":"discovered"}
{"attributes":{"entity_level":"task","scope_note":"Specialist molecular or omics evaluation; protocol details require review before numerical comparison.","task":"protein-glycan interaction prediction","version":null,"historical_missing_metadata":{"dataset_release":"unextracted","metric_implementation":"unextracted","split_manifest":"unextracted","version":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"This GlycanML task evaluates protein–glycan interaction prediction using glycan representations.","summary_source_ids":["src-discovery-glycanml-glycanml","evidence-benchmark-glycanml-bert-interaction-bert-yaml","evidence-benchmark-glycan-interaction-dataset"],"summary_source_locator":"Pinned README: Introduction; Experiment configurations; leaderboard; pinned configs/single_task/BERT/interaction_BERT.yaml, module/custom_datasets/glycan_interaction.py (task metric, dataset class and split methods)","sections":[{"title":"Evaluation methodology","body":"Protein–glycan interaction targets; the checked task is regression, not binary classification. Dataset loader retains the train/validation/test assignments supplied in the downloaded CSV; this inspection does not establish how those original assignments were constructed. MAE, RMSE and Spearman correlation in the checked interaction-regression configuration. Sequence-model CNN, ResNet, LSTM and BERT configurations; graph-model GCN, RGCN, GAT, GIN, CompGCN and MPNN configurations.","source_ids":["src-discovery-glycanml-glycanml","evidence-benchmark-glycanml-bert-interaction-bert-yaml","evidence-benchmark-glycan-interaction-dataset"],"source_locator":"Pinned README: Introduction; Experiment configurations; leaderboard; pinned configs/single_task/BERT/interaction_BERT.yaml, module/custom_datasets/glycan_interaction.py (task metric, dataset class and split methods)"}],"facts":[{"label":"Datasets","value":"Protein–glycan interaction targets; the checked task is regression, not binary classification.","status":"source_checked","source_ids":["src-discovery-glycanml-glycanml","evidence-benchmark-glycanml-bert-interaction-bert-yaml","evidence-benchmark-glycan-interaction-dataset"],"source_locator":"Pinned README: Introduction; Experiment configurations; leaderboard; pinned configs/single_task/BERT/interaction_BERT.yaml, module/custom_datasets/glycan_interaction.py (task metric, dataset class and split methods)"},{"label":"Splits","value":"Dataset loader retains the train/validation/test assignments supplied in the downloaded CSV; this inspection does not establish how those original assignments were constructed.","status":"source_checked","source_ids":["src-discovery-glycanml-glycanml","evidence-benchmark-glycanml-bert-interaction-bert-yaml","evidence-benchmark-glycan-interaction-dataset"],"source_locator":"Pinned README: Introduction; Experiment configurations; leaderboard; pinned configs/single_task/BERT/interaction_BERT.yaml, module/custom_datasets/glycan_interaction.py (task metric, dataset class and split methods)"},{"label":"Metrics","value":"MAE, RMSE and Spearman correlation in the checked interaction-regression configuration.","status":"source_checked","source_ids":["src-discovery-glycanml-glycanml","evidence-benchmark-glycanml-bert-interaction-bert-yaml","evidence-benchmark-glycan-interaction-dataset"],"source_locator":"Pinned README: Introduction; Experiment configurations; leaderboard; pinned configs/single_task/BERT/interaction_BERT.yaml, module/custom_datasets/glycan_interaction.py (task metric, dataset class and split methods)"},{"label":"Baselines","value":"Sequence-model CNN, ResNet, LSTM and BERT configurations; graph-model GCN, RGCN, GAT, GIN, CompGCN and MPNN configurations.","status":"source_checked","source_ids":["src-discovery-glycanml-glycanml","evidence-benchmark-glycanml-bert-interaction-bert-yaml","evidence-benchmark-glycan-interaction-dataset"],"source_locator":"Pinned README: Introduction; Experiment configurations; leaderboard; pinned configs/single_task/BERT/interaction_BERT.yaml, module/custom_datasets/glycan_interaction.py (task metric, dataset class and split methods)"},{"label":"Leakage controls","value":"MMseqs2 groups proteins at minimum within-cluster sequence identity 0.5; complete protein clusters are allocated 8:1:1, and protein–glycan pairs follow the protein split. The protocol targets unseen proteins, not necessarily unseen glycans.","status":"source_checked","source_ids":["evidence-discovery-final-glycanml"],"source_locator":"Section 3.4"},{"label":"Uncertainty","value":"Every experiment uses seeds 0, 1 and 2; reported summaries are the mean and standard deviation over those three runs.","status":"source_checked","source_ids":["evidence-discovery-final-glycanml"],"source_locator":"Sections 3.1–3.4 and 5.1; Tables 1 and 3"},{"label":"Entity type","value":"Constituent benchmark task: GlycanML protein-glycan interaction prediction","status":"source_checked","source_ids":["src-discovery-glycanml-glycanml","evidence-benchmark-glycanml-bert-interaction-bert-yaml","evidence-benchmark-glycan-interaction-dataset"],"source_locator":"Pinned README: Introduction; Experiment configurations; leaderboard; pinned configs/single_task/BERT/interaction_BERT.yaml, module/custom_datasets/glycan_interaction.py (task metric, dataset class and split methods)"},{"label":"Organisms","value":"Taxonomy tasks explicitly predict organism categories; species scope depends on the constituent dataset.","status":"source_checked","source_ids":["src-discovery-glycanml-glycanml","evidence-benchmark-glycanml-bert-interaction-bert-yaml","evidence-benchmark-glycan-interaction-dataset"],"source_locator":"Pinned README: Introduction; Experiment configurations; leaderboard; pinned configs/single_task/BERT/interaction_BERT.yaml, module/custom_datasets/glycan_interaction.py (task metric, dataset class and split methods)"},{"label":"Assays","value":"Taxonomy, immunogenicity, glycosylation-type and protein–glycan interaction annotations.","status":"source_checked","source_ids":["src-discovery-glycanml-glycanml","evidence-benchmark-glycanml-bert-interaction-bert-yaml","evidence-benchmark-glycan-interaction-dataset"],"source_locator":"Pinned README: Introduction; Experiment configurations; leaderboard; pinned configs/single_task/BERT/interaction_BERT.yaml, module/custom_datasets/glycan_interaction.py (task metric, dataset class and split methods)"},{"label":"Allowed inputs","value":"Protein–glycan pairs with interaction targets.","status":"source_checked","source_ids":["src-discovery-glycanml-glycanml","evidence-benchmark-glycanml-bert-interaction-bert-yaml","evidence-benchmark-glycan-interaction-dataset"],"source_locator":"Pinned README: Introduction; Experiment configurations; leaderboard; pinned configs/single_task/BERT/interaction_BERT.yaml, module/custom_datasets/glycan_interaction.py (task metric, dataset class and split methods)"},{"label":"Adaptation","value":"Separate single-task and multi-task training configurations are supplied.","status":"source_checked","source_ids":["src-discovery-glycanml-glycanml","evidence-benchmark-glycanml-bert-interaction-bert-yaml","evidence-benchmark-glycan-interaction-dataset"],"source_locator":"Pinned README: Introduction; Experiment configurations; leaderboard; pinned configs/single_task/BERT/interaction_BERT.yaml, module/custom_datasets/glycan_interaction.py (task metric, dataset class and split methods)"}],"strengths":[{"text":"Shared configurations allow controlled comparisons between single-task and multi-task learning.","source_ids":["src-discovery-glycanml-glycanml","evidence-benchmark-glycanml-bert-interaction-bert-yaml","evidence-benchmark-glycan-interaction-dataset"],"source_locator":"Pinned README: Introduction; Experiment configurations; leaderboard; pinned configs/single_task/BERT/interaction_BERT.yaml, module/custom_datasets/glycan_interaction.py (task metric, dataset class and split methods)"}],"limitations":[{"text":"Taxonomy, immunogenicity and glycosylation tasks hold out glycan motif clusters; interaction prediction holds out protein clusters. These boundaries do not imply that both proteins and glycans are unseen in the interaction task.","source_ids":["evidence-discovery-final-glycanml"],"source_locator":"Sections 3.1–3.4 and 5.1; Tables 1 and 3"}],"diagram":{"title":"Evaluation procedure","steps":["Allowed inputs: Protein–glycan pairs with interaction targets.","Splits: Dataset loader retains the train/validation/test assignments supplied in the downloaded CSV; this inspection does not establish how those original assignments were constructed.","Metrics: MAE, RMSE and Spearman correlation in the checked interaction-regression configuration."],"caption":"Conceptual procedure. Task variants and protocol versions retain their separate scoring conditions.","source_ids":["src-discovery-glycanml-glycanml","evidence-benchmark-glycanml-bert-interaction-bert-yaml","evidence-benchmark-glycan-interaction-dataset"],"source_locator":"Pinned README: Introduction; Experiment configurations; leaderboard; pinned configs/single_task/BERT/interaction_BERT.yaml, module/custom_datasets/glycan_interaction.py (task metric, dataset class and split methods)"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction."}}},"description":"protein-glycan interaction prediction","facets":{"areas":["glycomics"]},"id":"discovery-benchmark-glycanml-protein-glycan-interaction-prediction","kind":"benchmark","links":[{"relation":"parent","target_id":"discovery-benchmark-glycanml"},{"relation":"part_of","target_id":"discovery-benchmark-glycanml"}],"name":"GlycanML protein-glycan interaction prediction","source_ids":["src-discovery-glycanml-glycanml"],"status":"discovered"}
{"attributes":{"entity_level":"task","scope_note":"Specialist molecular or omics evaluation; protocol details require review before numerical comparison.","task":"taxonomy prediction","version":null,"historical_missing_metadata":{"dataset_release":"unextracted","metric_implementation":"unextracted","split_manifest":"unextracted","version":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"This GlycanML task evaluates taxonomy classification using glycan representations.","summary_source_ids":["src-discovery-glycanml-glycanml","evidence-benchmark-glycanml-bert-species-bert-yaml","evidence-benchmark-glycan-classification-dataset"],"summary_source_locator":"Pinned README: Introduction; Experiment configurations; leaderboard; pinned configs/single_task/BERT/species_BERT.yaml, module/custom_datasets/glycan_classification.py (task metric, dataset class and split methods)","sections":[{"title":"Evaluation methodology","body":"Hierarchical taxonomy labels span species, genus, family, order, class, phylum, kingdom and domain; the checked single-task configuration targets species. Dataset loader retains the train/validation/test assignments supplied in the downloaded CSV; this inspection does not establish how those original assignments were constructed. Accuracy and Matthews correlation coefficient in the checked species-classification configuration. Sequence-model CNN, ResNet, LSTM and BERT configurations; graph-model GCN, RGCN, GAT, GIN, CompGCN and MPNN configurations.","source_ids":["src-discovery-glycanml-glycanml","evidence-benchmark-glycanml-bert-species-bert-yaml","evidence-benchmark-glycan-classification-dataset"],"source_locator":"Pinned README: Introduction; Experiment configurations; leaderboard; pinned configs/single_task/BERT/species_BERT.yaml, module/custom_datasets/glycan_classification.py (task metric, dataset class and split methods)"}],"facts":[{"label":"Datasets","value":"Hierarchical taxonomy labels span species, genus, family, order, class, phylum, kingdom and domain; the checked single-task configuration targets species.","status":"source_checked","source_ids":["src-discovery-glycanml-glycanml","evidence-benchmark-glycanml-bert-species-bert-yaml","evidence-benchmark-glycan-classification-dataset"],"source_locator":"Pinned README: Introduction; Experiment configurations; leaderboard; pinned configs/single_task/BERT/species_BERT.yaml, module/custom_datasets/glycan_classification.py (task metric, dataset class and split methods)"},{"label":"Splits","value":"Glycans are represented by motif frequencies and clustered; motif groups are allocated to training, validation and test in an 8:1:1 grouping scheme. Table 1 preserves the resulting dataset-specific counts.","status":"source_checked","source_ids":["evidence-discovery-final-glycanml"],"source_locator":"Sections 3.1–3.3; Table 1"},{"label":"Metrics","value":"Accuracy and Matthews correlation coefficient in the checked species-classification configuration.","status":"source_checked","source_ids":["src-discovery-glycanml-glycanml","evidence-benchmark-glycanml-bert-species-bert-yaml","evidence-benchmark-glycan-classification-dataset"],"source_locator":"Pinned README: Introduction; Experiment configurations; leaderboard; pinned configs/single_task/BERT/species_BERT.yaml, module/custom_datasets/glycan_classification.py (task metric, dataset class and split methods)"},{"label":"Baselines","value":"Sequence-model CNN, ResNet, LSTM and BERT configurations; graph-model GCN, RGCN, GAT, GIN, CompGCN and MPNN configurations.","status":"source_checked","source_ids":["src-discovery-glycanml-glycanml","evidence-benchmark-glycanml-bert-species-bert-yaml","evidence-benchmark-glycan-classification-dataset"],"source_locator":"Pinned README: Introduction; Experiment configurations; leaderboard; pinned configs/single_task/BERT/species_BERT.yaml, module/custom_datasets/glycan_classification.py (task metric, dataset class and split methods)"},{"label":"Leakage controls","value":"Motif-based cluster separation tests transfer to structurally different glycans. This is a glycan-structure control, not a claim that all organisms or source studies are held out.","status":"source_checked","source_ids":["evidence-discovery-final-glycanml"],"source_locator":"Sections 3.1–3.3; Table 1"},{"label":"Uncertainty","value":"Every experiment uses seeds 0, 1 and 2; reported summaries are the mean and standard deviation over those three runs.","status":"source_checked","source_ids":["evidence-discovery-final-glycanml"],"source_locator":"Sections 3.1–3.4 and 5.1; Tables 1 and 3"},{"label":"Entity type","value":"Constituent benchmark task: GlycanML taxonomy prediction","status":"source_checked","source_ids":["src-discovery-glycanml-glycanml","evidence-benchmark-glycanml-bert-species-bert-yaml","evidence-benchmark-glycan-classification-dataset"],"source_locator":"Pinned README: Introduction; Experiment configurations; leaderboard; pinned configs/single_task/BERT/species_BERT.yaml, module/custom_datasets/glycan_classification.py (task metric, dataset class and split methods)"},{"label":"Organisms","value":"Taxonomy tasks explicitly predict organism categories; species scope depends on the constituent dataset.","status":"source_checked","source_ids":["src-discovery-glycanml-glycanml","evidence-benchmark-glycanml-bert-species-bert-yaml","evidence-benchmark-glycan-classification-dataset"],"source_locator":"Pinned README: Introduction; Experiment configurations; leaderboard; pinned configs/single_task/BERT/species_BERT.yaml, module/custom_datasets/glycan_classification.py (task metric, dataset class and split methods)"},{"label":"Assays","value":"Taxonomy, immunogenicity, glycosylation-type and protein–glycan interaction annotations.","status":"source_checked","source_ids":["src-discovery-glycanml-glycanml","evidence-benchmark-glycanml-bert-species-bert-yaml","evidence-benchmark-glycan-classification-dataset"],"source_locator":"Pinned README: Introduction; Experiment configurations; leaderboard; pinned configs/single_task/BERT/species_BERT.yaml, module/custom_datasets/glycan_classification.py (task metric, dataset class and split methods)"},{"label":"Allowed inputs","value":"Glycan sequence or graph representation with an organism-taxonomy target.","status":"source_checked","source_ids":["src-discovery-glycanml-glycanml","evidence-benchmark-glycanml-bert-species-bert-yaml","evidence-benchmark-glycan-classification-dataset"],"source_locator":"Pinned README: Introduction; Experiment configurations; leaderboard; pinned configs/single_task/BERT/species_BERT.yaml, module/custom_datasets/glycan_classification.py (task metric, dataset class and split methods)"},{"label":"Adaptation","value":"Separate single-task and multi-task training configurations are supplied.","status":"source_checked","source_ids":["src-discovery-glycanml-glycanml","evidence-benchmark-glycanml-bert-species-bert-yaml","evidence-benchmark-glycan-classification-dataset"],"source_locator":"Pinned README: Introduction; Experiment configurations; leaderboard; pinned configs/single_task/BERT/species_BERT.yaml, module/custom_datasets/glycan_classification.py (task metric, dataset class and split methods)"}],"strengths":[{"text":"Shared configurations allow controlled comparisons between single-task and multi-task learning.","source_ids":["src-discovery-glycanml-glycanml","evidence-benchmark-glycanml-bert-species-bert-yaml","evidence-benchmark-glycan-classification-dataset"],"source_locator":"Pinned README: Introduction; Experiment configurations; leaderboard; pinned configs/single_task/BERT/species_BERT.yaml, module/custom_datasets/glycan_classification.py (task metric, dataset class and split methods)"}],"limitations":[{"text":"Taxonomy, immunogenicity and glycosylation tasks hold out glycan motif clusters; interaction prediction holds out protein clusters. These boundaries do not imply that both proteins and glycans are unseen in the interaction task.","source_ids":["evidence-discovery-final-glycanml"],"source_locator":"Sections 3.1–3.4 and 5.1; Tables 1 and 3"}],"diagram":{"title":"Evaluation procedure","steps":["Allowed inputs: Glycan sequence or graph representation with an organism-taxonomy target.","Splits: Glycans are represented by motif frequencies and clustered; motif groups are allocated to training, validation and test in an 8:1:1 grouping scheme. Table 1 preserves the resulting dataset-specific counts.","Metrics: Accuracy and Matthews correlation coefficient in the checked species-classification configuration."],"caption":"Conceptual procedure. Task variants and protocol versions retain their separate scoring conditions.","source_ids":["src-discovery-glycanml-glycanml","evidence-benchmark-glycanml-bert-species-bert-yaml","evidence-benchmark-glycan-classification-dataset","evidence-discovery-final-glycanml"],"source_locator":"Pinned README: Introduction; Experiment configurations; leaderboard; pinned configs/single_task/BERT/species_BERT.yaml, module/custom_datasets/glycan_classification.py (task metric, dataset class and split methods); Sections 3.1–3.3; Table 1"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction."}}},"description":"taxonomy prediction","facets":{"areas":["glycomics"]},"id":"discovery-benchmark-glycanml-taxonomy-prediction","kind":"benchmark","links":[{"relation":"parent","target_id":"discovery-benchmark-glycanml"},{"relation":"part_of","target_id":"discovery-benchmark-glycanml"}],"name":"GlycanML taxonomy prediction","source_ids":["src-discovery-glycanml-glycanml"],"status":"discovered"}
{"attributes":{"entity_level":"suite","scope_note":"Specialist molecular or omics evaluation; protocol details require review before numerical comparison.","task":"Multi-species genome understanding tasks","version":null,"historical_missing_metadata":{"dataset_release":"unextracted","metric_implementation":"unextracted","split_manifest":"unextracted","version":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"GUE evaluates genome understanding across multiple datasets, task types and species.","summary_source_ids":["src-discovery-magics-lab-dnabert-2"],"summary_source_locator":"Pinned README: GUE section; data download and evaluation scripts","sections":[{"title":"Evaluation methodology","body":"GUE combines genome sequence classification tasks with supplied partitions and task-specific metrics. Models are fine-tuned on labelled training examples, selected with validation data and scored on test data. Random partitions occur in the suite, so a high score does not automatically demonstrate transfer to unrelated genomes.","source_ids":["evidence-discovery-final-gue"],"source_locator":"Section 5; Appendix C and Table 9: GUE task datasets"}],"facts":[{"label":"Datasets","value":"The benchmark archive contains separate sequence-classification datasets; model pretraining data are distributed separately.","status":"source_checked","source_ids":["src-discovery-magics-lab-dnabert-2"],"source_locator":"Pinned README: GUE section; data download and evaluation scripts"},{"label":"Splits","value":"Separate train/validation/test files and counts are defined per dataset. The yeast epigenetic tasks use an 8:1:1 random split; source-derived regulatory and viral tasks retain their own documented partitions. There is no universal chromosome holdout across GUE.","status":"source_checked","source_ids":["evidence-discovery-final-gue"],"source_locator":"Appendix C: task construction; Table 9"},{"label":"Metrics","value":"GUE uses Matthews correlation coefficient for the regulatory classification tasks and F1 for COVID variant classification; the exact task metric is tabulated in the dataset appendix.","status":"source_checked","source_ids":["evidence-discovery-final-gue"],"source_locator":"Section 5; Appendix C and Table 9: GUE task datasets"},{"label":"Baselines","value":"DNABERT-2 and other genomic representation models are compared in the associated benchmark.","status":"source_checked","source_ids":["src-discovery-magics-lab-dnabert-2"],"source_locator":"Pinned README: GUE section; data download and evaluation scripts"},{"label":"Leakage controls","value":"Datasets have explicit train/validation/test partitions, but some use random splits, including yeast epigenetic marks at 8:1:1. The paper’s masked-token leakage discussion concerns tokenization and must not be confused with proof of train/test genome independence.","status":"source_checked","source_ids":["evidence-discovery-final-gue"],"source_locator":"Section 5; Appendix C and Table 9: GUE task datasets"},{"label":"Uncertainty","value":"Models are fine-tuned with three different random seeds and the mean test result is reported. The stated protocol does not define a uniform confidence interval for every dataset.","status":"source_checked","source_ids":["evidence-discovery-final-gue"],"source_locator":"Section 5; Appendix C and Table 9: GUE task datasets"},{"label":"Entity type","value":"Genome Understanding Evaluation benchmark suite.","status":"source_checked","source_ids":["src-discovery-magics-lab-dnabert-2"],"source_locator":"Pinned README: GUE section; data download and evaluation scripts"},{"label":"Organisms","value":"Multiple species; the README describes four-species coverage.","status":"source_checked","source_ids":["src-discovery-magics-lab-dnabert-2"],"source_locator":"Pinned README: GUE section; data download and evaluation scripts"},{"label":"Assays","value":"Task-specific genomic classification labels.","status":"source_checked","source_ids":["src-discovery-magics-lab-dnabert-2"],"source_locator":"Pinned README: GUE section; data download and evaluation scripts"},{"label":"Allowed inputs","value":"DNA sequences from the benchmark archive, separate from model-pretraining data.","status":"source_checked","source_ids":["src-discovery-magics-lab-dnabert-2"],"source_locator":"Pinned README: GUE section; data download and evaluation scripts"},{"label":"Adaptation","value":"Supervised fine-tuning; scripts include model-specific training examples.","status":"source_checked","source_ids":["src-discovery-magics-lab-dnabert-2"],"source_locator":"Pinned README: GUE section; data download and evaluation scripts"}],"strengths":[{"text":"The benchmark distribution is separated from the pretraining corpus.","source_ids":["src-discovery-magics-lab-dnabert-2"],"source_locator":"Pinned README: GUE section; data download and evaluation scripts"}],"limitations":[{"text":"Dataset partitions and metrics are task-specific. Non-overlapping tokenization addresses masked-token information leakage, which is a different issue from biological homology or pretraining overlap.","source_ids":["evidence-discovery-final-gue"],"source_locator":"Section 5; Appendix C and Table 9: GUE task datasets"}],"diagram":{"title":"Evaluation procedure","steps":["Allowed inputs: DNA sequences from the benchmark archive, separate from model-pretraining data.","Splits: Separate train/validation/test files and counts are defined per dataset. The yeast epigenetic tasks use an 8:1:1 random split; source-derived regulatory and viral tasks retain their own documented partitions. There is no universal chromosome holdout across GUE.","Metrics: GUE uses Matthews correlation coefficient for the regulatory classification tasks and F1 for COVID variant classification; the exact task metric is tabulated in the dataset appendix."],"caption":"Conceptual procedure. Task variants and protocol versions retain their separate scoring conditions.","source_ids":["src-discovery-magics-lab-dnabert-2","evidence-discovery-final-gue"],"source_locator":"Pinned README: GUE section; data download and evaluation scripts; Appendix C: task construction; Table 9; Section 5; Appendix C and Table 9: GUE task datasets"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction."}}},"description":"Multi-species genome understanding tasks","facets":{"areas":["genomics"]},"id":"discovery-benchmark-gue","kind":"benchmark","links":[],"name":"GUE","source_ids":["src-discovery-magics-lab-dnabert-2"],"status":"discovered"}
{"attributes":{"entity_level":"suite","scope_note":"Specialist molecular or omics evaluation; protocol details require review before numerical comparison.","task":"Gene expression prediction from matched histology","version":null,"historical_missing_metadata":{"dataset_release":"unextracted","metric_implementation":"unextracted","split_manifest":"unextracted","version":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"HEST-Benchmark tests prediction of gene expression from histological image representations.","summary_source_ids":["src-discovery-mahmoodlab-hest"],"summary_source_locator":"Pinned README: HEST-Benchmark overview; evaluation notes","sections":[{"title":"Evaluation methodology","body":"HEST-Benchmark connects histology patches to spatially measured gene expression. A patch encoder supplies features to a regression model, which predicts highly variable genes. Patient-stratified folds evaluate transfer between individuals, and correlation is summarized across those folds.","source_ids":["evidence-discovery-final-hest"],"source_locator":"Sections 5.1–5.2; Appendix Table A11"}],"facts":[{"label":"Datasets","value":"Paired spatial-transcriptomic measurements and histology images from HEST resources.","status":"source_checked","source_ids":["src-discovery-mahmoodlab-hest"],"source_locator":"Pinned README: HEST-Benchmark overview; evaluation notes"},{"label":"Splits","value":"Patient-stratified cross-validation: one fold per patient, except ccRCC uses half as many folds because of its larger patient cohort.","status":"source_checked","source_ids":["evidence-discovery-final-hest"],"source_locator":"Sections 5.1–5.2; Appendix Table A11"},{"label":"Metrics","value":"Pearson correlation between predicted and measured log1p gene expression, using the 50 genes with highest normalized variance.","status":"source_checked","source_ids":["evidence-discovery-final-hest"],"source_locator":"Sections 5.1–5.2; Appendix Table A11"},{"label":"Baselines","value":"Ridge regression on PCA-reduced embeddings is the reported comparison setup.","status":"source_checked","source_ids":["src-discovery-mahmoodlab-hest"],"source_locator":"Pinned README: HEST-Benchmark overview; evaluation notes"},{"label":"Leakage controls","value":"All samples from a patient stay within their fold, preventing patch-level mixing of the same patient between training and test. Histology-encoder pretraining overlap is a separate concern.","status":"source_checked","source_ids":["evidence-discovery-final-hest"],"source_locator":"Sections 5.1–5.2; Appendix Table A11"},{"label":"Uncertainty","value":"The paper reports the mean and standard deviation across folds or patients, not a universal retraining-seed interval.","status":"source_checked","source_ids":["evidence-discovery-final-hest"],"source_locator":"Sections 5.1–5.2; Appendix Table A11"},{"label":"Entity type","value":"Spatial-transcriptomics prediction benchmark.","status":"source_checked","source_ids":["src-discovery-mahmoodlab-hest"],"source_locator":"Pinned README: HEST-Benchmark overview; evaluation notes"},{"label":"Organisms","value":"Multiple species selectable in the HEST metadata.","status":"source_checked","source_ids":["src-discovery-mahmoodlab-hest"],"source_locator":"Pinned README: HEST-Benchmark overview; evaluation notes"},{"label":"Assays","value":"Paired tissue histology and spatial gene-expression measurements.","status":"source_checked","source_ids":["src-discovery-mahmoodlab-hest"],"source_locator":"Pinned README: HEST-Benchmark overview; evaluation notes"},{"label":"Allowed inputs","value":"Histology patches for prediction; spatial expression supplies evaluation labels.","status":"source_checked","source_ids":["src-discovery-mahmoodlab-hest"],"source_locator":"Pinned README: HEST-Benchmark overview; evaluation notes"},{"label":"Adaptation","value":"Patch embeddings are evaluated through downstream expression prediction.","status":"source_checked","source_ids":["src-discovery-mahmoodlab-hest"],"source_locator":"Pinned README: HEST-Benchmark overview; evaluation notes"}],"strengths":[{"text":"Paired measurements connect image representations to molecular rather than image-only labels.","source_ids":["src-discovery-mahmoodlab-hest"],"source_locator":"Pinned README: HEST-Benchmark overview; evaluation notes"}],"limitations":[{"text":"The input is an image, but the evaluated output is spatial molecular expression. The selected genes, regression head and patient partitions are essential comparison conditions.","source_ids":["evidence-discovery-final-hest"],"source_locator":"Sections 5.1–5.2; Appendix Table A11"}],"diagram":{"title":"Evaluation procedure","steps":["Allowed inputs: Histology patches for prediction; spatial expression supplies evaluation labels.","Splits: Patient-stratified cross-validation: one fold per patient, except ccRCC uses half as many folds because of its larger patient cohort.","Metrics: Pearson correlation between predicted and measured log1p gene expression, using the 50 genes with highest normalized variance."],"caption":"Conceptual procedure. Task variants and protocol versions retain their separate scoring conditions.","source_ids":["src-discovery-mahmoodlab-hest","evidence-discovery-final-hest"],"source_locator":"Pinned README: HEST-Benchmark overview; evaluation notes; Sections 5.1–5.2; Appendix Table A11"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction."}}},"description":"Gene expression prediction from matched histology","facets":{"areas":["spatial-omics"]},"id":"discovery-benchmark-hest-benchmark","kind":"benchmark","links":[],"name":"HEST-Benchmark","source_ids":["src-discovery-mahmoodlab-hest"],"status":"discovered"}
{"attributes":{"entity_level":"suite","scope_note":"Specialist molecular or omics evaluation; protocol details require review before numerical comparison.","task":"Molecular identification from tandem mass spectra","version":null,"historical_missing_metadata":{"dataset_release":"unextracted","metric_implementation":"unextracted","split_manifest":"unextracted","version":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"MassSpecGym separates spectrum-to-structure generation, candidate retrieval and structure-to-spectrum simulation.","summary_source_ids":["src-discovery-pluskal-lab-massspecgym","evidence-benchmark-massspecgym-de-novo-base-py","evidence-benchmark-massspecgym-retrieval-base-py","evidence-benchmark-massspecgym-simulation-base-py"],"summary_source_locator":"Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned de_novo, retrieval and simulation base evaluation classes","sections":[{"title":"Evaluation methodology","body":"The released MassSpecGym MS/MS/molecule dataset, with task-specific inputs and candidate sets. MCES molecular clusters are grouped into fixed training, validation and test folds, stratified by acquisition metadata. Cross-fold molecular bond-edit distance is at least 10; all spectra follow the assigned molecular fold. Task evaluators distinguish molecular exact match/structural similarity, candidate-retrieval hit rate and spectrum similarity. These are separate readouts, not interchangeable scores. Maximum common edge subgraph (MCES) clustering keeps molecules connected by a bond-edit distance below 10 in the same fold. The split additionally balances instrument, collision-energy, adduct and molecule-frequency metadata; this is stronger than simply separating 2D InChIKeys.","source_ids":["src-discovery-pluskal-lab-massspecgym","evidence-benchmark-massspecgym-de-novo-base-py","evidence-benchmark-massspecgym-retrieval-base-py","evidence-benchmark-massspecgym-simulation-base-py","evidence-discovery-final-massspecgym"],"source_locator":"Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned de_novo, retrieval and simulation base evaluation classes; Section 3.4; Supplementary Information 2.5; Section 3.4; Supplementary Information 2.5; Tables 2–4"}],"facts":[{"label":"Datasets","value":"The released MassSpecGym MS/MS/molecule dataset, with task-specific inputs and candidate sets.","status":"source_checked","source_ids":["src-discovery-pluskal-lab-massspecgym","evidence-benchmark-massspecgym-de-novo-base-py","evidence-benchmark-massspecgym-retrieval-base-py","evidence-benchmark-massspecgym-simulation-base-py"],"source_locator":"Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned de_novo, retrieval and simulation base evaluation classes"},{"label":"Splits","value":"MCES molecular clusters are grouped into fixed training, validation and test folds, stratified by acquisition metadata. Cross-fold molecular bond-edit distance is at least 10; all spectra follow the assigned molecular fold.","status":"source_checked","source_ids":["evidence-discovery-final-massspecgym"],"source_locator":"Section 3.4; Supplementary Information 2.5"},{"label":"Metrics","value":"Task evaluators distinguish molecular exact match/structural similarity, candidate-retrieval hit rate and spectrum similarity. These are separate readouts, not interchangeable scores.","status":"source_checked","source_ids":["src-discovery-pluskal-lab-massspecgym","evidence-benchmark-massspecgym-de-novo-base-py","evidence-benchmark-massspecgym-retrieval-base-py","evidence-benchmark-massspecgym-simulation-base-py"],"source_locator":"Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned de_novo, retrieval and simulation base evaluation classes"},{"label":"Baselines","value":"The README illustrates a DeepSets-style spectrum-to-fingerprint retrieval baseline.","status":"source_checked","source_ids":["src-discovery-pluskal-lab-massspecgym","evidence-benchmark-massspecgym-de-novo-base-py","evidence-benchmark-massspecgym-retrieval-base-py","evidence-benchmark-massspecgym-simulation-base-py"],"source_locator":"Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned de_novo, retrieval and simulation base evaluation classes"},{"label":"Leakage controls","value":"Maximum common edge subgraph (MCES) clustering keeps molecules connected by a bond-edit distance below 10 in the same fold. The split additionally balances instrument, collision-energy, adduct and molecule-frequency metadata; this is stronger than simply separating 2D InChIKeys.","status":"source_checked","source_ids":["evidence-discovery-final-massspecgym"],"source_locator":"Section 3.4; Supplementary Information 2.5; Tables 2–4"},{"label":"Uncertainty","value":"Tables 2–4 report 99.9% bootstrap confidence intervals using 20,000 resamples. These intervals summarize test-example sampling, not variation across independently retrained models.","status":"source_checked","source_ids":["evidence-discovery-final-massspecgym"],"source_locator":"Section 3.4; Supplementary Information 2.5; Tables 2–4"},{"label":"Entity type","value":"Small-molecule MS/MS benchmark with three task directions.","status":"source_checked","source_ids":["src-discovery-pluskal-lab-massspecgym","evidence-benchmark-massspecgym-de-novo-base-py","evidence-benchmark-massspecgym-retrieval-base-py","evidence-benchmark-massspecgym-simulation-base-py"],"source_locator":"Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned de_novo, retrieval and simulation base evaluation classes"},{"label":"Organisms","value":"Molecule identity rather than organism classification defines these tasks.","status":"inapplicable","source_ids":["src-discovery-pluskal-lab-massspecgym","evidence-benchmark-massspecgym-de-novo-base-py","evidence-benchmark-massspecgym-retrieval-base-py","evidence-benchmark-massspecgym-simulation-base-py"],"source_locator":"Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned de_novo, retrieval and simulation base evaluation classes"},{"label":"Assays","value":"Tandem mass spectra paired with molecular structures.","status":"source_checked","source_ids":["src-discovery-pluskal-lab-massspecgym","evidence-benchmark-massspecgym-de-novo-base-py","evidence-benchmark-massspecgym-retrieval-base-py","evidence-benchmark-massspecgym-simulation-base-py"],"source_locator":"Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned de_novo, retrieval and simulation base evaluation classes"},{"label":"Allowed inputs","value":"Spectrum-to-molecule, spectrum-plus-candidates, or molecule-to-spectrum inputs depend on the selected task.","status":"source_checked","source_ids":["src-discovery-pluskal-lab-massspecgym","evidence-benchmark-massspecgym-de-novo-base-py","evidence-benchmark-massspecgym-retrieval-base-py","evidence-benchmark-massspecgym-simulation-base-py"],"source_locator":"Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned de_novo, retrieval and simulation base evaluation classes"},{"label":"Adaptation","value":"Supervised train/validation/test learning; pretrained or new models use the task-specific interfaces.","status":"source_checked","source_ids":["src-discovery-pluskal-lab-massspecgym","evidence-benchmark-massspecgym-de-novo-base-py","evidence-benchmark-massspecgym-retrieval-base-py","evidence-benchmark-massspecgym-simulation-base-py"],"source_locator":"Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned de_novo, retrieval and simulation base evaluation classes"}],"strengths":[{"text":"One release supplies explicit task interfaces and predefined splits for three complementary predictions.","source_ids":["src-discovery-pluskal-lab-massspecgym","evidence-benchmark-massspecgym-de-novo-base-py","evidence-benchmark-massspecgym-retrieval-base-py","evidence-benchmark-massspecgym-simulation-base-py"],"source_locator":"Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned de_novo, retrieval and simulation base evaluation classes"}],"limitations":[{"text":"Chemical-formula-assisted tasks provide extra input information and must remain separate from unassisted tasks. The MCES split constrains structural similarity but cannot establish independence from every external pretraining corpus.","source_ids":["evidence-discovery-final-massspecgym"],"source_locator":"Section 3.4; Supplementary Information 2.5; Tables 2–4"}],"diagram":{"title":"Evaluation procedure","steps":["Allowed inputs: Spectrum-to-molecule, spectrum-plus-candidates, or molecule-to-spectrum inputs depend on the selected task.","Splits: MCES molecular clusters are grouped into fixed training, validation and test folds, stratified by acquisition metadata. Cross-fold molecular bond-edit distance is at least 10; all spectra follow the assigned molecular fold.","Metrics: Task evaluators distinguish molecular exact match/structural similarity, candidate-retrieval hit rate and spectrum similarity. These are separate readouts, not interchangeable scores."],"caption":"Conceptual procedure. Task variants and protocol versions retain their separate scoring conditions.","source_ids":["src-discovery-pluskal-lab-massspecgym","evidence-benchmark-massspecgym-de-novo-base-py","evidence-benchmark-massspecgym-retrieval-base-py","evidence-benchmark-massspecgym-simulation-base-py","evidence-discovery-final-massspecgym"],"source_locator":"Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned de_novo, retrieval and simulation base evaluation classes; Section 3.4; Supplementary Information 2.5"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction."}}},"description":"Molecular identification from tandem mass spectra","facets":{"areas":["metabolomics"]},"id":"discovery-benchmark-massspecgym","kind":"benchmark","links":[],"name":"MassSpecGym","source_ids":["src-discovery-pluskal-lab-massspecgym"],"status":"discovered"}
{"attributes":{"entity_level":"task","scope_note":"Specialist molecular or omics evaluation; protocol details require review before numerical comparison.","task":"Generate molecular structures from tandem mass spectra","version":null,"historical_missing_metadata":{"dataset_release":"unextracted","metric_implementation":"unextracted","split_manifest":"unextracted","version":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"This task predicts molecular structures from an MS/MS spectrum.","summary_source_ids":["src-discovery-pluskal-lab-massspecgym","evidence-benchmark-massspecgym-de-novo-base-py"],"summary_source_locator":"Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned massspecgym/models/de_novo/base.py evaluation methods","sections":[{"title":"Evaluation methodology","body":"MS/MS spectrum input and molecular-structure target; formula-assisted variant is separate. MCES molecular clusters are grouped into fixed training, validation and test folds, stratified by acquisition metadata. Cross-fold molecular bond-edit distance is at least 10; all spectra follow the assigned molecular fold. Top-k molecular exact-match accuracy using InChIKey identity, maximum fingerprint Tanimoto similarity, minimum MCES distance and predicted-molecule validity. Maximum common edge subgraph (MCES) clustering keeps molecules connected by a bond-edit distance below 10 in the same fold. The split additionally balances instrument, collision-energy, adduct and molecule-frequency metadata; this is stronger than simply separating 2D InChIKeys.","source_ids":["src-discovery-pluskal-lab-massspecgym","evidence-benchmark-massspecgym-de-novo-base-py","evidence-discovery-final-massspecgym"],"source_locator":"Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned massspecgym/models/de_novo/base.py evaluation methods; Section 3.4; Supplementary Information 2.5; Section 3.4; Supplementary Information 2.5; Tables 2–4"}],"facts":[{"label":"Datasets","value":"MS/MS spectrum input and molecular-structure target; formula-assisted variant is separate.","status":"source_checked","source_ids":["src-discovery-pluskal-lab-massspecgym","evidence-benchmark-massspecgym-de-novo-base-py"],"source_locator":"Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned massspecgym/models/de_novo/base.py evaluation methods"},{"label":"Splits","value":"MCES molecular clusters are grouped into fixed training, validation and test folds, stratified by acquisition metadata. Cross-fold molecular bond-edit distance is at least 10; all spectra follow the assigned molecular fold.","status":"source_checked","source_ids":["evidence-discovery-final-massspecgym"],"source_locator":"Section 3.4; Supplementary Information 2.5"},{"label":"Metrics","value":"Top-k molecular exact-match accuracy using InChIKey identity, maximum fingerprint Tanimoto similarity, minimum MCES distance and predicted-molecule validity.","status":"source_checked","source_ids":["src-discovery-pluskal-lab-massspecgym","evidence-benchmark-massspecgym-de-novo-base-py"],"source_locator":"Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned massspecgym/models/de_novo/base.py evaluation methods"},{"label":"Baselines","value":"The README illustrates a DeepSets-style spectrum-to-fingerprint retrieval baseline.","status":"source_checked","source_ids":["src-discovery-pluskal-lab-massspecgym","evidence-benchmark-massspecgym-de-novo-base-py"],"source_locator":"Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned massspecgym/models/de_novo/base.py evaluation methods"},{"label":"Leakage controls","value":"Maximum common edge subgraph (MCES) clustering keeps molecules connected by a bond-edit distance below 10 in the same fold. The split additionally balances instrument, collision-energy, adduct and molecule-frequency metadata; this is stronger than simply separating 2D InChIKeys.","status":"source_checked","source_ids":["evidence-discovery-final-massspecgym"],"source_locator":"Section 3.4; Supplementary Information 2.5; Tables 2–4"},{"label":"Uncertainty","value":"Tables 2–4 report 99.9% bootstrap confidence intervals using 20,000 resamples. These intervals summarize test-example sampling, not variation across independently retrained models.","status":"source_checked","source_ids":["evidence-discovery-final-massspecgym"],"source_locator":"Section 3.4; Supplementary Information 2.5; Tables 2–4"},{"label":"Entity type","value":"Constituent benchmark task: MassSpecGym De novo molecule generation","status":"source_checked","source_ids":["src-discovery-pluskal-lab-massspecgym","evidence-benchmark-massspecgym-de-novo-base-py"],"source_locator":"Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned massspecgym/models/de_novo/base.py evaluation methods"},{"label":"Organisms","value":"Molecule identity rather than organism classification defines these tasks.","status":"inapplicable","source_ids":["src-discovery-pluskal-lab-massspecgym","evidence-benchmark-massspecgym-de-novo-base-py"],"source_locator":"Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned massspecgym/models/de_novo/base.py evaluation methods"},{"label":"Assays","value":"Tandem mass spectra paired with molecular structures.","status":"source_checked","source_ids":["src-discovery-pluskal-lab-massspecgym","evidence-benchmark-massspecgym-de-novo-base-py"],"source_locator":"Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned massspecgym/models/de_novo/base.py evaluation methods"},{"label":"Allowed inputs","value":"MS/MS spectrum; molecular formula is available only in the separately identified formula-assisted variant.","status":"source_checked","source_ids":["src-discovery-pluskal-lab-massspecgym","evidence-benchmark-massspecgym-de-novo-base-py"],"source_locator":"Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned massspecgym/models/de_novo/base.py evaluation methods"},{"label":"Adaptation","value":"Supervised train/validation/test learning; pretrained or new models use the task-specific interfaces.","status":"source_checked","source_ids":["src-discovery-pluskal-lab-massspecgym","evidence-benchmark-massspecgym-de-novo-base-py"],"source_locator":"Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned massspecgym/models/de_novo/base.py evaluation methods"}],"strengths":[{"text":"One release supplies explicit task interfaces and predefined splits for three complementary predictions.","source_ids":["src-discovery-pluskal-lab-massspecgym","evidence-benchmark-massspecgym-de-novo-base-py"],"source_locator":"Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned massspecgym/models/de_novo/base.py evaluation methods"}],"limitations":[{"text":"Chemical-formula-assisted tasks provide extra input information and must remain separate from unassisted tasks. The MCES split constrains structural similarity but cannot establish independence from every external pretraining corpus.","source_ids":["evidence-discovery-final-massspecgym"],"source_locator":"Section 3.4; Supplementary Information 2.5; Tables 2–4"}],"diagram":{"title":"Evaluation procedure","steps":["Allowed inputs: MS/MS spectrum; molecular formula is available only in the separately identified formula-assisted variant.","Splits: MCES molecular clusters are grouped into fixed training, validation and test folds, stratified by acquisition metadata. Cross-fold molecular bond-edit distance is at least 10; all spectra follow the assigned molecular fold.","Metrics: Top-k molecular exact-match accuracy using InChIKey identity, maximum fingerprint Tanimoto similarity, minimum MCES distance and predicted-molecule validity."],"caption":"Conceptual procedure. Task variants and protocol versions retain their separate scoring conditions.","source_ids":["src-discovery-pluskal-lab-massspecgym","evidence-benchmark-massspecgym-de-novo-base-py","evidence-discovery-final-massspecgym"],"source_locator":"Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned massspecgym/models/de_novo/base.py evaluation methods; Section 3.4; Supplementary Information 2.5"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction."}}},"description":"Generate molecular structures from tandem mass spectra","facets":{"areas":["metabolomics"]},"id":"discovery-benchmark-massspecgym-de-novo-molecule-generation","kind":"benchmark","links":[{"relation":"parent","target_id":"discovery-benchmark-massspecgym"},{"relation":"part_of","target_id":"discovery-benchmark-massspecgym"}],"name":"MassSpecGym De novo molecule generation","source_ids":["src-discovery-pluskal-lab-massspecgym"],"status":"discovered"}
{"attributes":{"entity_level":"task","scope_note":"Specialist molecular or omics evaluation; protocol details require review before numerical comparison.","task":"Rank candidate structures from a tandem mass spectrum","version":null,"historical_missing_metadata":{"dataset_release":"unextracted","metric_implementation":"unextracted","split_manifest":"unextracted","version":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"This task ranks candidate molecular structures for an observed MS/MS spectrum.","summary_source_ids":["src-discovery-pluskal-lab-massspecgym","evidence-benchmark-massspecgym-retrieval-base-py"],"summary_source_locator":"Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned massspecgym/models/retrieval/base.py evaluation methods","sections":[{"title":"Evaluation methodology","body":"MS/MS spectrum plus a candidate set; rank the matching molecule. MCES molecular clusters are grouped into fixed training, validation and test folds, stratified by acquisition metadata. Cross-fold molecular bond-edit distance is at least 10; all spectra follow the assigned molecular fold. Mean hit rate at configured top-k cutoffs, with optional MCES distance for the top-ranked candidate. Maximum common edge subgraph (MCES) clustering keeps molecules connected by a bond-edit distance below 10 in the same fold. The split additionally balances instrument, collision-energy, adduct and molecule-frequency metadata; this is stronger than simply separating 2D InChIKeys.","source_ids":["src-discovery-pluskal-lab-massspecgym","evidence-benchmark-massspecgym-retrieval-base-py","evidence-discovery-final-massspecgym"],"source_locator":"Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned massspecgym/models/retrieval/base.py evaluation methods; Section 3.4; Supplementary Information 2.5; Section 3.4; Supplementary Information 2.5; Tables 2–4"}],"facts":[{"label":"Datasets","value":"MS/MS spectrum plus a candidate set; rank the matching molecule.","status":"source_checked","source_ids":["src-discovery-pluskal-lab-massspecgym","evidence-benchmark-massspecgym-retrieval-base-py"],"source_locator":"Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned massspecgym/models/retrieval/base.py evaluation methods"},{"label":"Splits","value":"MCES molecular clusters are grouped into fixed training, validation and test folds, stratified by acquisition metadata. Cross-fold molecular bond-edit distance is at least 10; all spectra follow the assigned molecular fold.","status":"source_checked","source_ids":["evidence-discovery-final-massspecgym"],"source_locator":"Section 3.4; Supplementary Information 2.5"},{"label":"Metrics","value":"Mean hit rate at configured top-k cutoffs, with optional MCES distance for the top-ranked candidate.","status":"source_checked","source_ids":["src-discovery-pluskal-lab-massspecgym","evidence-benchmark-massspecgym-retrieval-base-py"],"source_locator":"Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned massspecgym/models/retrieval/base.py evaluation methods"},{"label":"Baselines","value":"The README illustrates a DeepSets-style spectrum-to-fingerprint retrieval baseline.","status":"source_checked","source_ids":["src-discovery-pluskal-lab-massspecgym","evidence-benchmark-massspecgym-retrieval-base-py"],"source_locator":"Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned massspecgym/models/retrieval/base.py evaluation methods"},{"label":"Leakage controls","value":"Maximum common edge subgraph (MCES) clustering keeps molecules connected by a bond-edit distance below 10 in the same fold. The split additionally balances instrument, collision-energy, adduct and molecule-frequency metadata; this is stronger than simply separating 2D InChIKeys.","status":"source_checked","source_ids":["evidence-discovery-final-massspecgym"],"source_locator":"Section 3.4; Supplementary Information 2.5; Tables 2–4"},{"label":"Uncertainty","value":"Tables 2–4 report 99.9% bootstrap confidence intervals using 20,000 resamples. These intervals summarize test-example sampling, not variation across independently retrained models.","status":"source_checked","source_ids":["evidence-discovery-final-massspecgym"],"source_locator":"Section 3.4; Supplementary Information 2.5; Tables 2–4"},{"label":"Entity type","value":"Constituent benchmark task: MassSpecGym Molecule retrieval","status":"source_checked","source_ids":["src-discovery-pluskal-lab-massspecgym","evidence-benchmark-massspecgym-retrieval-base-py"],"source_locator":"Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned massspecgym/models/retrieval/base.py evaluation methods"},{"label":"Organisms","value":"Molecule identity rather than organism classification defines these tasks.","status":"inapplicable","source_ids":["src-discovery-pluskal-lab-massspecgym","evidence-benchmark-massspecgym-retrieval-base-py"],"source_locator":"Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned massspecgym/models/retrieval/base.py evaluation methods"},{"label":"Assays","value":"Tandem mass spectra paired with molecular structures.","status":"source_checked","source_ids":["src-discovery-pluskal-lab-massspecgym","evidence-benchmark-massspecgym-retrieval-base-py"],"source_locator":"Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned massspecgym/models/retrieval/base.py evaluation methods"},{"label":"Allowed inputs","value":"MS/MS spectrum plus the supplied molecule candidate set.","status":"source_checked","source_ids":["src-discovery-pluskal-lab-massspecgym","evidence-benchmark-massspecgym-retrieval-base-py"],"source_locator":"Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned massspecgym/models/retrieval/base.py evaluation methods"},{"label":"Adaptation","value":"Supervised train/validation/test learning; pretrained or new models use the task-specific interfaces.","status":"source_checked","source_ids":["src-discovery-pluskal-lab-massspecgym","evidence-benchmark-massspecgym-retrieval-base-py"],"source_locator":"Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned massspecgym/models/retrieval/base.py evaluation methods"}],"strengths":[{"text":"One release supplies explicit task interfaces and predefined splits for three complementary predictions.","source_ids":["src-discovery-pluskal-lab-massspecgym","evidence-benchmark-massspecgym-retrieval-base-py"],"source_locator":"Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned massspecgym/models/retrieval/base.py evaluation methods"}],"limitations":[{"text":"Chemical-formula-assisted tasks provide extra input information and must remain separate from unassisted tasks. The MCES split constrains structural similarity but cannot establish independence from every external pretraining corpus.","source_ids":["evidence-discovery-final-massspecgym"],"source_locator":"Section 3.4; Supplementary Information 2.5; Tables 2–4"}],"diagram":{"title":"Evaluation procedure","steps":["Allowed inputs: MS/MS spectrum plus the supplied molecule candidate set.","Splits: MCES molecular clusters are grouped into fixed training, validation and test folds, stratified by acquisition metadata. Cross-fold molecular bond-edit distance is at least 10; all spectra follow the assigned molecular fold.","Metrics: Mean hit rate at configured top-k cutoffs, with optional MCES distance for the top-ranked candidate."],"caption":"Conceptual procedure. Task variants and protocol versions retain their separate scoring conditions.","source_ids":["src-discovery-pluskal-lab-massspecgym","evidence-benchmark-massspecgym-retrieval-base-py","evidence-discovery-final-massspecgym"],"source_locator":"Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned massspecgym/models/retrieval/base.py evaluation methods; Section 3.4; Supplementary Information 2.5"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction."}}},"description":"Rank candidate structures from a tandem mass spectrum","facets":{"areas":["metabolomics"]},"id":"discovery-benchmark-massspecgym-molecule-retrieval","kind":"benchmark","links":[{"relation":"parent","target_id":"discovery-benchmark-massspecgym"},{"relation":"part_of","target_id":"discovery-benchmark-massspecgym"}],"name":"MassSpecGym Molecule retrieval","source_ids":["src-discovery-pluskal-lab-massspecgym"],"status":"discovered"}
{"attributes":{"entity_level":"task","scope_note":"Specialist molecular or omics evaluation; protocol details require review before numerical comparison.","task":"Predict a tandem mass spectrum from molecular structure","version":null,"historical_missing_metadata":{"dataset_release":"unextracted","metric_implementation":"unextracted","split_manifest":"unextracted","version":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"This task predicts an MS/MS spectrum from molecular structure.","summary_source_ids":["src-discovery-pluskal-lab-massspecgym","evidence-benchmark-massspecgym-simulation-base-py"],"summary_source_locator":"Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned massspecgym/models/simulation/base.py evaluation methods","sections":[{"title":"Evaluation methodology","body":"Molecular-structure input and spectrum target; retrieval-based assessment is a distinct evaluation view. MCES molecular clusters are grouped into fixed training, validation and test folds, stratified by acquisition metadata. Cross-fold molecular bond-edit distance is at least 10; all spectra follow the assigned molecular fold. Configured cosine or Jensen–Shannon spectrum similarity, with intensity-transform variants; optional candidate-retrieval hit rates are a separate readout. Maximum common edge subgraph (MCES) clustering keeps molecules connected by a bond-edit distance below 10 in the same fold. The split additionally balances instrument, collision-energy, adduct and molecule-frequency metadata; this is stronger than simply separating 2D InChIKeys.","source_ids":["src-discovery-pluskal-lab-massspecgym","evidence-benchmark-massspecgym-simulation-base-py","evidence-discovery-final-massspecgym"],"source_locator":"Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned massspecgym/models/simulation/base.py evaluation methods; Section 3.4; Supplementary Information 2.5; Section 3.4; Supplementary Information 2.5; Tables 2–4"}],"facts":[{"label":"Datasets","value":"Molecular-structure input and spectrum target; retrieval-based assessment is a distinct evaluation view.","status":"source_checked","source_ids":["src-discovery-pluskal-lab-massspecgym","evidence-benchmark-massspecgym-simulation-base-py"],"source_locator":"Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned massspecgym/models/simulation/base.py evaluation methods"},{"label":"Splits","value":"MCES molecular clusters are grouped into fixed training, validation and test folds, stratified by acquisition metadata. Cross-fold molecular bond-edit distance is at least 10; all spectra follow the assigned molecular fold.","status":"source_checked","source_ids":["evidence-discovery-final-massspecgym"],"source_locator":"Section 3.4; Supplementary Information 2.5"},{"label":"Metrics","value":"Configured cosine or Jensen–Shannon spectrum similarity, with intensity-transform variants; optional candidate-retrieval hit rates are a separate readout.","status":"source_checked","source_ids":["src-discovery-pluskal-lab-massspecgym","evidence-benchmark-massspecgym-simulation-base-py"],"source_locator":"Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned massspecgym/models/simulation/base.py evaluation methods"},{"label":"Baselines","value":"The README illustrates a DeepSets-style spectrum-to-fingerprint retrieval baseline.","status":"source_checked","source_ids":["src-discovery-pluskal-lab-massspecgym","evidence-benchmark-massspecgym-simulation-base-py"],"source_locator":"Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned massspecgym/models/simulation/base.py evaluation methods"},{"label":"Leakage controls","value":"Maximum common edge subgraph (MCES) clustering keeps molecules connected by a bond-edit distance below 10 in the same fold. The split additionally balances instrument, collision-energy, adduct and molecule-frequency metadata; this is stronger than simply separating 2D InChIKeys.","status":"source_checked","source_ids":["evidence-discovery-final-massspecgym"],"source_locator":"Section 3.4; Supplementary Information 2.5; Tables 2–4"},{"label":"Uncertainty","value":"Tables 2–4 report 99.9% bootstrap confidence intervals using 20,000 resamples. These intervals summarize test-example sampling, not variation across independently retrained models.","status":"source_checked","source_ids":["evidence-discovery-final-massspecgym"],"source_locator":"Section 3.4; Supplementary Information 2.5; Tables 2–4"},{"label":"Entity type","value":"Constituent benchmark task: MassSpecGym Spectrum simulation","status":"source_checked","source_ids":["src-discovery-pluskal-lab-massspecgym","evidence-benchmark-massspecgym-simulation-base-py"],"source_locator":"Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned massspecgym/models/simulation/base.py evaluation methods"},{"label":"Organisms","value":"Molecule identity rather than organism classification defines these tasks.","status":"inapplicable","source_ids":["src-discovery-pluskal-lab-massspecgym","evidence-benchmark-massspecgym-simulation-base-py"],"source_locator":"Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned massspecgym/models/simulation/base.py evaluation methods"},{"label":"Assays","value":"Tandem mass spectra paired with molecular structures.","status":"source_checked","source_ids":["src-discovery-pluskal-lab-massspecgym","evidence-benchmark-massspecgym-simulation-base-py"],"source_locator":"Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned massspecgym/models/simulation/base.py evaluation methods"},{"label":"Allowed inputs","value":"Molecular structure, with the measured spectrum used only as the target.","status":"source_checked","source_ids":["src-discovery-pluskal-lab-massspecgym","evidence-benchmark-massspecgym-simulation-base-py"],"source_locator":"Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned massspecgym/models/simulation/base.py evaluation methods"},{"label":"Adaptation","value":"Supervised train/validation/test learning; pretrained or new models use the task-specific interfaces.","status":"source_checked","source_ids":["src-discovery-pluskal-lab-massspecgym","evidence-benchmark-massspecgym-simulation-base-py"],"source_locator":"Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned massspecgym/models/simulation/base.py evaluation methods"}],"strengths":[{"text":"One release supplies explicit task interfaces and predefined splits for three complementary predictions.","source_ids":["src-discovery-pluskal-lab-massspecgym","evidence-benchmark-massspecgym-simulation-base-py"],"source_locator":"Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned massspecgym/models/simulation/base.py evaluation methods"}],"limitations":[{"text":"Chemical-formula-assisted tasks provide extra input information and must remain separate from unassisted tasks. The MCES split constrains structural similarity but cannot establish independence from every external pretraining corpus.","source_ids":["evidence-discovery-final-massspecgym"],"source_locator":"Section 3.4; Supplementary Information 2.5; Tables 2–4"}],"diagram":{"title":"Evaluation procedure","steps":["Allowed inputs: Molecular structure, with the measured spectrum used only as the target.","Splits: MCES molecular clusters are grouped into fixed training, validation and test folds, stratified by acquisition metadata. Cross-fold molecular bond-edit distance is at least 10; all spectra follow the assigned molecular fold.","Metrics: Configured cosine or Jensen–Shannon spectrum similarity, with intensity-transform variants; optional candidate-retrieval hit rates are a separate readout."],"caption":"Conceptual procedure. Task variants and protocol versions retain their separate scoring conditions.","source_ids":["src-discovery-pluskal-lab-massspecgym","evidence-benchmark-massspecgym-simulation-base-py","evidence-discovery-final-massspecgym"],"source_locator":"Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned massspecgym/models/simulation/base.py evaluation methods; Section 3.4; Supplementary Information 2.5"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction."}}},"description":"Predict a tandem mass spectrum from molecular structure","facets":{"areas":["metabolomics"]},"id":"discovery-benchmark-massspecgym-spectrum-simulation","kind":"benchmark","links":[{"relation":"parent","target_id":"discovery-benchmark-massspecgym"},{"relation":"part_of","target_id":"discovery-benchmark-massspecgym"}],"name":"MassSpecGym Spectrum simulation","source_ids":["src-discovery-pluskal-lab-massspecgym"],"status":"discovered"}
{"attributes":{"entity_level":"suite","scope_note":"Specialist molecular or omics evaluation; protocol details require review before numerical comparison.","task":"mRNA embedding quality on downstream tasks","version":null,"historical_missing_metadata":{"dataset_release":"unextracted","metric_implementation":"unextracted","split_manifest":"unextracted","version":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"mRNABench assesses genomic-model embeddings on transcript-specific expression, stability and regulatory tasks.","summary_source_ids":["src-discovery-morrislab-mrnabench"],"summary_source_locator":"Pinned README: Overview; catalogue; dataset implementation requirements","sections":[{"title":"Evaluation methodology","body":"mRNABench evaluates mature-transcript representations on local sequence effects and global RNA properties. Linear probes use task-specific labels, with homology-based partitions where applicable. Chromosomal, k-mer and homology grouping are compared explicitly because random splits can overstate generalization.","source_ids":["evidence-discovery-final-mrnabench"],"source_locator":"Sections 3–4; Table 2; Appendix A and D"}],"facts":[{"label":"Datasets","value":"Named transcript datasets include translation efficiency, ribosome loading and RNA half-life.","status":"source_checked","source_ids":["src-discovery-morrislab-mrnabench"],"source_locator":"Pinned README: Overview; catalogue; dataset implementation requirements"},{"label":"Splits","value":"The library includes training split logic and supports homology-aware splitting using gene identifiers.","status":"source_checked","source_ids":["src-discovery-morrislab-mrnabench"],"source_locator":"Pinned README: Overview; catalogue; dataset implementation requirements"},{"label":"Metrics","value":"Task-specific AUPRC for classification and Pearson correlation for continuous RNA properties. Cross-task summaries Fisher-transform correlations before z-scoring; these derived summaries are distinct from the original per-assay metric.","status":"source_checked","source_ids":["evidence-discovery-final-mrnabench"],"source_locator":"Sections 3–4; Table 2; Appendix A and D"},{"label":"Baselines","value":"NaiveBaseline uses k-mer/GC/sequence statistics, with a six-track variant adding CDS length and exon count; NaiveMamba is an untrained fixed-seed reference.","status":"source_checked","source_ids":["src-discovery-morrislab-mrnabench"],"source_locator":"Pinned README: Overview; catalogue; dataset implementation requirements"},{"label":"Leakage controls","value":"Homology splitting requires explicit gene metadata; its presence in the library does not establish that every dataset uses it.","status":"source_checked","source_ids":["src-discovery-morrislab-mrnabench"],"source_locator":"Pinned README: Overview; catalogue; dataset implementation requirements"},{"label":"Uncertainty","value":"Linear-probe results are means over ten random splits/seeds. Appendix A and D document standard errors and the selected configurations; Table 2 also uses a Wilcoxon signed-rank comparison.","status":"source_checked","source_ids":["evidence-discovery-final-mrnabench"],"source_locator":"Sections 3–4; Table 2; Appendix A and D"},{"label":"Entity type","value":"mRNA representation benchmark suite.","status":"source_checked","source_ids":["src-discovery-morrislab-mrnabench"],"source_locator":"Pinned README: Overview; catalogue; dataset implementation requirements"},{"label":"Organisms","value":"Human and other dataset-specific transcript collections; the splitter example explicitly conditions on human homology.","status":"source_checked","source_ids":["src-discovery-morrislab-mrnabench"],"source_locator":"Pinned README: Overview; catalogue; dataset implementation requirements"},{"label":"Assays","value":"Translation efficiency, ribosome load, half-life and transcript-associated annotations.","status":"source_checked","source_ids":["src-discovery-morrislab-mrnabench"],"source_locator":"Pinned README: Overview; catalogue; dataset implementation requirements"},{"label":"Allowed inputs","value":"Transcript sequences; some feature baselines additionally use coding-region and splice annotations.","status":"source_checked","source_ids":["src-discovery-morrislab-mrnabench"],"source_locator":"Pinned README: Overview; catalogue; dataset implementation requirements"},{"label":"Adaptation","value":"Frozen embeddings with linear probes; split logic is selected independently of the embedding model.","status":"source_checked","source_ids":["src-discovery-morrislab-mrnabench"],"source_locator":"Pinned README: Overview; catalogue; dataset implementation requirements"}],"strengths":[{"text":"Naive sequence-feature and randomly initialized model baselines test whether pretraining adds useful information.","source_ids":["src-discovery-morrislab-mrnabench"],"source_locator":"Pinned README: Overview; catalogue; dataset implementation requirements"}],"limitations":[{"text":"Homology grouping is not used for every assay: the paper retains random splits for MRL-MPRA, MRL-HL-PAIR and variant effects. Derived cross-task z-scores should not replace the original biological metrics.","source_ids":["evidence-discovery-final-mrnabench"],"source_locator":"Sections 3–4; Table 2; Appendix A and D"}],"diagram":{"title":"Evaluation procedure","steps":["Allowed inputs: Transcript sequences; some feature baselines additionally use coding-region and splice annotations.","Splits: The library includes training split logic and supports homology-aware splitting using gene identifiers.","Metrics: Task-specific AUPRC for classification and Pearson correlation for continuous RNA properties. Cross-task summaries Fisher-transform correlations before z-scoring; these derived summaries are distinct from the original per-assay metric."],"caption":"Conceptual procedure. Task variants and protocol versions retain their separate scoring conditions.","source_ids":["src-discovery-morrislab-mrnabench","evidence-discovery-final-mrnabench"],"source_locator":"Pinned README: Overview; catalogue; dataset implementation requirements; Sections 3–4; Table 2; Appendix A and D"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction."}}},"description":"mRNA embedding quality on downstream tasks","facets":{"areas":["rna"]},"id":"discovery-benchmark-mrnabench","kind":"benchmark","links":[],"name":"mRNABench","source_ids":["src-discovery-morrislab-mrnabench"],"status":"discovered"}
{"attributes":{"entity_level":"suite","scope_note":"Specialist molecular or omics evaluation; protocol details require review before numerical comparison.","task":"DNA and RNA fitness prediction","version":null,"historical_missing_metadata":{"dataset_release":"unextracted","metric_implementation":"unextracted","split_manifest":"unextracted","version":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"NABench compares nucleotide foundation models on measured DNA/RNA sequence effects under multiple adaptation settings.","summary_source_ids":["src-discovery-mrzzmrzz-nabench"],"summary_source_locator":"Pinned README: Introduction; Evaluation settings; evaluation scripts","sections":[{"title":"Evaluation methodology","body":"NABench compares nucleic-acid fitness prediction across DMS and SELEX assays. Zero-shot sequence scores, supervised ridge probes and low-label settings are separate evaluation regimes. Random and contiguous-position folds distinguish interpolation from transfer to unseen mutational regions.","source_ids":["evidence-discovery-final-nabench"],"source_locator":"Sections on evaluation settings and metrics; dataset appendix"}],"facts":[{"label":"Datasets","value":"High-throughput assay collections spanning DNA and RNA families.","status":"source_checked","source_ids":["src-discovery-mrzzmrzz-nabench"],"source_locator":"Pinned README: Introduction; Evaluation settings; evaluation scripts"},{"label":"Splits","value":"Zero-shot, few-shot, supervised and transfer-learning settings are separate benchmark regimes.","status":"source_checked","source_ids":["src-discovery-mrzzmrzz-nabench"],"source_locator":"Pinned README: Introduction; Evaluation settings; evaluation scripts"},{"label":"Metrics","value":"Zero-shot evaluation reports Spearman correlation, NDCG, AUROC and MCC. Supervised and few-shot DMS evaluation emphasizes Spearman correlation, whereas SELEX evaluation uses AUROC.","status":"source_checked","source_ids":["evidence-discovery-final-nabench"],"source_locator":"Sections on evaluation settings and metrics; dataset appendix"},{"label":"Baselines","value":"BERT-like, GPT-like, Hyena and LLaMA-based model families are included.","status":"source_checked","source_ids":["src-discovery-mrzzmrzz-nabench"],"source_locator":"Pinned README: Introduction; Evaluation settings; evaluation scripts"},{"label":"Leakage controls","value":"Supervised probes use five-fold random or contiguous-position partitions. The contiguous version holds out variants mutated in a region of the wild-type sequence to test transfer across mutation positions; it is not a global homology or pretraining-overlap audit.","status":"source_checked","source_ids":["evidence-discovery-final-nabench"],"source_locator":"Sections on evaluation settings and metrics; dataset appendix"},{"label":"Uncertainty","value":"The inspected evaluation sections specify five-fold cross-validation and aggregate assay results, but do not define a uniform seed-based or bootstrap confidence interval for the suite.","status":"unreported","source_ids":["evidence-discovery-final-nabench"],"source_locator":"Sections on evaluation settings and metrics; dataset appendix"},{"label":"Entity type","value":"Nucleic-acid variant-effect benchmark suite.","status":"source_checked","source_ids":["src-discovery-mrzzmrzz-nabench"],"source_locator":"Pinned README: Introduction; Evaluation settings; evaluation scripts"},{"label":"Organisms","value":"Multiple natural DNA/RNA families and synthetic SELEX libraries, with assay-dependent experimental contexts. A synthetic selected sequence need not have a unique organism of origin.","status":"source_checked","source_ids":["evidence-discovery-final-nabench"],"source_locator":"Sections on evaluation settings and metrics; dataset appendix"},{"label":"Assays","value":"DMS and SELEX-derived nucleic-acid measurements; the README distinguishes released DMS data from pending processed SELEX data.","status":"source_checked","source_ids":["src-discovery-mrzzmrzz-nabench"],"source_locator":"Pinned README: Introduction; Evaluation settings; evaluation scripts"},{"label":"Allowed inputs","value":"DNA/RNA sequences and task-specific measured effects.","status":"source_checked","source_ids":["src-discovery-mrzzmrzz-nabench"],"source_locator":"Pinned README: Introduction; Evaluation settings; evaluation scripts"},{"label":"Adaptation","value":"Zero-shot, few-shot, supervised and transfer settings are evaluated separately.","status":"source_checked","source_ids":["src-discovery-mrzzmrzz-nabench"],"source_locator":"Pinned README: Introduction; Evaluation settings; evaluation scripts"}],"strengths":[{"text":"Separating label-access regimes exposes when supervised adaptation changes model comparisons.","source_ids":["src-discovery-mrzzmrzz-nabench"],"source_locator":"Pinned README: Introduction; Evaluation settings; evaluation scripts"}],"limitations":[{"text":"Fitness labels arise from different selection and reporter assays. The paper’s generalization controls do not establish that every underlying sequence was absent from model pretraining.","source_ids":["evidence-discovery-final-nabench"],"source_locator":"Sections on evaluation settings and metrics; dataset appendix"}],"diagram":{"title":"Evaluation procedure","steps":["Allowed inputs: DNA/RNA sequences and task-specific measured effects.","Splits: Zero-shot, few-shot, supervised and transfer-learning settings are separate benchmark regimes.","Metrics: Zero-shot evaluation reports Spearman correlation, NDCG, AUROC and MCC. Supervised and few-shot DMS evaluation emphasizes Spearman correlation, whereas SELEX evaluation uses AUROC."],"caption":"Conceptual procedure. Task variants and protocol versions retain their separate scoring conditions.","source_ids":["src-discovery-mrzzmrzz-nabench","evidence-discovery-final-nabench"],"source_locator":"Pinned README: Introduction; Evaluation settings; evaluation scripts; Sections on evaluation settings and metrics; dataset appendix"},"coverage":"limited","gaps":["Uncertainty: The inspected evaluation sections specify five-fold cross-validation and aggregate assay results, but do not define a uniform seed-based or bootstrap confidence interval for the suite."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction."}}},"description":"DNA and RNA fitness prediction","facets":{"areas":["rna"]},"id":"discovery-benchmark-nabench","kind":"benchmark","links":[],"name":"NABench","source_ids":["src-discovery-mrzzmrzz-nabench"],"status":"discovered"}
{"attributes":{"entity_level":"evaluator","scope_note":"Specialist molecular or omics evaluation; protocol details require review before numerical comparison.","task":"Metagenomic taxonomic profile evaluation","version":null,"historical_missing_metadata":{"dataset_release":"unextracted","metric_implementation":"unextracted","split_manifest":"unextracted","version":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"OPAL evaluates taxon presence and abundance profiles against a reference community.","summary_source_ids":["src-discovery-cami-challenge-opal"],"summary_source_locator":"Pinned README: introduction; Computed metrics; input format; example evaluations","sections":[{"title":"Evaluation methodology","body":"Predicted taxonomic abundance profiles and a gold-standard profile in supported formats. The evaluator scores supplied outputs; predictor train/test partitions belong to the dataset/run being evaluated. Precision, recall, F1, Jaccard, L1 error, UniFrac, Bray–Curtis and diversity measures. Multiple profiling tools can be compared; published example reports use CAMI and mock-community data. Uncertainty across samples, datasets or training runs must be defined by the evaluation study; this evaluator entry does not fix one experiment.","source_ids":["src-discovery-cami-challenge-opal"],"source_locator":"Pinned README: introduction; Computed metrics; input format; example evaluations"}],"facts":[{"label":"Datasets","value":"Predicted taxonomic abundance profiles and a gold-standard profile in supported formats.","status":"source_checked","source_ids":["src-discovery-cami-challenge-opal"],"source_locator":"Pinned README: introduction; Computed metrics; input format; example evaluations"},{"label":"Splits","value":"The evaluator scores supplied outputs; predictor train/test partitions belong to the dataset/run being evaluated.","status":"inapplicable","source_ids":["src-discovery-cami-challenge-opal"],"source_locator":"Pinned README: introduction; Computed metrics; input format; example evaluations"},{"label":"Metrics","value":"Precision, recall, F1, Jaccard, L1 error, UniFrac, Bray–Curtis and diversity measures.","status":"source_checked","source_ids":["src-discovery-cami-challenge-opal"],"source_locator":"Pinned README: introduction; Computed metrics; input format; example evaluations"},{"label":"Baselines","value":"Multiple profiling tools can be compared; published example reports use CAMI and mock-community data.","status":"source_checked","source_ids":["src-discovery-cami-challenge-opal"],"source_locator":"Pinned README: introduction; Computed metrics; input format; example evaluations"},{"label":"Leakage controls","value":"OPAL scores taxonomic profiles against a supplied gold standard; it does not define or audit predictor training data. Reference-database cutoffs and novelty controls must accompany the evaluated challenge.","status":"inapplicable","source_ids":["evidence-discovery-final-opal"],"source_locator":"Implementation: input data and evaluation metrics"},{"label":"Uncertainty","value":"Uncertainty across samples, datasets or training runs must be defined by the evaluation study; this evaluator entry does not fix one experiment.","status":"inapplicable","source_ids":["src-discovery-cami-challenge-opal"],"source_locator":"Pinned README: introduction; Computed metrics; input format; example evaluations"},{"label":"Entity type","value":"Evaluator for taxonomic abundance profiling.","status":"source_checked","source_ids":["src-discovery-cami-challenge-opal"],"source_locator":"Pinned README: introduction; Computed metrics; input format; example evaluations"},{"label":"Organisms","value":"The evaluator is not restricted to a single organism; profiles describe microbial communities.","status":"inapplicable","source_ids":["src-discovery-cami-challenge-opal"],"source_locator":"Pinned README: introduction; Computed metrics; input format; example evaluations"},{"label":"Assays","value":"Taxonomic abundance profiles derived from metagenomic analyses.","status":"source_checked","source_ids":["src-discovery-cami-challenge-opal"],"source_locator":"Pinned README: introduction; Computed metrics; input format; example evaluations"},{"label":"Allowed inputs","value":"Predicted and gold-standard taxon abundances in supported formats.","status":"source_checked","source_ids":["src-discovery-cami-challenge-opal"],"source_locator":"Pinned README: introduction; Computed metrics; input format; example evaluations"},{"label":"Adaptation","value":"OPAL evaluates profiles and does not define model adaptation.","status":"inapplicable","source_ids":["src-discovery-cami-challenge-opal"],"source_locator":"Pinned README: introduction; Computed metrics; input format; example evaluations"}],"strengths":[{"text":"Separates abundance agreement from presence/absence detection.","source_ids":["src-discovery-cami-challenge-opal"],"source_locator":"Pinned README: introduction; Computed metrics; input format; example evaluations"}],"limitations":[{"text":"Taxonomic profiles must use compatible taxonomy and abundance conventions. A low profiling error does not establish independence from the reference genomes.","source_ids":["evidence-discovery-final-opal"],"source_locator":"Implementation: input data and evaluation metrics"}],"diagram":{"title":"Evaluation procedure","steps":["Allowed inputs: Predicted and gold-standard taxon abundances in supported formats.","Splits: The evaluator scores supplied outputs; predictor train/test partitions belong to the dataset/run being evaluated.","Metrics: Precision, recall, F1, Jaccard, L1 error, UniFrac, Bray–Curtis and diversity measures."],"caption":"Conceptual procedure. Task variants and protocol versions retain their separate scoring conditions.","source_ids":["src-discovery-cami-challenge-opal"],"source_locator":"Pinned README: introduction; Computed metrics; input format; example evaluations"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction."}}},"description":"Metagenomic taxonomic profile evaluation","facets":{"areas":["microbiome"]},"id":"discovery-benchmark-opal","kind":"benchmark","links":[],"name":"OPAL","source_ids":["src-discovery-cami-challenge-opal"],"status":"discovered"}
{"attributes":{"entity_level":"suite","scope_note":"Specialist molecular or omics evaluation; protocol details require review before numerical comparison.","task":"Community single-cell analysis benchmarks","version":null,"historical_missing_metadata":{"dataset_release":"unextracted","metric_implementation":"unextracted","split_manifest":"unextracted","version":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Open Problems is an extensible platform hosting benchmark tasks and their datasets.","summary_source_ids":["src-discovery-openproblems-bio-openproblems"],"summary_source_locator":"Pinned README: platform description and benchmark/dataset resource links","sections":[{"title":"Evaluation methodology","body":"Open Problems is a collection of separately versioned single-cell evaluation tasks. Each task defines its inputs, reference data, methods, controls and metrics. For example, label projection learns from reference labels and predicts a held-out dataset, whereas integration measures how supplied batches are combined.","source_ids":["evidence-discovery-final-openproblems-label"],"source_locator":"Label Projection v1.0.0: task description, dataset variants and controls"}],"facts":[{"label":"Datasets","value":"The README links benchmark and dataset catalogues rather than fixing one data release.","status":"source_checked","source_ids":["src-discovery-openproblems-bio-openproblems"],"source_locator":"Pinned README: platform description and benchmark/dataset resource links"},{"label":"Splits","value":"No platform-wide split exists: select a hosted benchmark task and its dataset protocol.","status":"inapplicable","source_ids":["src-discovery-openproblems-bio-openproblems"],"source_locator":"Pinned README: platform description and benchmark/dataset resource links"},{"label":"Metrics","value":"No platform-wide scientific metric applies: hosted task definitions select their own evaluator.","status":"inapplicable","source_ids":["src-discovery-openproblems-bio-openproblems"],"source_locator":"Pinned README: platform description and benchmark/dataset resource links"},{"label":"Baselines","value":"No single platform-wide baseline: methods and references belong to each hosted task.","status":"inapplicable","source_ids":["src-discovery-openproblems-bio-openproblems"],"source_locator":"Pinned README: platform description and benchmark/dataset resource links"},{"label":"Leakage controls","value":"Controls are task dependent. Label Projection v1.0.0 distinguishes training/test batches and includes both random and batch-based CeNGEN dataset variants. Other tasks, such as batch integration, evaluate transductive processing of the supplied cells; a universal held-out-cell rule would be misleading.","status":"source_checked","source_ids":["evidence-discovery-final-openproblems-label"],"source_locator":"Label Projection v1.0.0: task description, dataset variants and controls"},{"label":"Uncertainty","value":"The inspected task pages and reporting configuration do not prescribe one platform-wide bootstrap or repeated-seed interval. Individual task versions define datasets, metrics and runs; model uncertainty from a method such as scANVI is not benchmark-score uncertainty.","status":"unreported","source_ids":["evidence-discovery-final-openproblems-label"],"source_locator":"Label Projection v1.0.0: task description, dataset variants and controls"},{"label":"Entity type","value":"Platform hosting computational biology benchmark tasks.","status":"source_checked","source_ids":["src-discovery-openproblems-bio-openproblems"],"source_locator":"Pinned README: platform description and benchmark/dataset resource links"},{"label":"Organisms","value":"Organisms belong to the selected task and dataset; the platform defines no unique organism.","status":"inapplicable","source_ids":["src-discovery-openproblems-bio-openproblems"],"source_locator":"Pinned README: platform description and benchmark/dataset resource links"},{"label":"Assays","value":"The platform does not prescribe one assay.","status":"inapplicable","source_ids":["src-discovery-openproblems-bio-openproblems"],"source_locator":"Pinned README: platform description and benchmark/dataset resource links"},{"label":"Allowed inputs","value":"Inputs are specified by each hosted task.","status":"source_checked","source_ids":["src-discovery-openproblems-bio-openproblems"],"source_locator":"Pinned README: platform description and benchmark/dataset resource links"},{"label":"Adaptation","value":"Adaptation rules belong to the task implementation; the platform is not a fitted predictor.","status":"inapplicable","source_ids":["src-discovery-openproblems-bio-openproblems"],"source_locator":"Pinned README: platform description and benchmark/dataset resource links"}],"strengths":[{"text":"The platform separates task definitions from methods submitted to them.","source_ids":["src-discovery-openproblems-bio-openproblems"],"source_locator":"Pinned README: platform description and benchmark/dataset resource links"}],"limitations":[{"text":"Results from different Open Problems tasks or split variants are not interchangeable. Perfect-label controls and random-label controls calibrate particular metrics; they are not measured biological performance ceilings.","source_ids":["evidence-discovery-final-openproblems-label"],"source_locator":"Label Projection v1.0.0: task description, dataset variants and controls"}],"diagram":{"title":"Evaluation procedure","steps":["Allowed inputs: Inputs are specified by each hosted task.","Splits: No platform-wide split exists: select a hosted benchmark task and its dataset protocol.","Metrics: No platform-wide scientific metric applies: hosted task definitions select their own evaluator."],"caption":"Conceptual procedure. Task variants and protocol versions retain their separate scoring conditions.","source_ids":["src-discovery-openproblems-bio-openproblems"],"source_locator":"Pinned README: platform description and benchmark/dataset resource links"},"coverage":"limited","gaps":["Uncertainty: The inspected task pages and reporting configuration do not prescribe one platform-wide bootstrap or repeated-seed interval. Individual task versions define datasets, metrics and runs; model uncertainty from a method such as scANVI is not benchmark-score uncertainty."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction."}}},"description":"Community single-cell analysis benchmarks","facets":{"areas":["single-cell"]},"id":"discovery-benchmark-open-problems","kind":"benchmark","links":[{"relation":"evaluates_task","target_id":"catalog-task-cell-batch-integration"}],"name":"Open Problems","source_ids":["src-discovery-openproblems-bio-openproblems"],"status":"discovered"}
{"attributes":{"entity_level":"suite","scope_note":"Specialist molecular or omics evaluation; protocol details require review before numerical comparison.","task":"Predicting cellular perturbation responses","version":null,"historical_missing_metadata":{"dataset_release":"unextracted","metric_implementation":"unextracted","split_manifest":"unextracted","version":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"PerturBench evaluates predicted single-cell perturbation responses with explicit aggregation and metric choices.","summary_source_ids":["src-discovery-altoslabs-perturbench"],"summary_source_locator":"Pinned README: Data; Model evaluation; custom dataset configuration","sections":[{"title":"Evaluation methodology","body":"PerturBench predicts single-cell responses in held-out perturbation–context combinations. It implements cross-covariate, combinatorial and inverse-combinatorial partitions, comparing learned models with simple controls. Rank-based metrics complement expression-error metrics to reveal models that fail to distinguish perturbations.","source_ids":["evidence-discovery-final-perturbench"],"source_locator":"Experimental setup; Appendix datasets and data splitting"}],"facts":[{"label":"Datasets","value":"Processed AnnData datasets with perturbation/covariate metadata and configurable feature selections.","status":"source_checked","source_ids":["src-discovery-altoslabs-perturbench"],"source_locator":"Pinned README: Data; Model evaluation; custom dataset configuration"},{"label":"Splits","value":"Cross-cell-type and combination-prediction splits are supported, along with explicit custom split files.","status":"source_checked","source_ids":["src-discovery-altoslabs-perturbench"],"source_locator":"Pinned README: Data; Model evaluation; custom dataset configuration"},{"label":"Metrics","value":"Expression/change aggregation precedes metrics such as cosine, Pearson, RMSE, MSE, MAE and R-squared; optional rank metrics form another view.","status":"source_checked","source_ids":["src-discovery-altoslabs-perturbench"],"source_locator":"Pinned README: Data; Model evaluation; custom dataset configuration"},{"label":"Baselines","value":"Reproduction configurations include linear reference models.","status":"source_checked","source_ids":["src-discovery-altoslabs-perturbench"],"source_locator":"Pinned README: Data; Model evaluation; custom dataset configuration"},{"label":"Leakage controls","value":"Split and covariate definitions are configuration inputs; exact evaluation files must be pinned.","status":"source_checked","source_ids":["src-discovery-altoslabs-perturbench"],"source_locator":"Pinned README: Data; Model evaluation; custom dataset configuration"},{"label":"Uncertainty","value":"For the best hyperparameter configuration the authors run four additional training seeds, yielding five runs. Error bars represent standard deviation of model performance across those runs.","status":"source_checked","source_ids":["evidence-discovery-final-perturbench"],"source_locator":"Experimental setup; Appendix datasets and data splitting"},{"label":"Entity type","value":"Perturbation-response evaluation framework.","status":"source_checked","source_ids":["src-discovery-altoslabs-perturbench"],"source_locator":"Pinned README: Data; Model evaluation; custom dataset configuration"},{"label":"Organisms","value":"The evaluated datasets use human cell-line perturbation systems, including McFaline-Figueroa’s glioblastoma cell contexts and Srivatsan’s chemical perturbation cell lines. The framework itself is not restricted to those organisms.","status":"source_checked","source_ids":["evidence-discovery-final-perturbench"],"source_locator":"Experimental setup; Appendix datasets and data splitting"},{"label":"Assays","value":"Single-cell perturbation response measurements.","status":"source_checked","source_ids":["src-discovery-altoslabs-perturbench"],"source_locator":"Pinned README: Data; Model evaluation; custom dataset configuration"},{"label":"Allowed inputs","value":"Predicted/observed expression and perturbation/covariate metadata in AnnData.","status":"source_checked","source_ids":["src-discovery-altoslabs-perturbench"],"source_locator":"Pinned README: Data; Model evaluation; custom dataset configuration"},{"label":"Adaptation","value":"Supports supervised response prediction with explicit cell-type and combination holdouts.","status":"source_checked","source_ids":["src-discovery-altoslabs-perturbench"],"source_locator":"Pinned README: Data; Model evaluation; custom dataset configuration"}],"strengths":[{"text":"Separating response aggregation from scoring makes a major source of metric disagreement explicit.","source_ids":["src-discovery-altoslabs-perturbench"],"source_locator":"Pinned README: Data; Model evaluation; custom dataset configuration"}],"limitations":[{"text":"A covariate-transfer split can expose a perturbation in another cell context during training; it should not be called completely unseen-perturbation prediction. Report the exact split and perturbation inputs.","source_ids":["evidence-discovery-final-perturbench"],"source_locator":"Experimental setup; Appendix datasets and data splitting"}],"diagram":{"title":"Evaluation procedure","steps":["Allowed inputs: Predicted/observed expression and perturbation/covariate metadata in AnnData.","Splits: Cross-cell-type and combination-prediction splits are supported, along with explicit custom split files.","Metrics: Expression/change aggregation precedes metrics such as cosine, Pearson, RMSE, MSE, MAE and R-squared; optional rank metrics form another view."],"caption":"Conceptual procedure. Task variants and protocol versions retain their separate scoring conditions.","source_ids":["src-discovery-altoslabs-perturbench"],"source_locator":"Pinned README: Data; Model evaluation; custom dataset configuration"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction."}}},"description":"Predicting cellular perturbation responses","facets":{"areas":["single-cell"]},"id":"discovery-benchmark-perturbench","kind":"benchmark","links":[],"name":"PerturBench","source_ids":["src-discovery-altoslabs-perturbench"],"status":"discovered"}
{"attributes":{"entity_level":"suite","scope_note":"Specialist molecular or omics evaluation; protocol details require review before numerical comparison.","task":"Parameter estimation for biological dynamical models","version":null,"historical_missing_metadata":{"dataset_release":"unextracted","metric_implementation":"unextracted","split_manifest":"unextracted","version":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"The PEtab collection supports evaluation of computational methods for fitting mathematical models to observations.","summary_source_ids":["src-discovery-benchmarking-initiative-benchmark-models-petab"],"summary_source_locator":"Pinned README: collection description; benchmark problem table","sections":[{"title":"Evaluation methodology","body":"The collection packages dynamical biological models with the experimental measurements and observation/noise assumptions needed for parameter estimation. A benchmark run specifies a model, solver and inference procedure against this fixed problem. Calibration fit, computational reliability and predictive validation are different assessment targets.","source_ids":["evidence-discovery-final-petab"],"source_locator":"Sections 2.2–2.4: observations, noise models and experimental conditions"}],"facts":[{"label":"Datasets","value":"Individual model/measurement problems in PEtab format, with problem-specific observables and noise assumptions.","status":"source_checked","source_ids":["src-discovery-benchmarking-initiative-benchmark-models-petab"],"source_locator":"Pinned README: collection description; benchmark problem table"},{"label":"Splits","value":"A supervised train/test split is not intrinsic to the parameter-estimation problem collection; the chosen study must define any held-out observations.","status":"inapplicable","source_ids":["src-discovery-benchmarking-initiative-benchmark-models-petab"],"source_locator":"Pinned README: collection description; benchmark problem table"},{"label":"Metrics","value":"Objectives depend on each PEtab measurement/noise model and the selected optimization-performance criterion; there is no universal prediction metric.","status":"inapplicable","source_ids":["src-discovery-benchmarking-initiative-benchmark-models-petab"],"source_locator":"Pinned README: collection description; benchmark problem table"},{"label":"Baselines","value":"The collection provides common problem definitions for comparing modeling/estimation methods, not a fixed universal baseline.","status":"source_checked","source_ids":["src-discovery-benchmarking-initiative-benchmark-models-petab"],"source_locator":"Pinned README: collection description; benchmark problem table"},{"label":"Leakage controls","value":"The collection provides mechanistic models with calibration data, observation functions and experimental conditions. It is intended to compare numerical inference methods; a predictive train/test exclusion policy must be defined by the study using each model.","status":"inapplicable","source_ids":["evidence-discovery-final-petab"],"source_locator":"Sections 2.2–2.4: observations, noise models and experimental conditions"},{"label":"Uncertainty","value":"Measurement errors can be fixed from experiments or estimated jointly through explicit noise models. Parameter uncertainty and optimizer variability are different quantities and require their own evaluation procedure.","status":"source_checked","source_ids":["evidence-discovery-final-petab"],"source_locator":"Sections 2.2–2.4: observations, noise models and experimental conditions"},{"label":"Entity type","value":"Collection of parameter-estimation benchmark problems.","status":"source_checked","source_ids":["src-discovery-benchmarking-initiative-benchmark-models-petab"],"source_locator":"Pinned README: collection description; benchmark problem table"},{"label":"Organisms","value":"Organism identity is problem-specific; the collection spans distinct systems.","status":"inapplicable","source_ids":["src-discovery-benchmarking-initiative-benchmark-models-petab"],"source_locator":"Pinned README: collection description; benchmark problem table"},{"label":"Assays","value":"Problem-specific observations with explicit measurement/noise models.","status":"source_checked","source_ids":["src-discovery-benchmarking-initiative-benchmark-models-petab"],"source_locator":"Pinned README: collection description; benchmark problem table"},{"label":"Allowed inputs","value":"PEtab model, parameter, condition, observable and measurement tables.","status":"source_checked","source_ids":["src-discovery-benchmarking-initiative-benchmark-models-petab"],"source_locator":"Pinned README: collection description; benchmark problem table"},{"label":"Adaptation","value":"Numerical parameter estimation against supplied observations; optimizer settings define the tested method.","status":"source_checked","source_ids":["src-discovery-benchmarking-initiative-benchmark-models-petab"],"source_locator":"Pinned README: collection description; benchmark problem table"}],"strengths":[{"text":"Standardized problem definitions expose differences in objectives and measurement-error assumptions.","source_ids":["src-discovery-benchmarking-initiative-benchmark-models-petab"],"source_locator":"Pinned README: collection description; benchmark problem table"}],"limitations":[{"text":"The original collection paper describes the scientific problem format; the current PEtab repository is a later versioned distribution. Fitting calibration data is not evidence of accuracy on an independent experimental condition.","source_ids":["evidence-discovery-final-petab"],"source_locator":"Sections 2.2–2.4: observations, noise models and experimental conditions"}],"diagram":{"title":"Evaluation procedure","steps":["Allowed inputs: PEtab model, parameter, condition, observable and measurement tables.","Splits: A supervised train/test split is not intrinsic to the parameter-estimation problem collection; the chosen study must define any held-out observations.","Metrics: Objectives depend on each PEtab measurement/noise model and the selected optimization-performance criterion; there is no universal prediction metric."],"caption":"Conceptual procedure. Task variants and protocol versions retain their separate scoring conditions.","source_ids":["src-discovery-benchmarking-initiative-benchmark-models-petab"],"source_locator":"Pinned README: collection description; benchmark problem table"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction."}}},"description":"Parameter estimation for biological dynamical models","facets":{"areas":["mechanistic-biology"]},"id":"discovery-benchmark-petab-benchmark-collection","kind":"benchmark","links":[],"name":"PEtab benchmark collection","source_ids":["src-discovery-benchmarking-initiative-benchmark-models-petab"],"status":"discovered"}
{"attributes":{"entity_level":"suite","scope_note":"Specialist molecular or omics evaluation; protocol details require review before numerical comparison.","task":"Protein representation evaluation","version":null,"historical_missing_metadata":{"dataset_release":"unextracted","metric_implementation":"unextracted","split_manifest":"unextracted","version":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"PFMBench is a configurable suite of protein-model downstream evaluations.","summary_source_ids":["src-discovery-biomap-research-pfmbench"],"summary_source_locator":"Pinned README: Overview; Features; repository architecture","sections":[{"title":"Evaluation methodology","body":"PFMBench evaluates protein representations across structural, functional, interaction and engineering tasks. Most datasets use sequence-similarity partitions, while mutation datasets retain their original assay splits. A stability screen identifies a core task subset, which must remain distinguishable from the full collection.","source_ids":["evidence-discovery-final-pfmbench"],"source_locator":"Benchmark construction and evaluation setup; Appendix task definitions"}],"facts":[{"label":"Datasets","value":"Tasks span structure, function, localization, interactions and other protein properties.","status":"source_checked","source_ids":["src-discovery-biomap-research-pfmbench"],"source_locator":"Pinned README: Overview; Features; repository architecture"},{"label":"Splits","value":"Most datasets are split 8:1:1 using a 30% protein sequence-similarity threshold. Mutation datasets are explicitly exempt and preserve their original train/validation/test partitions.","status":"source_checked","source_ids":["evidence-discovery-final-pfmbench"],"source_locator":"Benchmark construction"},{"label":"Metrics","value":"Task-specific metrics include AUROC for several binary function and interaction tasks, accuracy for categorical tasks, and Spearman correlation for continuous fitness, affinity and enzyme properties; the task appendix defines the metric for each dataset.","status":"source_checked","source_ids":["evidence-discovery-final-pfmbench"],"source_locator":"Benchmark construction and evaluation setup; Appendix task definitions"},{"label":"Baselines","value":"The framework supports both fine-tuning on labels and zero-shot evaluations.","status":"source_checked","source_ids":["src-discovery-biomap-research-pfmbench"],"source_locator":"Pinned README: Overview; Features; repository architecture"},{"label":"Leakage controls","value":"Most datasets use an 8:1:1 split with a 30% sequence-similarity threshold. Mutation datasets retain their original partitions. The paper also flags possible functional-label overlap for annotation-aware pretrained models.","status":"source_checked","source_ids":["evidence-discovery-final-pfmbench"],"source_locator":"Benchmark construction and evaluation setup; Appendix task definitions"},{"label":"Uncertainty","value":"ESM2-Adapter is evaluated over three runs to screen task stability. Its reported bias is the best-to-worst spread divided by mean performance; this is not a confidence interval or a rule proven for all models.","status":"source_checked","source_ids":["evidence-discovery-final-pfmbench"],"source_locator":"Benchmark construction and evaluation setup; Appendix task definitions"},{"label":"Entity type","value":"Configurable protein foundation-model evaluation suite.","status":"source_checked","source_ids":["src-discovery-biomap-research-pfmbench"],"source_locator":"Pinned README: Overview; Features; repository architecture"},{"label":"Organisms","value":"Dataset dependent: named tasks include human and yeast protein interactions, broader protein-property collections, molecular binding data and mutation assays. Species is a property of each source dataset, not one suite-wide organism.","status":"source_checked","source_ids":["evidence-discovery-final-pfmbench"],"source_locator":"Benchmark construction and evaluation setup; Appendix task definitions"},{"label":"Assays","value":"Task-specific structure, function, localization and interaction labels.","status":"source_checked","source_ids":["src-discovery-biomap-research-pfmbench"],"source_locator":"Pinned README: Overview; Features; repository architecture"},{"label":"Allowed inputs","value":"Protein task datasets via configurable loaders and prediction heads.","status":"source_checked","source_ids":["src-discovery-biomap-research-pfmbench"],"source_locator":"Pinned README: Overview; Features; repository architecture"},{"label":"Adaptation","value":"Both fine-tuning and zero-shot evaluation are supported; configurations define the adaptation budget.","status":"source_checked","source_ids":["src-discovery-biomap-research-pfmbench"],"source_locator":"Pinned README: Overview; Features; repository architecture"}],"strengths":[{"text":"The same task framework supports multiple representation and tuning strategies.","source_ids":["src-discovery-biomap-research-pfmbench"],"source_locator":"Pinned README: Overview; Features; repository architecture"}],"limitations":[{"text":"Core-task selection based on one adapter’s repeated runs does not guarantee equal reliability for every model. Functional-label pretraining and mutation-specific splits require separate overlap checks.","source_ids":["evidence-discovery-final-pfmbench"],"source_locator":"Benchmark construction and evaluation setup; Appendix task definitions"}],"diagram":{"title":"Evaluation procedure","steps":["Allowed inputs: Protein task datasets via configurable loaders and prediction heads.","Splits: Most datasets are split 8:1:1 using a 30% protein sequence-similarity threshold. Mutation datasets are explicitly exempt and preserve their original train/validation/test partitions.","Metrics: Task-specific metrics include AUROC for several binary function and interaction tasks, accuracy for categorical tasks, and Spearman correlation for continuous fitness, affinity and enzyme properties; the task appendix defines the metric for each dataset."],"caption":"Conceptual procedure. Task variants and protocol versions retain their separate scoring conditions.","source_ids":["src-discovery-biomap-research-pfmbench","evidence-discovery-final-pfmbench"],"source_locator":"Pinned README: Overview; Features; repository architecture; Benchmark construction; Benchmark construction and evaluation setup; Appendix task definitions"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction."}}},"description":"Protein representation evaluation","facets":{"areas":["protein-function"]},"id":"discovery-benchmark-pfmbench","kind":"benchmark","links":[],"name":"PFMBench","source_ids":["src-discovery-biomap-research-pfmbench"],"status":"discovered"}
{"attributes":{"entity_level":"suite","scope_note":"Specialist molecular or omics evaluation; protocol details require review before numerical comparison.","task":"Protein-ligand interaction evaluation","version":null,"historical_missing_metadata":{"dataset_release":"unextracted","metric_implementation":"unextracted","split_manifest":"unextracted","version":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"PLINDER supplies annotated protein–ligand systems and evaluation resources for docking.","summary_source_ids":["src-discovery-plinder-org-plinder"],"summary_source_locator":"Pinned README: Overview; Plinder versions; Known bugs","sections":[{"title":"Evaluation methodology","body":"PLINDER organizes protein–ligand complexes and their similarity relationships so a test set can be characterized by novelty of proteins, pockets, ligands and interactions. The evaluator compares predicted poses with reference systems and preserves matched-chain coverage. Dataset version, split and novelty stratum are essential parts of any reported result.","source_ids":["evidence-discovery-final-plinder-config0"],"source_locator":"Pinned docs/evaluation.md: per-pose scores and test stratification"}],"facts":[{"label":"Datasets","value":"Protein–ligand complexes with linked bound, unbound and predicted receptor structures.","status":"source_checked","source_ids":["src-discovery-plinder-org-plinder"],"source_locator":"Pinned README: Overview; Plinder versions; Known bugs"},{"label":"Splits","value":"Train/validation/test splits can be tuned to the learning task.","status":"source_checked","source_ids":["src-discovery-plinder-org-plinder"],"source_locator":"Pinned README: Overview; Plinder versions; Known bugs"},{"label":"Metrics","value":"Pose scoring includes ligand lDDT-PLI, binding-site-superposed symmetry-corrected RMSD and pocket lDDT. System summaries retain mapped-chain fractions and optionally score receptor lDDT, oligomer interfaces and PoseBusters validity. A pose confidence score is optional input, not benchmark uncertainty.","status":"source_checked","source_ids":["evidence-discovery-final-plinder-config0"],"source_locator":"Pinned docs/evaluation.md: Write scores"},{"label":"Baselines","value":"The official release history identifies a dataset version used to retrain DiffDock and points to the companion Moving Beyond Memorization study. Baseline identity must include the particular PLINDER release/split; a dataset entry does not define one universal reference score.","status":"source_checked","source_ids":["evidence-discovery-final-plinder-readme"],"source_locator":"README: dataset versions and Moving Beyond Memorization reference"},{"label":"Leakage controls","value":"Similarity annotations support task-dependent splitting; a specific split is required before claiming overlap exclusion.","status":"source_checked","source_ids":["src-discovery-plinder-org-plinder"],"source_locator":"Pinned README: Overview; Plinder versions; Known bugs"},{"label":"Uncertainty","value":"The inspected evaluator documentation defines per-pose metrics, system averages and similarity strata but no universal bootstrap or repeated-training interval. An uncertainty estimate must be attached to a specific evaluated model and dataset release.","status":"unreported","source_ids":["evidence-discovery-final-plinder-config0"],"source_locator":"Pinned docs/evaluation.md: per-pose scores and test stratification"},{"label":"Entity type","value":"Protein–ligand dataset and task-dependent split resource.","status":"source_checked","source_ids":["src-discovery-plinder-org-plinder"],"source_locator":"Pinned README: Overview; Plinder versions; Known bugs"},{"label":"Organisms","value":"Molecular systems define the collection; it is not tied to one organism.","status":"inapplicable","source_ids":["src-discovery-plinder-org-plinder"],"source_locator":"Pinned README: Overview; Plinder versions; Known bugs"},{"label":"Assays","value":"Protein–ligand structural data with bound, unbound and predicted receptor states.","status":"source_checked","source_ids":["src-discovery-plinder-org-plinder"],"source_locator":"Pinned README: Overview; Plinder versions; Known bugs"},{"label":"Allowed inputs","value":"Protein–ligand systems, receptor structures and curated metadata.","status":"source_checked","source_ids":["src-discovery-plinder-org-plinder"],"source_locator":"Pinned README: Overview; Plinder versions; Known bugs"},{"label":"Adaptation","value":"The chosen downstream model and split determine fitting; the data resource imposes no single adaptation scheme.","status":"source_checked","source_ids":["src-discovery-plinder-org-plinder"],"source_locator":"Pinned README: Overview; Plinder versions; Known bugs"}],"strengths":[{"text":"Similarity annotations make protein/ligand overlap inspectable when choosing splits.","source_ids":["src-discovery-plinder-org-plinder"],"source_locator":"Pinned README: Overview; Plinder versions; Known bugs"}],"limitations":[{"text":"A split can be novel by one molecular similarity measure and familiar by another. The documented evaluator produces scores and strata; it does not certify every model’s training history or impose a common uncertainty protocol.","source_ids":["evidence-discovery-final-plinder-config0"],"source_locator":"Pinned docs/evaluation.md: per-pose scores and test stratification"}],"diagram":{"title":"Evaluation procedure","steps":["Allowed inputs: Protein–ligand systems, receptor structures and curated metadata.","Splits: Train/validation/test splits can be tuned to the learning task.","Metrics: Pose scoring includes ligand lDDT-PLI, binding-site-superposed symmetry-corrected RMSD and pocket lDDT. System summaries retain mapped-chain fractions and optionally score receptor lDDT, oligomer interfaces and PoseBusters validity. A pose confidence score is optional input, not benchmark uncertainty."],"caption":"Conceptual procedure. Task variants and protocol versions retain their separate scoring conditions.","source_ids":["src-discovery-plinder-org-plinder","evidence-discovery-final-plinder-config0"],"source_locator":"Pinned README: Overview; Plinder versions; Known bugs; Pinned docs/evaluation.md: Write scores"},"coverage":"limited","gaps":["Uncertainty: The inspected evaluator documentation defines per-pose metrics, system averages and similarity strata but no universal bootstrap or repeated-training interval. An uncertainty estimate must be attached to a specific evaluated model and dataset release."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction."}}},"description":"Protein-ligand interaction evaluation","facets":{"areas":["molecular-interactions"]},"id":"discovery-benchmark-plinder","kind":"benchmark","links":[],"name":"PLINDER","source_ids":["src-discovery-plinder-org-plinder"],"status":"discovered"}
{"attributes":{"entity_level":"evaluator","scope_note":"Specialist molecular or omics evaluation; protocol details require review before numerical comparison.","task":"Geometric and chemical plausibility of molecular poses","version":null,"historical_missing_metadata":{"dataset_release":"unextracted","metric_implementation":"unextracted","split_manifest":"unextracted","version":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"PoseBusters checks the plausibility of predicted molecular poses.","summary_source_ids":["src-discovery-maabuu-posebusters"],"summary_source_locator":"Pinned README: description; Usage; paper/data links","sections":[{"title":"Evaluation methodology","body":"Predicted molecular coordinates, optionally with the conditioning protein, and paper-linked evaluation data. The evaluator scores supplied outputs; predictor train/test partitions belong to the dataset/run being evaluated. Chemical identity/stereochemistry, bond and angle geometry, aromatic planarity, internal clashes and protein–ligand clashes are checked separately. The paper evaluates native-like pose recovery jointly with passing the validity checks; its intramolecular tolerances are 25% for bond lengths/angles and 30% for nonbonded distances. The PoseBusters benchmark selects recent PDB complexes absent from the PDBbind v2020 training source used by evaluated learned docking methods. The paper further examines protein-sequence similarity to training data; new deposition date alone does not imply remote homology.","source_ids":["src-discovery-maabuu-posebusters","evidence-discovery-final-posebusters"],"source_locator":"Pinned README: description; Usage; paper/data links; Methods: chemical, intramolecular and intermolecular validity; Methods: benchmark construction and evaluation of generalization"}],"facts":[{"label":"Datasets","value":"Predicted molecular coordinates, optionally with the conditioning protein, and paper-linked evaluation data.","status":"source_checked","source_ids":["src-discovery-maabuu-posebusters"],"source_locator":"Pinned README: description; Usage; paper/data links"},{"label":"Splits","value":"The evaluator scores supplied outputs; predictor train/test partitions belong to the dataset/run being evaluated.","status":"inapplicable","source_ids":["src-discovery-maabuu-posebusters"],"source_locator":"Pinned README: description; Usage; paper/data links"},{"label":"Metrics","value":"Chemical identity/stereochemistry, bond and angle geometry, aromatic planarity, internal clashes and protein–ligand clashes are checked separately. The paper evaluates native-like pose recovery jointly with passing the validity checks; its intramolecular tolerances are 25% for bond lengths/angles and 30% for nonbonded distances.","status":"source_checked","source_ids":["evidence-discovery-final-posebusters"],"source_locator":"Methods: chemical, intramolecular and intermolecular validity"},{"label":"Baselines","value":"The evaluator does not prescribe a universal baseline predictor; comparisons require methods run on the same selected dataset.","status":"inapplicable","source_ids":["src-discovery-maabuu-posebusters"],"source_locator":"Pinned README: description; Usage; paper/data links"},{"label":"Leakage controls","value":"The PoseBusters benchmark selects recent PDB complexes absent from the PDBbind v2020 training source used by evaluated learned docking methods. The paper further examines protein-sequence similarity to training data; new deposition date alone does not imply remote homology.","status":"source_checked","source_ids":["evidence-discovery-final-posebusters"],"source_locator":"Methods: benchmark construction and evaluation of generalization"},{"label":"Uncertainty","value":"Uncertainty across samples, datasets or training runs must be defined by the evaluation study; this evaluator entry does not fix one experiment.","status":"inapplicable","source_ids":["src-discovery-maabuu-posebusters"],"source_locator":"Pinned README: description; Usage; paper/data links"},{"label":"Entity type","value":"Molecular-pose plausibility evaluator.","status":"source_checked","source_ids":["src-discovery-maabuu-posebusters"],"source_locator":"Pinned README: description; Usage; paper/data links"},{"label":"Organisms","value":"Plausibility checks concern molecular geometry rather than organism identity.","status":"inapplicable","source_ids":["src-discovery-maabuu-posebusters"],"source_locator":"Pinned README: description; Usage; paper/data links"},{"label":"Assays","value":"Predicted poses assessed against molecular validity criteria.","status":"source_checked","source_ids":["src-discovery-maabuu-posebusters"],"source_locator":"Pinned README: description; Usage; paper/data links"},{"label":"Allowed inputs","value":"Predicted molecule coordinates, optionally paired with a protein structure.","status":"source_checked","source_ids":["src-discovery-maabuu-posebusters"],"source_locator":"Pinned README: description; Usage; paper/data links"},{"label":"Adaptation","value":"The checker evaluates poses; it does not train or fine-tune the pose predictor.","status":"inapplicable","source_ids":["src-discovery-maabuu-posebusters"],"source_locator":"Pinned README: description; Usage; paper/data links"}],"strengths":[{"text":"Checks plausibility separately from pose agreement, preventing the two criteria being conflated.","source_ids":["src-discovery-maabuu-posebusters"],"source_locator":"Pinned README: description; Usage; paper/data links"}],"limitations":[{"text":"PoseBusters checks physical plausibility and geometric agreement. Passing those checks does not establish binding affinity, biological activity or complete independence from every external training corpus.","source_ids":["evidence-discovery-final-posebusters"],"source_locator":"Methods: benchmark construction and evaluation of generalization"}],"diagram":{"title":"Evaluation procedure","steps":["Allowed inputs: Predicted molecule coordinates, optionally paired with a protein structure.","Splits: The evaluator scores supplied outputs; predictor train/test partitions belong to the dataset/run being evaluated.","Metrics: Chemical identity/stereochemistry, bond and angle geometry, aromatic planarity, internal clashes and protein–ligand clashes are checked separately. The paper evaluates native-like pose recovery jointly with passing the validity checks; its intramolecular tolerances are 25% for bond lengths/angles and 30% for nonbonded distances."],"caption":"Conceptual procedure. Task variants and protocol versions retain their separate scoring conditions.","source_ids":["src-discovery-maabuu-posebusters","evidence-discovery-final-posebusters"],"source_locator":"Pinned README: description; Usage; paper/data links; Methods: chemical, intramolecular and intermolecular validity"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction."}}},"description":"Geometric and chemical plausibility of molecular poses","facets":{"areas":["molecular-interactions"]},"id":"discovery-benchmark-posebusters","kind":"benchmark","links":[],"name":"PoseBusters","source_ids":["src-discovery-maabuu-posebusters"],"status":"discovered"}
{"attributes":{"entity_level":"suite","scope_note":"Specialist molecular or omics evaluation; protocol details require review before numerical comparison.","task":"Protein prediction, design and dynamics evaluation","version":null,"historical_missing_metadata":{"dataset_release":"unextracted","metric_implementation":"unextracted","split_manifest":"unextracted","version":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"ProteinBench assesses multiple protein-model tasks using quality, novelty, diversity and robustness dimensions.","summary_source_ids":["evidence-benchmark-proteinbench-snapshot"],"summary_source_locator":"ProteinBench official website: Abstract; sequence/structure task evaluations; metric descriptions","sections":[{"title":"Evaluation methodology","body":"ProteinBench evaluates protein sequence and structure methods across design, folding and conformational tasks. It distinguishes quality, diversity and novelty instead of treating all protein capabilities as one accuracy score. Each task has its own reference data, sampling procedure and baseline family.","source_ids":["evidence-discovery-final-proteinbench"],"source_locator":"Task definitions; ATLAS evaluation; antibody-design setup; result tables"}],"facts":[{"label":"Datasets","value":"Task-specific structure, sequence and complex evaluation collections.","status":"source_checked","source_ids":["evidence-benchmark-proteinbench-snapshot"],"source_locator":"ProteinBench official website: Abstract; sequence/structure task evaluations; metric descriptions"},{"label":"Splits","value":"The framework distinguishes natural in-distribution structures from generated-backbone out-of-distribution evaluations.","status":"source_checked","source_ids":["evidence-benchmark-proteinbench-snapshot"],"source_locator":"ProteinBench official website: Abstract; sequence/structure task evaluations; metric descriptions"},{"label":"Metrics","value":"Metric sets vary by task and include structure-predictor confidence, structural similarity and diversity measures.","status":"source_checked","source_ids":["evidence-benchmark-proteinbench-snapshot"],"source_locator":"ProteinBench official website: Abstract; sequence/structure task evaluations; metric descriptions"},{"label":"Baselines","value":"Comparators vary by task: structure predictors, inverse-folding/design models, Rosetta-based antibody methods and molecular-dynamics or ensemble references. The paper defines the eligible method and input information separately for each task.","status":"source_checked","source_ids":["evidence-discovery-final-proteinbench"],"source_locator":"Task definitions; ATLAS evaluation; antibody-design setup; result tables"},{"label":"Leakage controls","value":"The antibody setup clusters CDR-H3 sequences at 40% similarity and excludes clusters containing RAbD test complexes from training/validation. The ATLAS ensemble task also applies a held-out protocol for models trained on ATLAS; these are task-specific controls, not a universal suite split.","status":"source_checked","source_ids":["evidence-discovery-final-proteinbench"],"source_locator":"Task definitions; ATLAS evaluation; antibody-design setup; result tables"},{"label":"Uncertainty","value":"Some task tables report repeated-experiment averages and standard deviations; others report medians.","status":"source_checked","source_ids":["evidence-benchmark-proteinbench-snapshot"],"source_locator":"ProteinBench official website: Abstract; sequence/structure task evaluations; metric descriptions"},{"label":"Entity type","value":"Protein-model evaluation framework spanning several tasks.","status":"source_checked","source_ids":["evidence-benchmark-proteinbench-snapshot"],"source_locator":"ProteinBench official website: Abstract; sequence/structure task evaluations; metric descriptions"},{"label":"Organisms","value":"No single organism defines its sequence, structure and complex task collections.","status":"inapplicable","source_ids":["evidence-benchmark-proteinbench-snapshot"],"source_locator":"ProteinBench official website: Abstract; sequence/structure task evaluations; metric descriptions"},{"label":"Assays","value":"Task-dependent protein structure and sequence/property references.","status":"source_checked","source_ids":["evidence-benchmark-proteinbench-snapshot"],"source_locator":"ProteinBench official website: Abstract; sequence/structure task evaluations; metric descriptions"},{"label":"Allowed inputs","value":"Task-specific sequences, structures or complexes.","status":"source_checked","source_ids":["evidence-benchmark-proteinbench-snapshot"],"source_locator":"ProteinBench official website: Abstract; sequence/structure task evaluations; metric descriptions"},{"label":"Adaptation","value":"Tasks define their own generation, prediction or adaptation regimes.","status":"source_checked","source_ids":["evidence-benchmark-proteinbench-snapshot"],"source_locator":"ProteinBench official website: Abstract; sequence/structure task evaluations; metric descriptions"}],"strengths":[{"text":"Separate task collections expose which capability a result measures.","source_ids":["evidence-benchmark-proteinbench-snapshot"],"source_locator":"ProteinBench official website: Abstract; sequence/structure task evaluations; metric descriptions"}],"limitations":[{"text":"Architectures receive different inputs and can generate different numbers of candidates. Task-specific exclusion rules and sampling budgets must accompany any comparison.","source_ids":["evidence-discovery-final-proteinbench"],"source_locator":"Task definitions; ATLAS evaluation; antibody-design setup; result tables"}],"diagram":{"title":"Evaluation procedure","steps":["Allowed inputs: Task-specific sequences, structures or complexes.","Splits: The framework distinguishes natural in-distribution structures from generated-backbone out-of-distribution evaluations.","Metrics: Metric sets vary by task and include structure-predictor confidence, structural similarity and diversity measures."],"caption":"Conceptual procedure. Task variants and protocol versions retain their separate scoring conditions.","source_ids":["evidence-benchmark-proteinbench-snapshot"],"source_locator":"ProteinBench official website: Abstract; sequence/structure task evaluations; metric descriptions"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction."}}},"description":"Protein prediction, design and dynamics evaluation","facets":{"areas":["protein-structure"]},"id":"discovery-benchmark-proteinbench","kind":"benchmark","links":[],"name":"ProteinBench","source_ids":["src-discovery-proteinbench"],"status":"discovered"}
{"attributes":{"entity_level":"suite","scope_note":"Specialist molecular or omics evaluation; protocol details require review before numerical comparison.","task":"Protein variant effect prediction","version":null,"historical_missing_metadata":{"dataset_release":"unextracted","metric_implementation":"unextracted","split_manifest":"unextracted","version":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"ProteinGym separates experimental variant-effect and clinical annotation tasks under supervised and zero-shot regimes.","summary_source_ids":["src-discovery-oatml-markslab-proteingym"],"summary_source_locator":"Pinned README: benchmark data; performance metrics and aggregation","sections":[{"title":"Evaluation methodology","body":"DMS assays and human clinical variants, with substitution and indel collections kept separate. Five-fold random, contiguous-position and modulo-position cross-validation are separate supervised DMS regimes. The original clinical analysis uses available ClinVar labels with explicit overlap warnings; zero-shot scoring does not fit on assay labels. Zero-shot DMS: Spearman, NDCG, AUC, MCC and top-K recall; supervised DMS: Spearman/MSE; clinical: AUC. Aggregation first groups assays by UniProt ID, then averages functional categories. Supervised DMS evaluations distinguish five-fold random, contiguous-position and modulo-position partitions. The original clinical benchmark explicitly warns that supervised methods may overlap ClinVar labels and that population-frequency training can leak information into benign-variant evaluation.","source_ids":["src-discovery-oatml-markslab-proteingym","evidence-discovery-final-proteingym"],"source_locator":"Pinned README: benchmark data; performance metrics and aggregation; Supervised DMS benchmarking; Sections on supervised DMS and clinical benchmarking"}],"facts":[{"label":"Datasets","value":"DMS assays and human clinical variants, with substitution and indel collections kept separate.","status":"source_checked","source_ids":["src-discovery-oatml-markslab-proteingym"],"source_locator":"Pinned README: benchmark data; performance metrics and aggregation"},{"label":"Splits","value":"Five-fold random, contiguous-position and modulo-position cross-validation are separate supervised DMS regimes. The original clinical analysis uses available ClinVar labels with explicit overlap warnings; zero-shot scoring does not fit on assay labels.","status":"source_checked","source_ids":["evidence-discovery-final-proteingym"],"source_locator":"Supervised DMS benchmarking"},{"label":"Metrics","value":"Zero-shot DMS: Spearman, NDCG, AUC, MCC and top-K recall; supervised DMS: Spearman/MSE; clinical: AUC. Aggregation first groups assays by UniProt ID, then averages functional categories.","status":"source_checked","source_ids":["src-discovery-oatml-markslab-proteingym"],"source_locator":"Pinned README: benchmark data; performance metrics and aggregation"},{"label":"Baselines","value":"The README distinguishes sequence-only baselines such as ESM-1v, alignment-based approaches such as DeepSequence/EVE, and sequence-plus-structure approaches such as SaProt; clinical baselines use dbNSFP 4.4a.","status":"source_checked","source_ids":["src-discovery-oatml-markslab-proteingym"],"source_locator":"Pinned README: benchmark data; performance metrics and aggregation"},{"label":"Leakage controls","value":"Supervised DMS evaluations distinguish five-fold random, contiguous-position and modulo-position partitions. The original clinical benchmark explicitly warns that supervised methods may overlap ClinVar labels and that population-frequency training can leak information into benign-variant evaluation.","status":"source_checked","source_ids":["evidence-discovery-final-proteingym"],"source_locator":"Sections on supervised DMS and clinical benchmarking"},{"label":"Uncertainty","value":"Bootstrapped standard errors are provided for aggregate metrics.","status":"source_checked","source_ids":["src-discovery-oatml-markslab-proteingym"],"source_locator":"Pinned README: benchmark data; performance metrics and aggregation"},{"label":"Entity type","value":"Benchmark suite with separate DMS/clinical, substitution/indel and supervised/zero-shot tracks.","status":"source_checked","source_ids":["src-discovery-oatml-markslab-proteingym"],"source_locator":"Pinned README: benchmark data; performance metrics and aggregation"},{"label":"Organisms","value":"DMS collections span taxa; the clinical track concerns human proteins. Taxa-specific performance files are provided.","status":"source_checked","source_ids":["src-discovery-oatml-markslab-proteingym"],"source_locator":"Pinned README: benchmark data; performance metrics and aggregation"},{"label":"Assays","value":"Deep mutational scanning measurements and curated benign/pathogenic clinical annotations.","status":"source_checked","source_ids":["src-discovery-oatml-markslab-proteingym"],"source_locator":"Pinned README: benchmark data; performance metrics and aggregation"},{"label":"Allowed inputs","value":"Variant and target protein sequences; comparator modalities separately include alignments, structures and function annotations.","status":"source_checked","source_ids":["src-discovery-oatml-markslab-proteingym"],"source_locator":"Pinned README: benchmark data; performance metrics and aggregation"},{"label":"Adaptation","value":"Separate zero-shot scoring and supervised learning regimes; labeled-data access must follow the chosen track.","status":"source_checked","source_ids":["src-discovery-oatml-markslab-proteingym"],"source_locator":"Pinned README: benchmark data; performance metrics and aggregation"}],"strengths":[{"text":"Protein-level and functional-category aggregation reduces over-weighting of proteins with many assays.","source_ids":["src-discovery-oatml-markslab-proteingym"],"source_locator":"Pinned README: benchmark data; performance metrics and aggregation"}],"limitations":[{"text":"ProteinGym’s DMS cross-validation controls and clinical-label overlap risks are different. Zero-shot, supervised, substitution and indel tasks must remain separate, with original assay identities preserved.","source_ids":["evidence-discovery-final-proteingym"],"source_locator":"Sections on supervised DMS and clinical benchmarking"}],"diagram":{"title":"Evaluation procedure","steps":["Allowed inputs: Variant and target protein sequences; comparator modalities separately include alignments, structures and function annotations.","Splits: Five-fold random, contiguous-position and modulo-position cross-validation are separate supervised DMS regimes. The original clinical analysis uses available ClinVar labels with explicit overlap warnings; zero-shot scoring does not fit on assay labels.","Metrics: Zero-shot DMS: Spearman, NDCG, AUC, MCC and top-K recall; supervised DMS: Spearman/MSE; clinical: AUC. Aggregation first groups assays by UniProt ID, then averages functional categories."],"caption":"Conceptual procedure. Task variants and protocol versions retain their separate scoring conditions.","source_ids":["src-discovery-oatml-markslab-proteingym","evidence-discovery-final-proteingym"],"source_locator":"Pinned README: benchmark data; performance metrics and aggregation; Supervised DMS benchmarking"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction."}}},"description":"Protein variant effect prediction","facets":{"areas":["protein-function"]},"id":"discovery-benchmark-proteingym","kind":"benchmark","links":[{"relation":"evaluates_task","target_id":"catalog-task-proteingym-effects"}],"name":"ProteinGym","source_ids":["src-discovery-oatml-markslab-proteingym"],"status":"discovered"}
{"attributes":{"entity_level":"suite","scope_note":"Specialist molecular or omics evaluation; protocol details require review before numerical comparison.","task":"Single-cell integration evaluation","version":null,"historical_missing_metadata":{"dataset_release":"unextracted","metric_implementation":"unextracted","split_manifest":"unextracted","version":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"scIB evaluates single-cell integration by checking batch removal and biological conservation separately.","summary_source_ids":["src-discovery-theislab-scib"],"summary_source_locator":"Pinned README: package purpose; Metrics; Integration Tools","sections":[{"title":"Evaluation methodology","body":"Annotated single-cell data in AnnData form, with preprocessing and integration output types selected for the evaluation. The evaluator scores supplied outputs; predictor train/test partitions belong to the dataset/run being evaluated. The metric module separates batch-correction and biological-conservation measures. Listed integrations include Harmony, MNN/FastMNN, scVI/scANVI, Scanorama, BBKNN and Seurat. Uncertainty across samples, datasets or training runs must be defined by the evaluation study; this evaluator entry does not fix one experiment.","source_ids":["src-discovery-theislab-scib"],"source_locator":"Pinned README: package purpose; Metrics; Integration Tools"}],"facts":[{"label":"Datasets","value":"Annotated single-cell data in AnnData form, with preprocessing and integration output types selected for the evaluation.","status":"source_checked","source_ids":["src-discovery-theislab-scib"],"source_locator":"Pinned README: package purpose; Metrics; Integration Tools"},{"label":"Splits","value":"The evaluator scores supplied outputs; predictor train/test partitions belong to the dataset/run being evaluated.","status":"inapplicable","source_ids":["src-discovery-theislab-scib"],"source_locator":"Pinned README: package purpose; Metrics; Integration Tools"},{"label":"Metrics","value":"The metric module separates batch-correction and biological-conservation measures.","status":"source_checked","source_ids":["src-discovery-theislab-scib"],"source_locator":"Pinned README: package purpose; Metrics; Integration Tools"},{"label":"Baselines","value":"Listed integrations include Harmony, MNN/FastMNN, scVI/scANVI, Scanorama, BBKNN and Seurat.","status":"source_checked","source_ids":["src-discovery-theislab-scib"],"source_locator":"Pinned README: package purpose; Metrics; Integration Tools"},{"label":"Leakage controls","value":"scIB evaluates integration of the supplied batches together. Some methods use cell-type labels and others do not; the paper reports this distinction. This transductive integration setting is not a held-out-cell classifier test, so supervised train/test leakage terminology cannot be applied without specifying the method.","status":"source_checked","source_ids":["evidence-discovery-final-scib"],"source_locator":"Methods and Results: integration inputs, label use and biological-conservation metrics"},{"label":"Uncertainty","value":"Uncertainty across samples, datasets or training runs must be defined by the evaluation study; this evaluator entry does not fix one experiment.","status":"inapplicable","source_ids":["src-discovery-theislab-scib"],"source_locator":"Pinned README: package purpose; Metrics; Integration Tools"},{"label":"Entity type","value":"Single-cell integration evaluator.","status":"source_checked","source_ids":["src-discovery-theislab-scib"],"source_locator":"Pinned README: package purpose; Metrics; Integration Tools"},{"label":"Organisms","value":"Dataset choice supplies the organism; the metric package does not prescribe one.","status":"inapplicable","source_ids":["src-discovery-theislab-scib"],"source_locator":"Pinned README: package purpose; Metrics; Integration Tools"},{"label":"Assays","value":"Single-cell expression with batch and biological annotations.","status":"source_checked","source_ids":["src-discovery-theislab-scib"],"source_locator":"Pinned README: package purpose; Metrics; Integration Tools"},{"label":"Allowed inputs","value":"AnnData and integration outputs; label-dependent metrics additionally require biological labels.","status":"source_checked","source_ids":["src-discovery-theislab-scib"],"source_locator":"Pinned README: package purpose; Metrics; Integration Tools"},{"label":"Adaptation","value":"scIB scores integration outputs; adaptation occurs in the compared integration methods.","status":"inapplicable","source_ids":["src-discovery-theislab-scib"],"source_locator":"Pinned README: package purpose; Metrics; Integration Tools"}],"strengths":[{"text":"Separate batch-removal and biological-conservation scores reveal their trade-off.","source_ids":["src-discovery-theislab-scib"],"source_locator":"Pinned README: package purpose; Metrics; Integration Tools"}],"limitations":[{"text":"Batch mixing and preservation of biological variation must be considered together. Annotation-assisted methods and unsupervised methods receive different input information, and integration of known batches does not establish transfer to unseen batches.","source_ids":["evidence-discovery-final-scib"],"source_locator":"Methods and Results: integration inputs, label use and biological-conservation metrics"}],"diagram":{"title":"Evaluation procedure","steps":["Allowed inputs: AnnData and integration outputs; label-dependent metrics additionally require biological labels.","Splits: The evaluator scores supplied outputs; predictor train/test partitions belong to the dataset/run being evaluated.","Metrics: The metric module separates batch-correction and biological-conservation measures."],"caption":"Conceptual procedure. Task variants and protocol versions retain their separate scoring conditions.","source_ids":["src-discovery-theislab-scib"],"source_locator":"Pinned README: package purpose; Metrics; Integration Tools"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction."}}},"description":"Single-cell integration evaluation","facets":{"areas":["single-cell"]},"id":"discovery-benchmark-scib","kind":"benchmark","links":[{"relation":"evaluates_task","target_id":"catalog-task-cell-batch-integration"}],"name":"scIB","source_ids":["src-discovery-theislab-scib"],"status":"discovered"}
{"attributes":{"entity_level":"evaluator","scope_note":"Specialist molecular or omics evaluation; protocol details require review before numerical comparison.","task":"Perturbation prediction metric calibration","version":null,"historical_missing_metadata":{"dataset_release":"unextracted","metric_implementation":"unextracted","split_manifest":"unextracted","version":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"scPertEval evaluates and calibrates scoring protocols for single-cell perturbation predictions.","summary_source_ids":["src-discovery-virtual-cell-research-community-scperteval"],"summary_source_locator":"Pinned README: toolkit purpose; scoring/calibration/DE actions; protocol taxonomy","sections":[{"title":"Evaluation methodology","body":"scPertEval supplies explicit scoring protocols for perturbation predictions. A metric can compare one perturbation or operate across the complete perturbation panel, depending on its declared scope. Ground-truth references and context are passed to the evaluator, so protocol and preprocessing choices remain part of the result.","source_ids":["evidence-discovery-final-scperteval0"],"source_locator":"Pinned src/scperteval/protocols/metrics.py: metric input contract and context"}],"facts":[{"label":"Datasets","value":"Predicted and observed perturbation responses; the associated study evaluates protocols across public datasets.","status":"source_checked","source_ids":["src-discovery-virtual-cell-research-community-scperteval"],"source_locator":"Pinned README: toolkit purpose; scoring/calibration/DE actions; protocol taxonomy"},{"label":"Splits","value":"The evaluator scores supplied outputs; predictor train/test partitions belong to the dataset/run being evaluated.","status":"inapplicable","source_ids":["src-discovery-virtual-cell-research-community-scperteval"],"source_locator":"Pinned README: toolkit purpose; scoring/calibration/DE actions; protocol taxonomy"},{"label":"Metrics","value":"Protocol scoring is separated from calibration against empirical positive/negative controls using DRF and BDS.","status":"source_checked","source_ids":["src-discovery-virtual-cell-research-community-scperteval"],"source_locator":"Pinned README: toolkit purpose; scoring/calibration/DE actions; protocol taxonomy"},{"label":"Baselines","value":"Empirical controls calibrate how well a protocol distinguishes expected response quality.","status":"source_checked","source_ids":["src-discovery-virtual-cell-research-community-scperteval"],"source_locator":"Pinned README: toolkit purpose; scoring/calibration/DE actions; protocol taxonomy"},{"label":"Leakage controls","value":"The metric implementation receives ground truth, predictions and an evaluation context; it does not construct model-training partitions or audit training data. Predictor leakage controls belong to the protocol and dataset used to produce the submitted predictions.","status":"inapplicable","source_ids":["evidence-discovery-final-scperteval0"],"source_locator":"Pinned src/scperteval/protocols/metrics.py: metric input contract and context"},{"label":"Uncertainty","value":"Uncertainty across samples, datasets or training runs must be defined by the evaluation study; this evaluator entry does not fix one experiment.","status":"inapplicable","source_ids":["src-discovery-virtual-cell-research-community-scperteval"],"source_locator":"Pinned README: toolkit purpose; scoring/calibration/DE actions; protocol taxonomy"},{"label":"Entity type","value":"Perturbation scoring-protocol calibration toolkit.","status":"source_checked","source_ids":["src-discovery-virtual-cell-research-community-scperteval"],"source_locator":"Pinned README: toolkit purpose; scoring/calibration/DE actions; protocol taxonomy"},{"label":"Organisms","value":"Organism scope belongs to the selected perturbation dataset.","status":"inapplicable","source_ids":["src-discovery-virtual-cell-research-community-scperteval"],"source_locator":"Pinned README: toolkit purpose; scoring/calibration/DE actions; protocol taxonomy"},{"label":"Assays","value":"Observed single-cell perturbation responses and empirical positive/negative controls.","status":"source_checked","source_ids":["src-discovery-virtual-cell-research-community-scperteval"],"source_locator":"Pinned README: toolkit purpose; scoring/calibration/DE actions; protocol taxonomy"},{"label":"Allowed inputs","value":"Predicted/observed responses plus a protocol specifying representation, metric, transformation and reporting.","status":"source_checked","source_ids":["src-discovery-virtual-cell-research-community-scperteval"],"source_locator":"Pinned README: toolkit purpose; scoring/calibration/DE actions; protocol taxonomy"},{"label":"Adaptation","value":"The toolkit scores and calibrates evaluation protocols; it does not impose predictor fine-tuning.","status":"inapplicable","source_ids":["src-discovery-virtual-cell-research-community-scperteval"],"source_locator":"Pinned README: toolkit purpose; scoring/calibration/DE actions; protocol taxonomy"}],"strengths":[{"text":"Calibration tests whether a score distinguishes empirical controls before using it to rank models.","source_ids":["src-discovery-virtual-cell-research-community-scperteval"],"source_locator":"Pinned README: toolkit purpose; scoring/calibration/DE actions; protocol taxonomy"}],"limitations":[{"text":"Metric computation alone cannot establish whether a predictor saw a held-out perturbation or cell context during training. That evidence must come from the model and dataset protocol.","source_ids":["evidence-discovery-final-scperteval0"],"source_locator":"Pinned src/scperteval/protocols/metrics.py: metric input contract and context"}],"diagram":{"title":"Evaluation procedure","steps":["Allowed inputs: Predicted/observed responses plus a protocol specifying representation, metric, transformation and reporting.","Splits: The evaluator scores supplied outputs; predictor train/test partitions belong to the dataset/run being evaluated.","Metrics: Protocol scoring is separated from calibration against empirical positive/negative controls using DRF and BDS."],"caption":"Conceptual procedure. Task variants and protocol versions retain their separate scoring conditions.","source_ids":["src-discovery-virtual-cell-research-community-scperteval"],"source_locator":"Pinned README: toolkit purpose; scoring/calibration/DE actions; protocol taxonomy"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction."}}},"description":"Perturbation prediction metric calibration","facets":{"areas":["single-cell"]},"id":"discovery-benchmark-scperteval","kind":"benchmark","links":[],"name":"scPertEval","source_ids":["src-discovery-virtual-cell-research-community-scperteval"],"status":"discovered"}
{"attributes":{"entity_level":"suite","scope_note":"Specialist molecular or omics evaluation; protocol details require review before numerical comparison.","task":"Protein representation learning tasks","version":null,"historical_missing_metadata":{"dataset_release":"unextracted","metric_implementation":"unextracted","split_manifest":"unextracted","version":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"TAPE evaluates protein representations through five supervised downstream tasks.","summary_source_ids":["src-discovery-songlab-cal-tape"],"summary_source_locator":"Pinned README: compatibility notice; Evaluating a Downstream Model; Data; task leaderboards","sections":[{"title":"Evaluation methodology","body":"TAPE evaluates protein representations on five tasks: local secondary structure, contacts, remote homology, fluorescence and stability. Each has a biologically motivated supervised partition, and the original paper compares pretrained representations with untrained and alignment-based controls using task-specific prediction heads.","source_ids":["evidence-discovery-final-tape"],"source_locator":"Section 3; Appendix A.1"}],"facts":[{"label":"Datasets","value":"Secondary structure, contacts, remote homology, fluorescence and stability; a Pfam pretraining corpus is supplied separately.","status":"source_checked","source_ids":["src-discovery-songlab-cal-tape"],"source_locator":"Pinned README: compatibility notice; Evaluating a Downstream Model; Data; task leaderboards"},{"label":"Splits","value":"Secondary structure uses 25% identity filtering; contacts use ProteinNet/CASP12 with 30% filtering; remote homology holds out superfamilies; fluorescence holds out greater mutation distances; stability holds out selected mutation neighbourhoods.","status":"source_checked","source_ids":["evidence-discovery-final-tape"],"source_locator":"Section 3; Appendix A.1"},{"label":"Metrics","value":"Task leaderboards use three-class accuracy, contact ranking, top-1 homology accuracy and Spearman correlation for fluorescence/stability.","status":"source_checked","source_ids":["src-discovery-songlab-cal-tape"],"source_locator":"Pinned README: compatibility notice; Evaluating a Downstream Model; Data; task leaderboards"},{"label":"Baselines","value":"Transformer, LSTM, ResNet, UniRep and one-hot baselines.","status":"source_checked","source_ids":["src-discovery-songlab-cal-tape"],"source_locator":"Pinned README: compatibility notice; Evaluating a Downstream Model; Data; task leaderboards"},{"label":"Leakage controls","value":"The five tasks use different supervised generalization boundaries. Protein identity, evolutionary groups and mutational distance are not interchangeable, and none is a universal audit of unsupervised pretraining overlap.","status":"source_checked","source_ids":["evidence-discovery-final-tape"],"source_locator":"Section 3; Appendix A.1"},{"label":"Uncertainty","value":"The original paper reports point estimates in its task result tables. Methods and task appendices do not define a suite-wide repeated-seed or bootstrap interval; individual later evaluations must supply their own uncertainty.","status":"unreported","source_ids":["evidence-discovery-final-tape"],"source_locator":"Section 3; Appendix A.1; Tables 1–2"},{"label":"Entity type","value":"Protein-representation benchmark suite.","status":"source_checked","source_ids":["src-discovery-songlab-cal-tape"],"source_locator":"Pinned README: compatibility notice; Evaluating a Downstream Model; Data; task leaderboards"},{"label":"Organisms","value":"Mixed protein-domain and structural sources rather than a species-held-out benchmark; the engineering tasks use green fluorescent protein variants and designed stability landscapes.","status":"source_checked","source_ids":["evidence-discovery-final-tape"],"source_locator":"Section 3; Appendix A.1; Tables 1–2"},{"label":"Assays","value":"Protein structure/homology annotations and experimental fluorescence/stability measurements.","status":"source_checked","source_ids":["src-discovery-songlab-cal-tape"],"source_locator":"Pinned README: compatibility notice; Evaluating a Downstream Model; Data; task leaderboards"},{"label":"Allowed inputs","value":"Protein amino-acid sequences and task-specific labels.","status":"source_checked","source_ids":["src-discovery-songlab-cal-tape"],"source_locator":"Pinned README: compatibility notice; Evaluating a Downstream Model; Data; task leaderboards"},{"label":"Adaptation","value":"Unsupervised pretraining followed by supervised downstream training; the README warns that downstream hyperparameters require task-specific tuning.","status":"source_checked","source_ids":["src-discovery-songlab-cal-tape"],"source_locator":"Pinned README: compatibility notice; Evaluating a Downstream Model; Data; task leaderboards"}],"strengths":[{"text":"Distinct sequence, residue and protein-level tasks avoid relying on language-model perplexity as a proxy for transfer.","source_ids":["src-discovery-songlab-cal-tape"],"source_locator":"Pinned README: compatibility notice; Evaluating a Downstream Model; Data; task leaderboards"}],"limitations":[{"text":"The original TAPE paper and the later PyTorch implementation are distinct versions. Preserve the task split and implementation; pretraining exposure is separate from supervised sequence-identity filtering.","source_ids":["evidence-discovery-final-tape"],"source_locator":"Section 3; Appendix A.1; Tables 1–2"}],"diagram":{"title":"Evaluation procedure","steps":["Allowed inputs: Protein amino-acid sequences and task-specific labels.","Splits: Secondary structure uses 25% identity filtering; contacts use ProteinNet/CASP12 with 30% filtering; remote homology holds out superfamilies; fluorescence holds out greater mutation distances; stability holds out selected mutation neighbourhoods.","Metrics: Task leaderboards use three-class accuracy, contact ranking, top-1 homology accuracy and Spearman correlation for fluorescence/stability."],"caption":"Conceptual procedure. Task variants and protocol versions retain their separate scoring conditions.","source_ids":["src-discovery-songlab-cal-tape","evidence-discovery-final-tape"],"source_locator":"Pinned README: compatibility notice; Evaluating a Downstream Model; Data; task leaderboards; Section 3; Appendix A.1"},"coverage":"limited","gaps":["Uncertainty: The original paper reports point estimates in its task result tables. Methods and task appendices do not define a suite-wide repeated-seed or bootstrap interval; individual later evaluations must supply their own uncertainty."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction."}}},"description":"Protein representation learning tasks","facets":{"areas":["protein-function"]},"id":"discovery-benchmark-tape","kind":"benchmark","links":[],"name":"TAPE","source_ids":["src-discovery-songlab-cal-tape"],"status":"discovered"}
{"attributes":{"entity_level":"task","scope_note":"Specialist molecular or omics evaluation; protocol details require review before numerical comparison.","task":"Contact Prediction","version":null,"historical_missing_metadata":{"dataset_release":"unextracted","metric_implementation":"unextracted","split_manifest":"unextracted","version":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"The TAPE protein residue-contact prediction task evaluates a trained protein representation.","summary_source_ids":["src-discovery-songlab-cal-tape"],"summary_source_locator":"Pinned README: compatibility notice; Evaluating a Downstream Model; Data; task leaderboards; README task-specific leaderboard and data-source references","sections":[{"title":"Evaluation methodology","body":"ProteinNet training/validation partitions filtered at 30% sequence identity; the held-out evaluation uses CASP12 targets. Sequence-identity filtering and the CASP12 target set define the supervised generalization boundary.","source_ids":["evidence-discovery-final-tape"],"source_locator":"Section 3 task definitions; Appendix A.1.1–A.1.5"}],"facts":[{"label":"Datasets","value":"ProteinNet structural data with CASP12 test targets.","status":"source_checked","source_ids":["evidence-discovery-final-tape"],"source_locator":"Appendix A.1"},{"label":"Splits","value":"ProteinNet training/validation partitions filtered at 30% sequence identity; the held-out evaluation uses CASP12 targets.","status":"source_checked","source_ids":["evidence-discovery-final-tape"],"source_locator":"Section 3 task definitions; Appendix A.1.1–A.1.5"},{"label":"Metrics","value":"Precision at L/5 for medium- and long-range contacts.","status":"source_checked","source_ids":["src-discovery-songlab-cal-tape"],"source_locator":"Pinned README: compatibility notice; Evaluating a Downstream Model; Data; task leaderboards; README task-specific leaderboard and data-source references"},{"label":"Baselines","value":"Task leaderboard comparisons include Transformer, LSTM, UniRep, ResNet, Bepler and one-hot baselines; alignment-augmented references appear where applicable.","status":"source_checked","source_ids":["src-discovery-songlab-cal-tape"],"source_locator":"Pinned README: compatibility notice; Evaluating a Downstream Model; Data; task leaderboards; README task-specific leaderboard and data-source references"},{"label":"Leakage controls","value":"Sequence-identity filtering and the CASP12 target set define the supervised generalization boundary.","status":"source_checked","source_ids":["evidence-discovery-final-tape"],"source_locator":"Section 3 task definitions; Appendix A.1.1–A.1.5"},{"label":"Uncertainty","value":"The original paper reports point estimates in its task result tables. Methods and task appendices do not define a suite-wide repeated-seed or bootstrap interval; individual later evaluations must supply their own uncertainty.","status":"unreported","source_ids":["evidence-discovery-final-tape"],"source_locator":"Section 3; Appendix A.1; Tables 1–2"},{"label":"Entity type","value":"Constituent benchmark task: TAPE Contact Prediction","status":"source_checked","source_ids":["src-discovery-songlab-cal-tape"],"source_locator":"Pinned README: compatibility notice; Evaluating a Downstream Model; Data; task leaderboards; README task-specific leaderboard and data-source references"},{"label":"Organisms","value":"Mixed-organism ProteinNet/CASP protein structures; organism identity is not the partitioning key.","status":"source_checked","source_ids":["evidence-discovery-final-tape"],"source_locator":"Section 3 task definitions; Appendix A.1.1–A.1.5"},{"label":"Assays","value":"Protein structure/homology annotations and experimental fluorescence/stability measurements.","status":"source_checked","source_ids":["src-discovery-songlab-cal-tape"],"source_locator":"Pinned README: compatibility notice; Evaluating a Downstream Model; Data; task leaderboards; README task-specific leaderboard and data-source references"},{"label":"Allowed inputs","value":"Protein sequence with a residue-contact target derived from ProteinNet.","status":"source_checked","source_ids":["src-discovery-songlab-cal-tape"],"source_locator":"Pinned README: compatibility notice; Evaluating a Downstream Model; Data; task leaderboards; README task-specific leaderboard and data-source references"},{"label":"Adaptation","value":"Unsupervised pretraining followed by supervised downstream training; the README warns that downstream hyperparameters require task-specific tuning.","status":"source_checked","source_ids":["src-discovery-songlab-cal-tape"],"source_locator":"Pinned README: compatibility notice; Evaluating a Downstream Model; Data; task leaderboards; README task-specific leaderboard and data-source references"}],"strengths":[{"text":"Distinct sequence, residue and protein-level tasks avoid relying on language-model perplexity as a proxy for transfer.","source_ids":["src-discovery-songlab-cal-tape"],"source_locator":"Pinned README: compatibility notice; Evaluating a Downstream Model; Data; task leaderboards; README task-specific leaderboard and data-source references"}],"limitations":[{"text":"The original TAPE paper and the later PyTorch implementation are distinct versions. Preserve the task split and implementation; pretraining exposure is separate from supervised sequence-identity filtering.","source_ids":["evidence-discovery-final-tape"],"source_locator":"Section 3; Appendix A.1; Tables 1–2"}],"diagram":{"title":"Evaluation procedure","steps":["Allowed inputs: Protein sequence with a residue-contact target derived from ProteinNet.","Splits: ProteinNet training/validation partitions filtered at 30% sequence identity; the held-out evaluation uses CASP12 targets.","Metrics: Precision at L/5 for medium- and long-range contacts."],"caption":"Conceptual procedure. Task variants and protocol versions retain their separate scoring conditions.","source_ids":["src-discovery-songlab-cal-tape","evidence-discovery-final-tape"],"source_locator":"Pinned README: compatibility notice; Evaluating a Downstream Model; Data; task leaderboards; README task-specific leaderboard and data-source references; Section 3 task definitions; Appendix A.1.1–A.1.5"},"coverage":"limited","gaps":["Uncertainty: The original paper reports point estimates in its task result tables. Methods and task appendices do not define a suite-wide repeated-seed or bootstrap interval; individual later evaluations must supply their own uncertainty."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction."}}},"description":"Contact Prediction","facets":{"areas":["protein-function"]},"id":"discovery-benchmark-tape-contact-prediction","kind":"benchmark","links":[{"relation":"parent","target_id":"discovery-benchmark-tape"},{"relation":"part_of","target_id":"discovery-benchmark-tape"}],"name":"TAPE Contact Prediction","source_ids":["src-discovery-songlab-cal-tape"],"status":"discovered"}
{"attributes":{"entity_level":"task","scope_note":"Specialist molecular or omics evaluation; protocol details require review before numerical comparison.","task":"Fluorescence","version":null,"historical_missing_metadata":{"dataset_release":"unextracted","metric_implementation":"unextracted","split_manifest":"unextracted","version":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"The TAPE protein fluorescence prediction task evaluates a trained protein representation.","summary_source_ids":["src-discovery-songlab-cal-tape"],"summary_source_locator":"Pinned README: compatibility notice; Evaluating a Downstream Model; Data; task leaderboards; README task-specific leaderboard and data-source references","sections":[{"title":"Evaluation methodology","body":"Training and validation use GFP variants within three mutations of the parent; testing uses variants with four to fifteen mutations. Mutation-distance separation tests extrapolation away from the same parent protein rather than independence of protein families.","source_ids":["evidence-discovery-final-tape"],"source_locator":"Section 3 task definitions; Appendix A.1.1–A.1.5"}],"facts":[{"label":"Datasets","value":"Sarkisyan green fluorescent protein mutagenesis assay, using the original TAPE mutation-distance partition.","status":"source_checked","source_ids":["evidence-discovery-final-tape"],"source_locator":"Appendix A.1"},{"label":"Splits","value":"Training and validation use GFP variants within three mutations of the parent; testing uses variants with four to fifteen mutations.","status":"source_checked","source_ids":["evidence-discovery-final-tape"],"source_locator":"Section 3 task definitions; Appendix A.1.1–A.1.5"},{"label":"Metrics","value":"Spearman rank correlation.","status":"source_checked","source_ids":["src-discovery-songlab-cal-tape"],"source_locator":"Pinned README: compatibility notice; Evaluating a Downstream Model; Data; task leaderboards; README task-specific leaderboard and data-source references"},{"label":"Baselines","value":"Task leaderboard comparisons include Transformer, LSTM, UniRep, ResNet, Bepler and one-hot baselines; alignment-augmented references appear where applicable.","status":"source_checked","source_ids":["src-discovery-songlab-cal-tape"],"source_locator":"Pinned README: compatibility notice; Evaluating a Downstream Model; Data; task leaderboards; README task-specific leaderboard and data-source references"},{"label":"Leakage controls","value":"Mutation-distance separation tests extrapolation away from the same parent protein rather than independence of protein families.","status":"source_checked","source_ids":["evidence-discovery-final-tape"],"source_locator":"Section 3 task definitions; Appendix A.1.1–A.1.5"},{"label":"Uncertainty","value":"The original paper reports point estimates in its task result tables. Methods and task appendices do not define a suite-wide repeated-seed or bootstrap interval; individual later evaluations must supply their own uncertainty.","status":"unreported","source_ids":["evidence-discovery-final-tape"],"source_locator":"Section 3; Appendix A.1; Tables 1–2"},{"label":"Entity type","value":"Constituent benchmark task: TAPE Fluorescence","status":"source_checked","source_ids":["src-discovery-songlab-cal-tape"],"source_locator":"Pinned README: compatibility notice; Evaluating a Downstream Model; Data; task leaderboards; README task-specific leaderboard and data-source references"},{"label":"Organisms","value":"Variants of the parent green fluorescent protein in the Sarkisyan mutagenesis assay.","status":"source_checked","source_ids":["evidence-discovery-final-tape"],"source_locator":"Section 3 task definitions; Appendix A.1.1–A.1.5"},{"label":"Assays","value":"Protein structure/homology annotations and experimental fluorescence/stability measurements.","status":"source_checked","source_ids":["src-discovery-songlab-cal-tape"],"source_locator":"Pinned README: compatibility notice; Evaluating a Downstream Model; Data; task leaderboards; README task-specific leaderboard and data-source references"},{"label":"Allowed inputs","value":"Protein sequence with a measured fluorescence target.","status":"source_checked","source_ids":["src-discovery-songlab-cal-tape"],"source_locator":"Pinned README: compatibility notice; Evaluating a Downstream Model; Data; task leaderboards; README task-specific leaderboard and data-source references"},{"label":"Adaptation","value":"Unsupervised pretraining followed by supervised downstream training; the README warns that downstream hyperparameters require task-specific tuning.","status":"source_checked","source_ids":["src-discovery-songlab-cal-tape"],"source_locator":"Pinned README: compatibility notice; Evaluating a Downstream Model; Data; task leaderboards; README task-specific leaderboard and data-source references"}],"strengths":[{"text":"Distinct sequence, residue and protein-level tasks avoid relying on language-model perplexity as a proxy for transfer.","source_ids":["src-discovery-songlab-cal-tape"],"source_locator":"Pinned README: compatibility notice; Evaluating a Downstream Model; Data; task leaderboards; README task-specific leaderboard and data-source references"}],"limitations":[{"text":"The original TAPE paper and the later PyTorch implementation are distinct versions. Preserve the task split and implementation; pretraining exposure is separate from supervised sequence-identity filtering.","source_ids":["evidence-discovery-final-tape"],"source_locator":"Section 3; Appendix A.1; Tables 1–2"}],"diagram":{"title":"Evaluation procedure","steps":["Allowed inputs: Protein sequence with a measured fluorescence target.","Splits: Training and validation use GFP variants within three mutations of the parent; testing uses variants with four to fifteen mutations.","Metrics: Spearman rank correlation."],"caption":"Conceptual procedure. Task variants and protocol versions retain their separate scoring conditions.","source_ids":["src-discovery-songlab-cal-tape","evidence-discovery-final-tape"],"source_locator":"Pinned README: compatibility notice; Evaluating a Downstream Model; Data; task leaderboards; README task-specific leaderboard and data-source references; Section 3 task definitions; Appendix A.1.1–A.1.5"},"coverage":"limited","gaps":["Uncertainty: The original paper reports point estimates in its task result tables. Methods and task appendices do not define a suite-wide repeated-seed or bootstrap interval; individual later evaluations must supply their own uncertainty."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction."}}},"description":"Fluorescence","facets":{"areas":["protein-function"]},"id":"discovery-benchmark-tape-fluorescence","kind":"benchmark","links":[{"relation":"parent","target_id":"discovery-benchmark-tape"},{"relation":"part_of","target_id":"discovery-benchmark-tape"}],"name":"TAPE Fluorescence","source_ids":["src-discovery-songlab-cal-tape"],"status":"discovered"}
{"attributes":{"entity_level":"task","scope_note":"Specialist molecular or omics evaluation; protocol details require review before numerical comparison.","task":"Remote Homology Detection","version":null,"historical_missing_metadata":{"dataset_release":"unextracted","metric_implementation":"unextracted","split_manifest":"unextracted","version":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"The TAPE remote protein-homology classification task evaluates a trained protein representation.","summary_source_ids":["src-discovery-songlab-cal-tape"],"summary_source_locator":"Pinned README: compatibility notice; Evaluating a Downstream Model; Data; task leaderboards; README task-specific leaderboard and data-source references","sections":[{"title":"Evaluation methodology","body":"SCOP 1.75 protein domains are grouped by evolutionary hierarchy; entire superfamilies are held out for fold-level classification. Holding out superfamilies tests remote homologues without transferring examples from the same superfamily between train and test.","source_ids":["evidence-discovery-final-tape"],"source_locator":"Section 3 task definitions; Appendix A.1.1–A.1.5"}],"facts":[{"label":"Datasets","value":"SCOP 1.75 domains with fold labels and held-out superfamilies.","status":"source_checked","source_ids":["evidence-discovery-final-tape"],"source_locator":"Appendix A.1"},{"label":"Splits","value":"SCOP 1.75 protein domains are grouped by evolutionary hierarchy; entire superfamilies are held out for fold-level classification.","status":"source_checked","source_ids":["evidence-discovery-final-tape"],"source_locator":"Section 3 task definitions; Appendix A.1.1–A.1.5"},{"label":"Metrics","value":"Top-1 class accuracy.","status":"source_checked","source_ids":["src-discovery-songlab-cal-tape"],"source_locator":"Pinned README: compatibility notice; Evaluating a Downstream Model; Data; task leaderboards; README task-specific leaderboard and data-source references"},{"label":"Baselines","value":"Task leaderboard comparisons include Transformer, LSTM, UniRep, ResNet, Bepler and one-hot baselines; alignment-augmented references appear where applicable.","status":"source_checked","source_ids":["src-discovery-songlab-cal-tape"],"source_locator":"Pinned README: compatibility notice; Evaluating a Downstream Model; Data; task leaderboards; README task-specific leaderboard and data-source references"},{"label":"Leakage controls","value":"Holding out superfamilies tests remote homologues without transferring examples from the same superfamily between train and test.","status":"source_checked","source_ids":["evidence-discovery-final-tape"],"source_locator":"Section 3 task definitions; Appendix A.1.1–A.1.5"},{"label":"Uncertainty","value":"The original paper reports point estimates in its task result tables. Methods and task appendices do not define a suite-wide repeated-seed or bootstrap interval; individual later evaluations must supply their own uncertainty.","status":"unreported","source_ids":["evidence-discovery-final-tape"],"source_locator":"Section 3; Appendix A.1; Tables 1–2"},{"label":"Entity type","value":"Constituent benchmark task: TAPE Remote Homology Detection","status":"source_checked","source_ids":["src-discovery-songlab-cal-tape"],"source_locator":"Pinned README: compatibility notice; Evaluating a Downstream Model; Data; task leaderboards; README task-specific leaderboard and data-source references"},{"label":"Organisms","value":"SCOP protein-domain collection across organisms; the evaluated label is structural fold, not species.","status":"source_checked","source_ids":["evidence-discovery-final-tape"],"source_locator":"Section 3 task definitions; Appendix A.1.1–A.1.5"},{"label":"Assays","value":"Protein structure/homology annotations and experimental fluorescence/stability measurements.","status":"source_checked","source_ids":["src-discovery-songlab-cal-tape"],"source_locator":"Pinned README: compatibility notice; Evaluating a Downstream Model; Data; task leaderboards; README task-specific leaderboard and data-source references"},{"label":"Allowed inputs","value":"Protein sequence with a remote-homology class.","status":"source_checked","source_ids":["src-discovery-songlab-cal-tape"],"source_locator":"Pinned README: compatibility notice; Evaluating a Downstream Model; Data; task leaderboards; README task-specific leaderboard and data-source references"},{"label":"Adaptation","value":"Unsupervised pretraining followed by supervised downstream training; the README warns that downstream hyperparameters require task-specific tuning.","status":"source_checked","source_ids":["src-discovery-songlab-cal-tape"],"source_locator":"Pinned README: compatibility notice; Evaluating a Downstream Model; Data; task leaderboards; README task-specific leaderboard and data-source references"}],"strengths":[{"text":"Distinct sequence, residue and protein-level tasks avoid relying on language-model perplexity as a proxy for transfer.","source_ids":["src-discovery-songlab-cal-tape"],"source_locator":"Pinned README: compatibility notice; Evaluating a Downstream Model; Data; task leaderboards; README task-specific leaderboard and data-source references"}],"limitations":[{"text":"The original TAPE paper and the later PyTorch implementation are distinct versions. Preserve the task split and implementation; pretraining exposure is separate from supervised sequence-identity filtering.","source_ids":["evidence-discovery-final-tape"],"source_locator":"Section 3; Appendix A.1; Tables 1–2"}],"diagram":{"title":"Evaluation procedure","steps":["Allowed inputs: Protein sequence with a remote-homology class.","Splits: SCOP 1.75 protein domains are grouped by evolutionary hierarchy; entire superfamilies are held out for fold-level classification.","Metrics: Top-1 class accuracy."],"caption":"Conceptual procedure. Task variants and protocol versions retain their separate scoring conditions.","source_ids":["src-discovery-songlab-cal-tape","evidence-discovery-final-tape"],"source_locator":"Pinned README: compatibility notice; Evaluating a Downstream Model; Data; task leaderboards; README task-specific leaderboard and data-source references; Section 3 task definitions; Appendix A.1.1–A.1.5"},"coverage":"limited","gaps":["Uncertainty: The original paper reports point estimates in its task result tables. Methods and task appendices do not define a suite-wide repeated-seed or bootstrap interval; individual later evaluations must supply their own uncertainty."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction."}}},"description":"Remote Homology Detection","facets":{"areas":["protein-function"]},"id":"discovery-benchmark-tape-remote-homology-detection","kind":"benchmark","links":[{"relation":"parent","target_id":"discovery-benchmark-tape"},{"relation":"part_of","target_id":"discovery-benchmark-tape"}],"name":"TAPE Remote Homology Detection","source_ids":["src-discovery-songlab-cal-tape"],"status":"discovered"}
{"attributes":{"entity_level":"task","scope_note":"Specialist molecular or omics evaluation; protocol details require review before numerical comparison.","task":"Secondary Structure","version":null,"historical_missing_metadata":{"dataset_release":"unextracted","metric_implementation":"unextracted","split_manifest":"unextracted","version":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"The TAPE residue secondary-structure classification task evaluates a trained protein representation.","summary_source_ids":["src-discovery-songlab-cal-tape"],"summary_source_locator":"Pinned README: compatibility notice; Evaluating a Downstream Model; Data; task leaderboards; README task-specific leaderboard and data-source references","sections":[{"title":"Evaluation methodology","body":"Training/validation and CB513, CASP12 and TS115 test proteins are filtered at 25% sequence identity. Identity filtering excludes close train/test homologues; it is not a species holdout.","source_ids":["evidence-discovery-final-tape"],"source_locator":"Section 3 task definitions; Appendix A.1.1–A.1.5"}],"facts":[{"label":"Datasets","value":"Klausen training/validation proteins with CB513, CASP12 and TS115 structure-label test collections.","status":"source_checked","source_ids":["evidence-discovery-final-tape"],"source_locator":"Appendix A.1"},{"label":"Splits","value":"Training/validation and CB513, CASP12 and TS115 test proteins are filtered at 25% sequence identity.","status":"source_checked","source_ids":["evidence-discovery-final-tape"],"source_locator":"Section 3 task definitions; Appendix A.1.1–A.1.5"},{"label":"Metrics","value":"Three-class residue accuracy.","status":"source_checked","source_ids":["src-discovery-songlab-cal-tape"],"source_locator":"Pinned README: compatibility notice; Evaluating a Downstream Model; Data; task leaderboards; README task-specific leaderboard and data-source references"},{"label":"Baselines","value":"Task leaderboard comparisons include Transformer, LSTM, UniRep, ResNet, Bepler and one-hot baselines; alignment-augmented references appear where applicable.","status":"source_checked","source_ids":["src-discovery-songlab-cal-tape"],"source_locator":"Pinned README: compatibility notice; Evaluating a Downstream Model; Data; task leaderboards; README task-specific leaderboard and data-source references"},{"label":"Leakage controls","value":"Identity filtering excludes close train/test homologues; it is not a species holdout.","status":"source_checked","source_ids":["evidence-discovery-final-tape"],"source_locator":"Section 3 task definitions; Appendix A.1.1–A.1.5"},{"label":"Uncertainty","value":"The original paper reports point estimates in its task result tables. Methods and task appendices do not define a suite-wide repeated-seed or bootstrap interval; individual later evaluations must supply their own uncertainty.","status":"unreported","source_ids":["evidence-discovery-final-tape"],"source_locator":"Section 3; Appendix A.1; Tables 1–2"},{"label":"Entity type","value":"Constituent benchmark task: TAPE Secondary Structure","status":"source_checked","source_ids":["src-discovery-songlab-cal-tape"],"source_locator":"Pinned README: compatibility notice; Evaluating a Downstream Model; Data; task leaderboards; README task-specific leaderboard and data-source references"},{"label":"Organisms","value":"Protein structures from the mixed-organism Klausen/CB513/CASP12/TS115 collections.","status":"source_checked","source_ids":["evidence-discovery-final-tape"],"source_locator":"Section 3 task definitions; Appendix A.1.1–A.1.5"},{"label":"Assays","value":"Protein structure/homology annotations and experimental fluorescence/stability measurements.","status":"source_checked","source_ids":["src-discovery-songlab-cal-tape"],"source_locator":"Pinned README: compatibility notice; Evaluating a Downstream Model; Data; task leaderboards; README task-specific leaderboard and data-source references"},{"label":"Allowed inputs","value":"Protein amino-acid sequence with residue-level secondary-structure labels.","status":"source_checked","source_ids":["src-discovery-songlab-cal-tape"],"source_locator":"Pinned README: compatibility notice; Evaluating a Downstream Model; Data; task leaderboards; README task-specific leaderboard and data-source references"},{"label":"Adaptation","value":"Unsupervised pretraining followed by supervised downstream training; the README warns that downstream hyperparameters require task-specific tuning.","status":"source_checked","source_ids":["src-discovery-songlab-cal-tape"],"source_locator":"Pinned README: compatibility notice; Evaluating a Downstream Model; Data; task leaderboards; README task-specific leaderboard and data-source references"}],"strengths":[{"text":"Distinct sequence, residue and protein-level tasks avoid relying on language-model perplexity as a proxy for transfer.","source_ids":["src-discovery-songlab-cal-tape"],"source_locator":"Pinned README: compatibility notice; Evaluating a Downstream Model; Data; task leaderboards; README task-specific leaderboard and data-source references"}],"limitations":[{"text":"The original TAPE paper and the later PyTorch implementation are distinct versions. Preserve the task split and implementation; pretraining exposure is separate from supervised sequence-identity filtering.","source_ids":["evidence-discovery-final-tape"],"source_locator":"Section 3; Appendix A.1; Tables 1–2"}],"diagram":{"title":"Evaluation procedure","steps":["Allowed inputs: Protein amino-acid sequence with residue-level secondary-structure labels.","Splits: Training/validation and CB513, CASP12 and TS115 test proteins are filtered at 25% sequence identity.","Metrics: Three-class residue accuracy."],"caption":"Conceptual procedure. Task variants and protocol versions retain their separate scoring conditions.","source_ids":["src-discovery-songlab-cal-tape","evidence-discovery-final-tape"],"source_locator":"Pinned README: compatibility notice; Evaluating a Downstream Model; Data; task leaderboards; README task-specific leaderboard and data-source references; Section 3 task definitions; Appendix A.1.1–A.1.5"},"coverage":"limited","gaps":["Uncertainty: The original paper reports point estimates in its task result tables. Methods and task appendices do not define a suite-wide repeated-seed or bootstrap interval; individual later evaluations must supply their own uncertainty."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction."}}},"description":"Secondary Structure","facets":{"areas":["protein-function"]},"id":"discovery-benchmark-tape-secondary-structure","kind":"benchmark","links":[{"relation":"parent","target_id":"discovery-benchmark-tape"},{"relation":"part_of","target_id":"discovery-benchmark-tape"}],"name":"TAPE Secondary Structure","source_ids":["src-discovery-songlab-cal-tape"],"status":"discovered"}
{"attributes":{"entity_level":"task","scope_note":"Specialist molecular or omics evaluation; protocol details require review before numerical comparison.","task":"Stability","version":null,"historical_missing_metadata":{"dataset_release":"unextracted","metric_implementation":"unextracted","split_manifest":"unextracted","version":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"The TAPE protein stability prediction task evaluates a trained protein representation.","summary_source_ids":["src-discovery-songlab-cal-tape"],"summary_source_locator":"Pinned README: compatibility notice; Evaluating a Downstream Model; Data; task leaderboards; README task-specific leaderboard and data-source references","sections":[{"title":"Evaluation methodology","body":"Training/validation use four rounds of designed-protein stability experiments; testing uses seventeen one-mutation neighbourhoods around selected promising proteins. Designed round-to-neighbourhood generalization is intentional; test variants can be close to parent proteins observed during training.","source_ids":["evidence-discovery-final-tape"],"source_locator":"Section 3 task definitions; Appendix A.1.1–A.1.5"}],"facts":[{"label":"Datasets","value":"Rocklin designed-protein stability measurements, using the original TAPE round-to-neighbourhood partition.","status":"source_checked","source_ids":["evidence-discovery-final-tape"],"source_locator":"Appendix A.1"},{"label":"Splits","value":"Training/validation use four rounds of designed-protein stability experiments; testing uses seventeen one-mutation neighbourhoods around selected promising proteins.","status":"source_checked","source_ids":["evidence-discovery-final-tape"],"source_locator":"Section 3 task definitions; Appendix A.1.1–A.1.5"},{"label":"Metrics","value":"Spearman rank correlation.","status":"source_checked","source_ids":["src-discovery-songlab-cal-tape"],"source_locator":"Pinned README: compatibility notice; Evaluating a Downstream Model; Data; task leaderboards; README task-specific leaderboard and data-source references"},{"label":"Baselines","value":"Task leaderboard comparisons include Transformer, LSTM, UniRep, ResNet, Bepler and one-hot baselines; alignment-augmented references appear where applicable.","status":"source_checked","source_ids":["src-discovery-songlab-cal-tape"],"source_locator":"Pinned README: compatibility notice; Evaluating a Downstream Model; Data; task leaderboards; README task-specific leaderboard and data-source references"},{"label":"Leakage controls","value":"Designed round-to-neighbourhood generalization is intentional; test variants can be close to parent proteins observed during training.","status":"source_checked","source_ids":["evidence-discovery-final-tape"],"source_locator":"Section 3 task definitions; Appendix A.1.1–A.1.5"},{"label":"Uncertainty","value":"The original paper reports point estimates in its task result tables. Methods and task appendices do not define a suite-wide repeated-seed or bootstrap interval; individual later evaluations must supply their own uncertainty.","status":"unreported","source_ids":["evidence-discovery-final-tape"],"source_locator":"Section 3; Appendix A.1; Tables 1–2"},{"label":"Entity type","value":"Constituent benchmark task: TAPE Stability","status":"source_checked","source_ids":["src-discovery-songlab-cal-tape"],"source_locator":"Pinned README: compatibility notice; Evaluating a Downstream Model; Data; task leaderboards; README task-specific leaderboard and data-source references"},{"label":"Organisms","value":"Designed protein sequences measured by the Rocklin stability assay; a single natural organism label is inapplicable.","status":"source_checked","source_ids":["evidence-discovery-final-tape"],"source_locator":"Section 3 task definitions; Appendix A.1.1–A.1.5"},{"label":"Assays","value":"Protein structure/homology annotations and experimental fluorescence/stability measurements.","status":"source_checked","source_ids":["src-discovery-songlab-cal-tape"],"source_locator":"Pinned README: compatibility notice; Evaluating a Downstream Model; Data; task leaderboards; README task-specific leaderboard and data-source references"},{"label":"Allowed inputs","value":"Protein sequence with a measured stability target.","status":"source_checked","source_ids":["src-discovery-songlab-cal-tape"],"source_locator":"Pinned README: compatibility notice; Evaluating a Downstream Model; Data; task leaderboards; README task-specific leaderboard and data-source references"},{"label":"Adaptation","value":"Unsupervised pretraining followed by supervised downstream training; the README warns that downstream hyperparameters require task-specific tuning.","status":"source_checked","source_ids":["src-discovery-songlab-cal-tape"],"source_locator":"Pinned README: compatibility notice; Evaluating a Downstream Model; Data; task leaderboards; README task-specific leaderboard and data-source references"}],"strengths":[{"text":"Distinct sequence, residue and protein-level tasks avoid relying on language-model perplexity as a proxy for transfer.","source_ids":["src-discovery-songlab-cal-tape"],"source_locator":"Pinned README: compatibility notice; Evaluating a Downstream Model; Data; task leaderboards; README task-specific leaderboard and data-source references"}],"limitations":[{"text":"The original TAPE paper and the later PyTorch implementation are distinct versions. Preserve the task split and implementation; pretraining exposure is separate from supervised sequence-identity filtering.","source_ids":["evidence-discovery-final-tape"],"source_locator":"Section 3; Appendix A.1; Tables 1–2"}],"diagram":{"title":"Evaluation procedure","steps":["Allowed inputs: Protein sequence with a measured stability target.","Splits: Training/validation use four rounds of designed-protein stability experiments; testing uses seventeen one-mutation neighbourhoods around selected promising proteins.","Metrics: Spearman rank correlation."],"caption":"Conceptual procedure. Task variants and protocol versions retain their separate scoring conditions.","source_ids":["src-discovery-songlab-cal-tape","evidence-discovery-final-tape"],"source_locator":"Pinned README: compatibility notice; Evaluating a Downstream Model; Data; task leaderboards; README task-specific leaderboard and data-source references; Section 3 task definitions; Appendix A.1.1–A.1.5"},"coverage":"limited","gaps":["Uncertainty: The original paper reports point estimates in its task result tables. Methods and task appendices do not define a suite-wide repeated-seed or bootstrap interval; individual later evaluations must supply their own uncertainty."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction."}}},"description":"Stability","facets":{"areas":["protein-function"]},"id":"discovery-benchmark-tape-stability","kind":"benchmark","links":[{"relation":"parent","target_id":"discovery-benchmark-tape"},{"relation":"part_of","target_id":"discovery-benchmark-tape"}],"name":"TAPE Stability","source_ids":["src-discovery-songlab-cal-tape"],"status":"discovered"}
{"attributes":{"entity_level":"suite","scope_note":"Specialist molecular or omics evaluation; protocol details require review before numerical comparison.","task":"Molecular binding, biochemical activity and related specialist tasks","version":null,"historical_missing_metadata":{"dataset_release":"unextracted","metric_implementation":"unextracted","split_manifest":"unextracted","version":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"TDC organizes molecular prediction tasks into datasets and benchmark groups with explicit splitting and evaluation interfaces.","summary_source_ids":["src-discovery-mims-harvard-tdc"],"summary_source_locator":"Pinned README: data functions; dataset splits; evaluators; benchmark groups","sections":[{"title":"Evaluation methodology","body":"Therapeutics Data Commons supplies separate molecular tasks and curated benchmark groups. Each group specifies datasets, prediction units, partitions and metrics; the original ADMET example uses scaffold splits and simple descriptor or sequence baselines. Only the molecular and mechanistic tasks within rewire’s scope belong in this catalogue.","source_ids":["evidence-discovery-final-tdc"],"source_locator":"Section 9 and Tables 3–4: benchmark groups"}],"facts":[{"label":"Datasets","value":"Multiple task-specific molecular datasets, including grouped benchmark themes.","status":"source_checked","source_ids":["src-discovery-mims-harvard-tdc"],"source_locator":"Pinned README: data functions; dataset splits; evaluators; benchmark groups"},{"label":"Splits","value":"Random and scaffold-based splits are supported with recorded seeds/fractions; benchmark groups expose default split routines.","status":"source_checked","source_ids":["src-discovery-mims-harvard-tdc"],"source_locator":"Pinned README: data functions; dataset splits; evaluators; benchmark groups"},{"label":"Metrics","value":"A named evaluator computes task metrics; ROC-AUC is one documented example, not a universal TDC metric.","status":"source_checked","source_ids":["src-discovery-mims-harvard-tdc"],"source_locator":"Pinned README: data functions; dataset splits; evaluators; benchmark groups"},{"label":"Baselines","value":"The original molecular ADMET example compares RDKit2D-descriptor MLPs, Morgan-fingerprint MLPs and SMILES CNNs. Other in-scope molecular tasks require their own comparator set; these are not universal baselines for all of TDC.","status":"source_checked","source_ids":["evidence-discovery-final-tdc"],"source_locator":"Section 9 and Tables 3–4: benchmark groups"},{"label":"Leakage controls","value":"Scaffold holdout is available for relevant molecular tasks; it is not implied for every TDC dataset.","status":"source_checked","source_ids":["src-discovery-mims-harvard-tdc"],"source_locator":"Pinned README: data functions; dataset splits; evaluators; benchmark groups"},{"label":"Uncertainty","value":"The original ADMET table prints ± values but its caption and accompanying protocol do not define a suite-wide resampling or uncertainty rule. A result must retain the exact benchmark-group submission protocol before those values are interpreted.","status":"unreported","source_ids":["evidence-discovery-final-tdc"],"source_locator":"Section 9 and Tables 3–4: benchmark groups"},{"label":"Entity type","value":"Task and benchmark-group platform for molecular prediction.","status":"source_checked","source_ids":["src-discovery-mims-harvard-tdc"],"source_locator":"Pinned README: data functions; dataset splits; evaluators; benchmark groups"},{"label":"Organisms","value":"Organism scope is dataset-specific; the umbrella platform does not define one species.","status":"inapplicable","source_ids":["src-discovery-mims-harvard-tdc"],"source_locator":"Pinned README: data functions; dataset splits; evaluators; benchmark groups"},{"label":"Assays","value":"Task-specific molecular and therapeutic-property labels.","status":"source_checked","source_ids":["src-discovery-mims-harvard-tdc"],"source_locator":"Pinned README: data functions; dataset splits; evaluators; benchmark groups"},{"label":"Allowed inputs","value":"Named dataset inputs and labels; the benchmark group identifies the relevant molecular representation.","status":"source_checked","source_ids":["src-discovery-mims-harvard-tdc"],"source_locator":"Pinned README: data functions; dataset splits; evaluators; benchmark groups"},{"label":"Adaptation","value":"Dataset-specific supervised evaluation with default or explicitly selected split methods.","status":"source_checked","source_ids":["src-discovery-mims-harvard-tdc"],"source_locator":"Pinned README: data functions; dataset splits; evaluators; benchmark groups"}],"strengths":[{"text":"Benchmark groups expose standard split and evaluator routines while permitting explicit alternatives.","source_ids":["src-discovery-mims-harvard-tdc"],"source_locator":"Pinned README: data functions; dataset splits; evaluators; benchmark groups"}],"limitations":[{"text":"TDC also contains tasks outside this catalogue’s molecular remit. Shared software access does not make datasets, labels or evaluation protocols scientifically interchangeable.","source_ids":["evidence-discovery-final-tdc"],"source_locator":"Section 9 and Tables 3–4: benchmark groups"}],"diagram":{"title":"Evaluation procedure","steps":["Allowed inputs: Named dataset inputs and labels; the benchmark group identifies the relevant molecular representation.","Splits: Random and scaffold-based splits are supported with recorded seeds/fractions; benchmark groups expose default split routines.","Metrics: A named evaluator computes task metrics; ROC-AUC is one documented example, not a universal TDC metric."],"caption":"Conceptual procedure. Task variants and protocol versions retain their separate scoring conditions.","source_ids":["src-discovery-mims-harvard-tdc"],"source_locator":"Pinned README: data functions; dataset splits; evaluators; benchmark groups"},"coverage":"limited","gaps":["Uncertainty: The original ADMET table prints ± values but its caption and accompanying protocol do not define a suite-wide resampling or uncertainty rule. A result must retain the exact benchmark-group submission protocol before those values are interpreted."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction."}}},"description":"Molecular binding, biochemical activity and related specialist tasks","facets":{"areas":["molecular-interactions"]},"id":"discovery-benchmark-tdc-molecular-tasks","kind":"benchmark","links":[],"name":"TDC molecular tasks","source_ids":["src-discovery-mims-harvard-tdc"],"status":"discovered"}
{"attributes":{"entity_level":"challenge","scope_note":"Specialist molecular or omics evaluation; protocol details require review before numerical comparison.","task":"Zero-shot perturbation prediction in unseen cellular contexts","version":null,"historical_missing_metadata":{"dataset_release":"unextracted","metric_implementation":"unextracted","split_manifest":"unextracted","version":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"The 2026 Virtual Cell Challenge evaluates perturbation-response prediction in unseen cellular contexts.","summary_source_ids":["evidence-benchmark-vcc2026-snapshot"],"summary_source_locator":"Arc Institute 2026 Virtual Cell Challenge announcement: task; context holdouts; inputs; scoring","sections":[{"title":"Evaluation methodology","body":"The 2026 challenge predicts CRISPRi responses in cellular contexts with no challenge-specific perturbation training set. Participants receive unperturbed cells and target-gene identifiers; three contexts support validation and three different contexts support final testing. Scores are calibrated against baseline and split-half experimental references.","source_ids":["evidence-discovery-final-vcc-guide"],"source_locator":"VCC CLI guide: context labels and interpretation of the six metrics"}],"facts":[{"label":"Datasets","value":"Challenge-specific cellular-context evaluation data; the previous year’s released dataset is a separate resource.","status":"source_checked","source_ids":["evidence-benchmark-vcc2026-snapshot"],"source_locator":"Arc Institute 2026 Virtual Cell Challenge announcement: task; context holdouts; inputs; scoring"},{"label":"Splits","value":"Validation contexts support the live leaderboard and other contexts are held for final testing; there is no challenge-specific training set.","status":"source_checked","source_ids":["evidence-benchmark-vcc2026-snapshot"],"source_locator":"Arc Institute 2026 Virtual Cell Challenge announcement: task; context holdouts; inputs; scoring"},{"label":"Metrics","value":"Six calibrated components: perturbation discrimination, expression accuracy, differential-expression log-fold-change accuracy, direction fidelity, direction reach and significance overlap. Their unweighted mean is the overall score; each result also needs its partition, perturbation panel and anchor-set identity.","status":"source_checked","source_ids":["evidence-discovery-final-vcc-guide"],"source_locator":"VCC CLI guide: score interpretation and partition/panel/anchor stamp"},{"label":"Baselines","value":"Participants may choose modeling strategies and train on public or their own datasets.","status":"source_checked","source_ids":["evidence-benchmark-vcc2026-snapshot"],"source_locator":"Arc Institute 2026 Virtual Cell Challenge announcement: task; context holdouts; inputs; scoring"},{"label":"Leakage controls","value":"Unseen-context generalization defines the challenge; exact admissibility and overlap rules require the detailed competition protocol.","status":"source_checked","source_ids":["evidence-benchmark-vcc2026-snapshot"],"source_locator":"Arc Institute 2026 Virtual Cell Challenge announcement: task; context holdouts; inputs; scoring"},{"label":"Uncertainty","value":"The score scale uses a mean-effect baseline and a split-half replicate of the real experiment as reference points. Replicate noise is therefore represented in calibration, but the guide does not define a universal confidence interval on the final leaderboard score.","status":"source_checked","source_ids":["evidence-discovery-final-vcc-guide"],"source_locator":"VCC CLI guide: context labels and interpretation of the six metrics"},{"label":"Entity type","value":"Dated cellular perturbation prediction challenge.","status":"source_checked","source_ids":["evidence-benchmark-vcc2026-snapshot"],"source_locator":"Arc Institute 2026 Virtual Cell Challenge announcement: task; context holdouts; inputs; scoring"},{"label":"Organisms","value":"The released contexts are deliberately anonymized cell lines A/B/C for validation and D/E/F for final testing. The inspected public announcement and CLI guide do not provide their exact line identities or a per-context organism manifest; do not infer them from the 2025 H1 dataset.","status":"unreported","source_ids":["evidence-discovery-final-vcc-guide"],"source_locator":"VCC CLI guide: context labels and interpretation of the six metrics"},{"label":"Assays","value":"Challenge-specific perturbation response data; prior-year data form a different resource.","status":"source_checked","source_ids":["evidence-benchmark-vcc2026-snapshot"],"source_locator":"Arc Institute 2026 Virtual Cell Challenge announcement: task; context holdouts; inputs; scoring"},{"label":"Allowed inputs","value":"Released challenge inputs and the submission schema for the 2026 edition.","status":"source_checked","source_ids":["evidence-benchmark-vcc2026-snapshot"],"source_locator":"Arc Institute 2026 Virtual Cell Challenge announcement: task; context holdouts; inputs; scoring"},{"label":"Adaptation","value":"Rules and deadlines belong to this challenge edition; prior-year adaptation conditions are not automatically inherited.","status":"source_checked","source_ids":["evidence-benchmark-vcc2026-snapshot"],"source_locator":"Arc Institute 2026 Virtual Cell Challenge announcement: task; context holdouts; inputs; scoring"}],"strengths":[{"text":"A dated challenge identity prevents historical datasets being mistaken for the current evaluation.","source_ids":["evidence-benchmark-vcc2026-snapshot"],"source_locator":"Arc Institute 2026 Virtual Cell Challenge announcement: task; context holdouts; inputs; scoring"}],"limitations":[{"text":"Validation and final phases use different contexts, perturbation panels and scoring bundles. Their scores are not directly comparable, and a replicate-calibrated value is neither a percentage correct nor an absolute biological ceiling.","source_ids":["evidence-discovery-final-vcc-guide"],"source_locator":"VCC CLI guide: context labels and interpretation of the six metrics"}],"diagram":{"title":"Evaluation procedure","steps":["Allowed inputs: Released challenge inputs and the submission schema for the 2026 edition.","Splits: Validation contexts support the live leaderboard and other contexts are held for final testing; there is no challenge-specific training set.","Metrics: Six calibrated components: perturbation discrimination, expression accuracy, differential-expression log-fold-change accuracy, direction fidelity, direction reach and significance overlap. Their unweighted mean is the overall score; each result also needs its partition, perturbation panel and anchor-set identity."],"caption":"Conceptual procedure. Task variants and protocol versions retain their separate scoring conditions.","source_ids":["evidence-benchmark-vcc2026-snapshot","evidence-discovery-final-vcc-guide"],"source_locator":"Arc Institute 2026 Virtual Cell Challenge announcement: task; context holdouts; inputs; scoring; VCC CLI guide: score interpretation and partition/panel/anchor stamp"},"coverage":"limited","gaps":["Organisms: The released contexts are deliberately anonymized cell lines A/B/C for validation and D/E/F for final testing. The inspected public announcement and CLI guide do not provide their exact line identities or a per-context organism manifest; do not infer them from the 2025 H1 dataset."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction."}}},"description":"Zero-shot perturbation prediction in unseen cellular contexts","facets":{"areas":["single-cell"]},"id":"discovery-benchmark-virtual-cell-challenge-2026","kind":"benchmark","links":[],"name":"Virtual Cell Challenge 2026","source_ids":["src-discovery-vcc2026"],"status":"discovered"}
{"attributes":{"assay":"mass spectrometry","missing_metadata":{"split":"not_applicable","version":"unextracted"},"scope_note":"Reference library; a leakage-aware benchmark split and scoring protocol must be defined separately.","split":null,"version":null},"description":"Experimental lipid reference spectra for identification assessment.","facets":{"areas":["lipidomics"]},"id":"discovery-dataset-lipid-maps-standards-spectra","kind":"dataset","links":[],"name":"LIPID MAPS Standards Spectra","source_ids":["src-discovery-lipidmaps-spectra"],"status":"discovered"}
{"attributes":{"missing_metadata":{"denominator":"unextracted","split":"unextracted","version":"unreported"},"split":null,"version":null},"description":"Dataset used by the historical TAPE leaderboard.","facets":{"areas":["protein-function"]},"id":"discovery-dataset-tape-fluorescence-source-dataset","kind":"dataset","links":[],"name":"TAPE Fluorescence source dataset","source_ids":["src-discovery-songlab-cal-tape"],"status":"discovered"}
{"attributes":{"missing_metadata":{"denominator":"unextracted","split":"unextracted","version":"unreported"},"split":null,"version":null},"description":"Dataset used by the historical TAPE leaderboard.","facets":{"areas":["protein-function"]},"id":"discovery-dataset-tape-stability-source-dataset","kind":"dataset","links":[],"name":"TAPE Stability source dataset","source_ids":["src-discovery-songlab-cal-tape"],"status":"discovered"}
{"attributes":{"comparison":{"adaptation":null,"aggregation":null,"budget":null,"dataset_version":null,"inputs":null,"metric_implementation":null,"population":null,"protocol_id":"discovery-benchmark-tape-fluorescence","split":null},"missing_metadata":{"checkpoint":"unreported","denominator":"unextracted","seeds":"unreported","split_manifest":"unextracted"},"origin":"author_reported","protocol":"TAPE Fluorescence","version":"6d345c2b2bbf52cd32cf179325c222afd92aec7e"},"description":"Historical benchmark-maintainer report; experiments have not been rerun.","facets":{"areas":["protein-function"]},"id":"discovery-evaluation-tape-fluorescence-bepler-leaderboard-evaluation","kind":"evaluation","links":[{"relation":"model","target_id":"discovery-model-tape-bepler"},{"relation":"benchmark","target_id":"discovery-benchmark-tape-fluorescence"},{"relation":"dataset","target_id":"discovery-dataset-tape-fluorescence-source-dataset"}],"name":"TAPE Fluorescence Bepler leaderboard evaluation","source_ids":["src-discovery-songlab-cal-tape"],"status":"discovered"}
{"attributes":{"comparison":{"adaptation":null,"aggregation":null,"budget":null,"dataset_version":null,"inputs":null,"metric_implementation":null,"population":null,"protocol_id":"discovery-benchmark-tape-fluorescence","split":null},"missing_metadata":{"checkpoint":"unreported","denominator":"unextracted","seeds":"unreported","split_manifest":"unextracted"},"origin":"author_reported","protocol":"TAPE Fluorescence","version":"6d345c2b2bbf52cd32cf179325c222afd92aec7e"},"description":"Historical benchmark-maintainer report; experiments have not been rerun.","facets":{"areas":["protein-function"]},"id":"discovery-evaluation-tape-fluorescence-lstm-leaderboard-evaluation","kind":"evaluation","links":[{"relation":"model","target_id":"discovery-model-tape-lstm"},{"relation":"benchmark","target_id":"discovery-benchmark-tape-fluorescence"},{"relation":"dataset","target_id":"discovery-dataset-tape-fluorescence-source-dataset"}],"name":"TAPE Fluorescence LSTM leaderboard evaluation","source_ids":["src-discovery-songlab-cal-tape"],"status":"discovered"}
{"attributes":{"comparison":{"adaptation":null,"aggregation":null,"budget":null,"dataset_version":null,"inputs":null,"metric_implementation":null,"population":null,"protocol_id":"discovery-benchmark-tape-fluorescence","split":null},"missing_metadata":{"checkpoint":"unreported","denominator":"unextracted","seeds":"unreported","split_manifest":"unextracted"},"origin":"author_reported","protocol":"TAPE Fluorescence","version":"6d345c2b2bbf52cd32cf179325c222afd92aec7e"},"description":"Historical benchmark-maintainer report; experiments have not been rerun.","facets":{"areas":["protein-function"]},"id":"discovery-evaluation-tape-fluorescence-one-hot-leaderboard-evaluation","kind":"evaluation","links":[{"relation":"model","target_id":"discovery-model-tape-one-hot"},{"relation":"benchmark","target_id":"discovery-benchmark-tape-fluorescence"},{"relation":"dataset","target_id":"discovery-dataset-tape-fluorescence-source-dataset"}],"name":"TAPE Fluorescence One Hot leaderboard evaluation","source_ids":["src-discovery-songlab-cal-tape"],"status":"discovered"}
{"attributes":{"comparison":{"adaptation":null,"aggregation":null,"budget":null,"dataset_version":null,"inputs":null,"metric_implementation":null,"population":null,"protocol_id":"discovery-benchmark-tape-fluorescence","split":null},"missing_metadata":{"checkpoint":"unreported","denominator":"unextracted","seeds":"unreported","split_manifest":"unextracted"},"origin":"author_reported","protocol":"TAPE Fluorescence","version":"6d345c2b2bbf52cd32cf179325c222afd92aec7e"},"description":"Historical benchmark-maintainer report; experiments have not been rerun.","facets":{"areas":["protein-function"]},"id":"discovery-evaluation-tape-fluorescence-resnet-leaderboard-evaluation","kind":"evaluation","links":[{"relation":"model","target_id":"discovery-model-tape-resnet"},{"relation":"benchmark","target_id":"discovery-benchmark-tape-fluorescence"},{"relation":"dataset","target_id":"discovery-dataset-tape-fluorescence-source-dataset"}],"name":"TAPE Fluorescence ResNet leaderboard evaluation","source_ids":["src-discovery-songlab-cal-tape"],"status":"discovered"}
{"attributes":{"comparison":{"adaptation":null,"aggregation":null,"budget":null,"dataset_version":null,"inputs":null,"metric_implementation":null,"population":null,"protocol_id":"discovery-benchmark-tape-fluorescence","split":null},"missing_metadata":{"checkpoint":"unreported","denominator":"unextracted","seeds":"unreported","split_manifest":"unextracted"},"origin":"author_reported","protocol":"TAPE Fluorescence","version":"6d345c2b2bbf52cd32cf179325c222afd92aec7e"},"description":"Historical benchmark-maintainer report; experiments have not been rerun.","facets":{"areas":["protein-function"]},"id":"discovery-evaluation-tape-fluorescence-transformer-leaderboard-evaluation","kind":"evaluation","links":[{"relation":"model","target_id":"discovery-model-tape-transformer"},{"relation":"benchmark","target_id":"discovery-benchmark-tape-fluorescence"},{"relation":"dataset","target_id":"discovery-dataset-tape-fluorescence-source-dataset"}],"name":"TAPE Fluorescence Transformer leaderboard evaluation","source_ids":["src-discovery-songlab-cal-tape"],"status":"discovered"}
{"attributes":{"comparison":{"adaptation":null,"aggregation":null,"budget":null,"dataset_version":null,"inputs":null,"metric_implementation":null,"population":null,"protocol_id":"discovery-benchmark-tape-fluorescence","split":null},"missing_metadata":{"checkpoint":"unreported","denominator":"unextracted","seeds":"unreported","split_manifest":"unextracted"},"origin":"author_reported","protocol":"TAPE Fluorescence","version":"6d345c2b2bbf52cd32cf179325c222afd92aec7e"},"description":"Historical benchmark-maintainer report; experiments have not been rerun.","facets":{"areas":["protein-function"]},"id":"discovery-evaluation-tape-fluorescence-unirep-leaderboard-evaluation","kind":"evaluation","links":[{"relation":"model","target_id":"discovery-model-tape-unirep"},{"relation":"benchmark","target_id":"discovery-benchmark-tape-fluorescence"},{"relation":"dataset","target_id":"discovery-dataset-tape-fluorescence-source-dataset"}],"name":"TAPE Fluorescence Unirep leaderboard evaluation","source_ids":["src-discovery-songlab-cal-tape"],"status":"discovered"}
{"attributes":{"comparison":{"adaptation":null,"aggregation":null,"budget":null,"dataset_version":null,"inputs":null,"metric_implementation":null,"population":null,"protocol_id":"discovery-benchmark-tape-stability","split":null},"missing_metadata":{"checkpoint":"unreported","denominator":"unextracted","seeds":"unreported","split_manifest":"unextracted"},"origin":"author_reported","protocol":"TAPE Stability","version":"6d345c2b2bbf52cd32cf179325c222afd92aec7e"},"description":"Historical benchmark-maintainer report; experiments have not been rerun.","facets":{"areas":["protein-function"]},"id":"discovery-evaluation-tape-stability-bepler-leaderboard-evaluation","kind":"evaluation","links":[{"relation":"model","target_id":"discovery-model-tape-bepler"},{"relation":"benchmark","target_id":"discovery-benchmark-tape-stability"},{"relation":"dataset","target_id":"discovery-dataset-tape-stability-source-dataset"}],"name":"TAPE Stability Bepler leaderboard evaluation","source_ids":["src-discovery-songlab-cal-tape"],"status":"discovered"}
{"attributes":{"comparison":{"adaptation":null,"aggregation":null,"budget":null,"dataset_version":null,"inputs":null,"metric_implementation":null,"population":null,"protocol_id":"discovery-benchmark-tape-stability","split":null},"missing_metadata":{"checkpoint":"unreported","denominator":"unextracted","seeds":"unreported","split_manifest":"unextracted"},"origin":"author_reported","protocol":"TAPE Stability","version":"6d345c2b2bbf52cd32cf179325c222afd92aec7e"},"description":"Historical benchmark-maintainer report; experiments have not been rerun.","facets":{"areas":["protein-function"]},"id":"discovery-evaluation-tape-stability-lstm-leaderboard-evaluation","kind":"evaluation","links":[{"relation":"model","target_id":"discovery-model-tape-lstm"},{"relation":"benchmark","target_id":"discovery-benchmark-tape-stability"},{"relation":"dataset","target_id":"discovery-dataset-tape-stability-source-dataset"}],"name":"TAPE Stability LSTM leaderboard evaluation","source_ids":["src-discovery-songlab-cal-tape"],"status":"discovered"}
{"attributes":{"comparison":{"adaptation":null,"aggregation":null,"budget":null,"dataset_version":null,"inputs":null,"metric_implementation":null,"population":null,"protocol_id":"discovery-benchmark-tape-stability","split":null},"missing_metadata":{"checkpoint":"unreported","denominator":"unextracted","seeds":"unreported","split_manifest":"unextracted"},"origin":"author_reported","protocol":"TAPE Stability","version":"6d345c2b2bbf52cd32cf179325c222afd92aec7e"},"description":"Historical benchmark-maintainer report; experiments have not been rerun.","facets":{"areas":["protein-function"]},"id":"discovery-evaluation-tape-stability-one-hot-leaderboard-evaluation","kind":"evaluation","links":[{"relation":"model","target_id":"discovery-model-tape-one-hot"},{"relation":"benchmark","target_id":"discovery-benchmark-tape-stability"},{"relation":"dataset","target_id":"discovery-dataset-tape-stability-source-dataset"}],"name":"TAPE Stability One Hot leaderboard evaluation","source_ids":["src-discovery-songlab-cal-tape"],"status":"discovered"}
{"attributes":{"comparison":{"adaptation":null,"aggregation":null,"budget":null,"dataset_version":null,"inputs":null,"metric_implementation":null,"population":null,"protocol_id":"discovery-benchmark-tape-stability","split":null},"missing_metadata":{"checkpoint":"unreported","denominator":"unextracted","seeds":"unreported","split_manifest":"unextracted"},"origin":"author_reported","protocol":"TAPE Stability","version":"6d345c2b2bbf52cd32cf179325c222afd92aec7e"},"description":"Historical benchmark-maintainer report; experiments have not been rerun.","facets":{"areas":["protein-function"]},"id":"discovery-evaluation-tape-stability-resnet-leaderboard-evaluation","kind":"evaluation","links":[{"relation":"model","target_id":"discovery-model-tape-resnet"},{"relation":"benchmark","target_id":"discovery-benchmark-tape-stability"},{"relation":"dataset","target_id":"discovery-dataset-tape-stability-source-dataset"}],"name":"TAPE Stability ResNet leaderboard evaluation","source_ids":["src-discovery-songlab-cal-tape"],"status":"discovered"}
{"attributes":{"comparison":{"adaptation":null,"aggregation":null,"budget":null,"dataset_version":null,"inputs":null,"metric_implementation":null,"population":null,"protocol_id":"discovery-benchmark-tape-stability","split":null},"missing_metadata":{"checkpoint":"unreported","denominator":"unextracted","seeds":"unreported","split_manifest":"unextracted"},"origin":"author_reported","protocol":"TAPE Stability","version":"6d345c2b2bbf52cd32cf179325c222afd92aec7e"},"description":"Historical benchmark-maintainer report; experiments have not been rerun.","facets":{"areas":["protein-function"]},"id":"discovery-evaluation-tape-stability-transformer-leaderboard-evaluation","kind":"evaluation","links":[{"relation":"model","target_id":"discovery-model-tape-transformer"},{"relation":"benchmark","target_id":"discovery-benchmark-tape-stability"},{"relation":"dataset","target_id":"discovery-dataset-tape-stability-source-dataset"}],"name":"TAPE Stability Transformer leaderboard evaluation","source_ids":["src-discovery-songlab-cal-tape"],"status":"discovered"}
{"attributes":{"comparison":{"adaptation":null,"aggregation":null,"budget":null,"dataset_version":null,"inputs":null,"metric_implementation":null,"population":null,"protocol_id":"discovery-benchmark-tape-stability","split":null},"missing_metadata":{"checkpoint":"unreported","denominator":"unextracted","seeds":"unreported","split_manifest":"unextracted"},"origin":"author_reported","protocol":"TAPE Stability","version":"6d345c2b2bbf52cd32cf179325c222afd92aec7e"},"description":"Historical benchmark-maintainer report; experiments have not been rerun.","facets":{"areas":["protein-function"]},"id":"discovery-evaluation-tape-stability-unirep-leaderboard-evaluation","kind":"evaluation","links":[{"relation":"model","target_id":"discovery-model-tape-unirep"},{"relation":"benchmark","target_id":"discovery-benchmark-tape-stability"},{"relation":"dataset","target_id":"discovery-dataset-tape-stability-source-dataset"}],"name":"TAPE Stability Unirep leaderboard evaluation","source_ids":["src-discovery-songlab-cal-tape"],"status":"discovered"}
{"attributes":{"access":"official_source_linked","benchmark_applicability":"candidate; not evidence of a reported evaluation","candidate_benchmark_ids":["discovery-benchmark-geneb"],"entity_level":"family","reported_name":"Agro Nucleotide Transformer","version":null,"historical_missing_metadata":{"checkpoint":"unextracted","code_licence":"unextracted","parameters":"unextracted","training_cutoff":"unextracted","training_data":"unextracted","version":"unextracted","weights_licence":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"AgroNT learns DNA representations from plant reference genomes for plant molecular prediction tasks.","summary_source_ids":["evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-4603e2d255507e025c80","evidence-official-04ac37076f906b8470cb","evidence-official-576c2ecba240ddfda8f2"],"summary_source_locator":"AgroNT paper Methods: Architecture, Pre-training dataset and Pre-training strategy; official model-card licence metadata","sections":[{"title":"How it works","body":"AgroNT learns DNA representations from plant reference genomes for plant molecular prediction tasks. One-billion-parameter encoder-only transformer with 40 attention blocks, hidden width 1,500, learned positional embeddings and a six-mer masked-language-model head. The documented inputs are plant DNA sequence, with standalone tokens for ambiguous or remainder bases. The output consists of DNA embeddings for downstream regulatory, RNA-processing or expression tasks.","source_ids":["evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-4603e2d255507e025c80","evidence-official-04ac37076f906b8470cb","evidence-official-576c2ecba240ddfda8f2"],"source_locator":"AgroNT paper Methods: Architecture, Pre-training dataset and Pre-training strategy; official model-card licence metadata"},{"title":"Versions and reproducibility","body":"1B_agro_nt pretrained model. 1,024 tokens, approximately 6kb of unambiguous sequence rather than an unconditional 6,144-base guarantee.","source_ids":["evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-4603e2d255507e025c80","evidence-official-04ac37076f906b8470cb","evidence-official-576c2ecba240ddfda8f2"],"source_locator":"AgroNT paper Methods: Architecture, Pre-training dataset and Pre-training strategy; official model-card licence metadata"}],"facts":[{"label":"Model type","value":"Plant DNA transformer encoder","status":"source_checked","source_ids":["evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-4603e2d255507e025c80","evidence-official-04ac37076f906b8470cb","evidence-official-576c2ecba240ddfda8f2"],"source_locator":"AgroNT paper Methods: Architecture, Pre-training dataset and Pre-training strategy; official model-card licence metadata"},{"label":"Architecture","value":"One-billion-parameter encoder-only transformer with 40 attention blocks, hidden width 1,500, learned positional embeddings and a six-mer masked-language-model head.","status":"source_checked","source_ids":["evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-4603e2d255507e025c80","evidence-official-04ac37076f906b8470cb","evidence-official-576c2ecba240ddfda8f2"],"source_locator":"AgroNT paper Methods: Architecture, Pre-training dataset and Pre-training strategy; official model-card licence metadata"},{"label":"Inputs","value":"Plant DNA sequence, with standalone tokens for ambiguous or remainder bases.","status":"source_checked","source_ids":["evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-4603e2d255507e025c80","evidence-official-04ac37076f906b8470cb","evidence-official-576c2ecba240ddfda8f2"],"source_locator":"AgroNT paper Methods: Architecture, Pre-training dataset and Pre-training strategy; official model-card licence metadata"},{"label":"Outputs","value":"DNA embeddings for downstream regulatory, RNA-processing or expression tasks.","status":"source_checked","source_ids":["evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-4603e2d255507e025c80","evidence-official-04ac37076f906b8470cb","evidence-official-576c2ecba240ddfda8f2"],"source_locator":"AgroNT paper Methods: Architecture, Pre-training dataset and Pre-training strategy; official model-card licence metadata"},{"label":"Parameters","value":"1 billion.","status":"source_checked","source_ids":["evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-4603e2d255507e025c80","evidence-official-04ac37076f906b8470cb","evidence-official-576c2ecba240ddfda8f2"],"source_locator":"AgroNT paper Methods: Architecture, Pre-training dataset and Pre-training strategy; official model-card licence metadata"},{"label":"Known versions","value":"1B_agro_nt pretrained model.","status":"source_checked","source_ids":["evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-4603e2d255507e025c80","evidence-official-04ac37076f906b8470cb","evidence-official-576c2ecba240ddfda8f2"],"source_locator":"AgroNT paper Methods: Architecture, Pre-training dataset and Pre-training strategy; official model-card licence metadata"},{"label":"Training data","value":"Approximately 10.5M sequences from reference genomes of 48 plant species in Ensembl Plants.","status":"source_checked","source_ids":["evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-4603e2d255507e025c80","evidence-official-04ac37076f906b8470cb","evidence-official-576c2ecba240ddfda8f2"],"source_locator":"AgroNT paper Methods: Architecture, Pre-training dataset and Pre-training strategy; official model-card licence metadata"},{"label":"Training cutoff","value":"The inspected Methods identifies 48 Ensembl Plants reference species. It does not state one latest-deposition date for their combined genomic sequences.","status":"unreported","source_ids":["evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-4603e2d255507e025c80","evidence-official-04ac37076f906b8470cb","evidence-official-576c2ecba240ddfda8f2"],"source_locator":"AgroNT paper Methods: Architecture, Pre-training dataset and Pre-training strategy; official model-card licence metadata"},{"label":"Context limits","value":"1,024 tokens, approximately 6kb of unambiguous sequence rather than an unconditional 6,144-base guarantee.","status":"source_checked","source_ids":["evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-4603e2d255507e025c80","evidence-official-04ac37076f906b8470cb","evidence-official-576c2ecba240ddfda8f2"],"source_locator":"AgroNT paper Methods: Architecture, Pre-training dataset and Pre-training strategy; official model-card licence metadata"},{"label":"Weights licence","value":"CC-BY-NC-SA-4.0 declared by the official agro-nucleotide-transformer-1b model card.","status":"source_checked","source_ids":["evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-4603e2d255507e025c80","evidence-official-04ac37076f906b8470cb","evidence-official-576c2ecba240ddfda8f2"],"source_locator":"AgroNT paper Methods: Architecture, Pre-training dataset and Pre-training strategy; official model-card licence metadata"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/instadeepai/nucleotide-transformer","status":"source_checked","source_ids":["evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-4603e2d255507e025c80","evidence-official-04ac37076f906b8470cb","evidence-official-576c2ecba240ddfda8f2"],"source_locator":"AgroNT paper Methods: Architecture, Pre-training dataset and Pre-training strategy; official model-card licence metadata"},{"label":"Code licence","value":"CC-BY-NC-SA-4.0","status":"source_checked","source_ids":["evidence-official-7e4b193e47ba209860a1"],"source_locator":"LICENSE.md: licence text"}],"strengths":[{"text":"Pretraining explicitly targets plant reference genomes, predominantly crop species.","source_ids":["evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-4603e2d255507e025c80","evidence-official-04ac37076f906b8470cb","evidence-official-576c2ecba240ddfda8f2"],"source_locator":"AgroNT paper Methods: Architecture, Pre-training dataset and Pre-training strategy; official model-card licence metadata"}],"limitations":[{"text":"The context is counted in tokens: the class token and ambiguous bases reduce the maximum number of ordinary bases represented. Predictions outside the studied species/tasks need separate evaluation.","source_ids":["evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-4603e2d255507e025c80","evidence-official-04ac37076f906b8470cb","evidence-official-576c2ecba240ddfda8f2"],"source_locator":"AgroNT paper Methods: Architecture, Pre-training dataset and Pre-training strategy; official model-card licence metadata"}],"diagram":{"title":"Agro Nucleotide Transformer workflow","steps":["Plant DNA","6-mer tokenizer","Masked-language transformer","Plant DNA embeddings"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-4603e2d255507e025c80","evidence-official-04ac37076f906b8470cb","evidence-official-576c2ecba240ddfda8f2"],"source_locator":"AgroNT paper Methods: Architecture, Pre-training dataset and Pre-training strategy; official model-card licence metadata"},"coverage":"limited","gaps":["Training cutoff: The inspected Methods identifies 48 Ensembl Plants reference species. It does not state one latest-deposition date for their combined genomic sequences."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}},"description":"Plant genomic representation family","facets":{"areas":["genomics"]},"id":"discovery-model-agro-nucleotide-transformer","kind":"model","links":[{"relation":"applicable_to","target_id":"discovery-benchmark-geneb"}],"name":"Agro Nucleotide Transformer","source_ids":["src-discovery-instadeepai-nucleotide-transformer"],"status":"discovered"}
{"attributes":{"access":"official_source_linked","benchmark_applicability":"candidate; not evidence of a reported evaluation","candidate_benchmark_ids":["discovery-benchmark-plinder"],"entity_level":"family","reported_name":"AlphaFold 3","version":null,"historical_missing_metadata":{"checkpoint":"unextracted","code_licence":"unextracted","parameters":"unextracted","training_cutoff":"unextracted","training_data":"unextracted","version":"unextracted","weights_licence":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"AlphaFold 3 predicts three-dimensional structures of complexes containing proteins, nucleic acids and other molecular components. It combines a Pairformer representation network with an atomic-coordinate diffusion model. This entry describes the model and local implementation; the hosted AlphaFold Server has a separate profile.","summary_source_ids":["evidence-alphafold-paper","evidence-alphafold-readme"],"summary_source_locator":"Abstract; Model architecture; README: Installation and Usage","sections":[{"title":"Joint structure prediction","body":"Sequence, chemical and evolutionary features feed a Pairformer, which builds representations of individual tokens and their relationships. A diffusion module then predicts atomic coordinates. Separate heads estimate confidence. The paper describes 48 Pairformer blocks; the architecture models complexes jointly rather than treating every partner as a separately folded structure.","source_ids":["evidence-alphafold-paper"],"source_locator":"Main text: Model architecture; Fig. 1d and Fig. 2"},{"title":"Training and evaluation context","body":"The standard model uses a structural training cutoff of 30 September 2021. The dedicated PoseBusters Methods section reports a separate model with a 30 September 2019 cutoff, although other training passages disagree (see limitations). Model seeds, templates, input information and ranking also affect the reported comparison; paper evaluation variants are not automatically identical to current downloadable weights.","source_ids":["evidence-alphafold-paper","evidence-alphafold-supplement"],"source_locator":"Main paper Methods: Training regime, Inference regime and PoseBusters; supplement Section 5.2"},{"title":"Local access","body":"The pinned repository provides inference code and a direct Google-hosted weights download. Its README is more current on access than the server FAQ, which still describes an application form. Code and model parameters have different licences; the server output terms should not be substituted for the local weights terms.","source_ids":["evidence-alphafold-readme","evidence-alphafold-license","evidence-alphafold-weights-terms-of-use","evidence-alphafold-server-faq"],"source_locator":"README: Obtaining Model Parameters and Licences; LICENSE; weights terms: Key things to know; FAQ: model access"},{"title":"Training data and provenance","body":"Training combines experimental PDB structures with approximately 41 million predicted protein monomers, about 25,000 disorder-focused protein complexes and about 65,000 predicted RNA structures. The supplement also lists transcription-factor examples used during fine-tuning. Its sequence-search resources include UniRef90, UniProt, BFD/Uniclust30, MGnify, Rfam and RNAcentral. These resources have different versions and dates: the structural training cutoff is not a cutoff for every sequence database.","source_ids":["evidence-alphafold-supplement"],"source_locator":"Sections 2.2 and 2.5; Table 3; Section 2.5.2 distillation datasets (PDF pages 8–9, printed pages 3–4)"}],"facts":[{"label":"Model type","value":"Pairformer plus diffusion model for joint biomolecular structure prediction.","status":"source_checked","source_ids":["evidence-alphafold-paper"],"source_locator":"Model architecture; Fig. 1"},{"label":"Architecture","value":"A 48-block Pairformer builds token and pair representations; a diffusion module predicts atomic coordinates and separate heads estimate confidence.","status":"source_checked","source_ids":["evidence-alphafold-paper"],"source_locator":"Model architecture; Fig. 1d and Fig. 2"},{"label":"Known versions","value":"AlphaFold 3; inference code/documentation reviewed at commit c0f97eda2f1f482fd94d3a38bece18c7069b4a5c. This is a software revision, not a weight-file checksum. Paper evaluation variants are distinct configurations.","status":"source_checked","source_ids":["evidence-alphafold-readme","evidence-alphafold-paper"],"source_locator":"Pinned repository revision; paper evaluation distinctions described in Methods"},{"label":"Inputs","value":"Protein, DNA and RNA sequences; chemical components; optional MSAs and structural templates.","status":"source_checked","source_ids":["evidence-alphafold-docs-input"],"source_locator":"Top-level structure; protein, RNA, DNA and ligand inputs"},{"label":"Outputs","value":"Predicted structures in mmCIF plus confidence outputs, including pLDDT, PAE, pTM and ipTM.","status":"source_checked","source_ids":["evidence-alphafold-docs-output"],"source_locator":"Output directory structure; confidence outputs"},{"label":"Training cutoff","value":"Experimental PDB structures plus protein and RNA distillation sets. The standard structural cutoff is 2021-09-30. The dedicated PoseBusters Methods section specifies a separate 2019-09-30 model; other training passages conflict with that date (see below).","status":"source_checked","source_ids":["evidence-alphafold-paper","evidence-alphafold-supplement"],"source_locator":"Main paper Methods: Training regime and PoseBusters; Supplement Sections 2.5 and 5.2"},{"label":"Training data","value":"Experimental PDB structures, protein and RNA distillation sets, and transcription-factor examples used during fine-tuning. Sequence-search databases are separately versioned input resources.","status":"source_checked","source_ids":["evidence-alphafold-supplement"],"source_locator":"Sections 2.2 and 2.5; Table 3; training-data discussion below"},{"label":"Context limits","value":"The default largest compilation bucket is 5,120 tokens. The documentation supports larger inputs by configuration, subject to memory; this is not a universal architectural context limit.","status":"source_checked","source_ids":["evidence-alphafold-docs-performance"],"source_locator":"Compilation buckets; predicting structures with more than 5,120 tokens"},{"label":"Access","value":"Public inference implementation; weights downloaded directly from Google under separate non-commercial terms.","status":"source_checked","source_ids":["evidence-alphafold-readme"],"source_locator":"Obtaining Model Parameters; Installation and Usage"},{"label":"Code licence","value":"Apache License 2.0.","status":"source_checked","source_ids":["evidence-alphafold-license"],"source_locator":"LICENSE"},{"label":"Weights licence","value":"Custom AlphaFold 3 Model Parameters Terms of Use, last modified 2024-11-09. Non-commercial use by or for non-commercial organisations; additional output and redistribution restrictions apply.","status":"source_checked","source_ids":["evidence-alphafold-weights-terms-of-use"],"source_locator":"Key things to know; Use restrictions"},{"label":"Parameters","value":"No total trainable-parameter count is reported in the inspected main paper, supplementary architecture/training sections or implementation documentation. Layer dimensions do not establish a complete checkpoint total.","status":"unreported","source_ids":["evidence-alphafold-paper","evidence-alphafold-supplement"],"source_locator":"Main paper Model architecture; supplement Sections 3–5 and full-text parameter search; implementation documentation"}],"strengths":[{"text":"One architecture handles several molecular component types and their joint structures. This is a capability description, not evidence that every complex will be accurate.","source_ids":["evidence-alphafold-paper"],"source_locator":"Abstract; Model architecture"},{"text":"The local implementation exposes input, template and output specifications, allowing an evaluation configuration to be documented.","source_ids":["evidence-alphafold-docs-input"],"source_locator":"Input format and optional input fields"}],"limitations":[{"text":"Predictions can contain incorrect chirality, atomic clashes or spurious structure in disordered regions. Confidence and structural plausibility need separate inspection.","source_ids":["evidence-alphafold-paper"],"source_locator":"Model limitations; Fig. 5"},{"text":"Sampled structures are not a calibrated solution-state ensemble. Prediction confidence does not establish binding affinity or experimental function.","source_ids":["evidence-alphafold-paper"],"source_locator":"Model limitations: dynamics and conformational states; confidence outputs are structure-quality estimates"},{"text":"The public code licence does not remove the separate non-commercial restrictions on weights and outputs.","source_ids":["evidence-alphafold-weights-terms-of-use"],"source_locator":"Key things to know"},{"text":"The source is internally inconsistent about the PoseBusters training cutoff. Its dedicated PoseBusters Methods section and Results specify 2019-09-30, while the general Training regime and supplement Section 5.2 say 2021-09-30. The separate evaluation variant is retained; this profile does not resolve the discrepancy or assign its result to a current downloadable checkpoint.","source_ids":["evidence-alphafold-paper","evidence-alphafold-supplement"],"source_locator":"Main paper Results and Methods: PoseBusters versus Training regime; supplement Section 5.2, printed page 29"}],"diagram":{"title":"AlphaFold 3 architecture","steps":["Molecular sequences and chemical features","MSA and template features","Pairformer token and pair representations","Diffusion predicts atomic coordinates","Confidence heads and ranked structures"],"caption":"Conceptual architecture based on the paper. Exact preprocessing and sampling settings belong to each evaluation.","source_ids":["evidence-alphafold-paper"],"source_locator":"Fig. 1d; Fig. 2; Model architecture"},"coverage":"reviewed","gaps":["The inspected sources do not report a complete checkpoint parameter total; no weights were downloaded or counted.","The PoseBusters cutoff disagreement is preserved explicitly. The currently downloaded weights have not been mapped to the paper’s separate evaluation variants."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Read primary paper XML, pinned official repository documentation and licences. Reviewed public server FAQ separately. No model run, independent performance replication or human review. Supplementary PDF reviewed, including visual checks of Tables 3 and 6. Conflicting cutoff statements remain explicit. A second automated reviewer checked the AlphaFold source claims and service/model distinction; this is not human review or experimental reproduction."}}},"description":"Biomolecular complex structure prediction","facets":{"areas":["protein-structure"]},"id":"discovery-model-alphafold-3","kind":"model","links":[{"relation":"applicable_to","target_id":"discovery-benchmark-plinder"}],"name":"AlphaFold 3","source_ids":["src-discovery-google-deepmind-alphafold3"],"status":"discovered"}
{"attributes":{"access":"official_source_linked","benchmark_applicability":"candidate; not evidence of a reported evaluation","candidate_benchmark_ids":["discovery-benchmark-posebusters"],"entity_level":"method","reported_name":"AutoDock Vina","version":null,"historical_missing_metadata":{"checkpoint":"unextracted","code_licence":"unextracted","parameters":"unextracted","training_cutoff":"unextracted","training_data":"unextracted","version":"unextracted","weights_licence":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"AutoDock Vina searches for ligand conformations and poses in a molecular docking problem.","summary_source_ids":["evidence-official-23e72f7e93a0a633ab2b"],"summary_source_locator":"README.md: introduction, feature list, license and Citations","sections":[{"title":"How it works","body":"AutoDock Vina searches for ligand conformations and poses in a molecular docking problem. Scoring functions coupled to gradient-based conformational optimization and search. The documented inputs are prepared receptor and ligand structures with the configured search space and scoring function. The output consists of candidate docking poses and corresponding docking scores.","source_ids":["evidence-official-23e72f7e93a0a633ab2b"],"source_locator":"README.md: introduction, feature list, license and Citations"},{"title":"Versions and reproducibility","body":"AutoDock Vina; README cites the 1.2.0 feature expansion separately from the original 2010 method. Molecular geometry and search-box constraints rather than a sequence context window.","source_ids":["evidence-official-23e72f7e93a0a633ab2b"],"source_locator":"README.md: introduction, feature list, license and Citations"}],"facts":[{"label":"Model type","value":"Classical molecular docking software","status":"source_checked","source_ids":["evidence-official-23e72f7e93a0a633ab2b"],"source_locator":"README.md: introduction, feature list, license and Citations"},{"label":"Architecture","value":"Scoring functions coupled to gradient-based conformational optimization and search.","status":"source_checked","source_ids":["evidence-official-23e72f7e93a0a633ab2b"],"source_locator":"README.md: introduction, feature list, license and Citations"},{"label":"Inputs","value":"Prepared receptor and ligand structures with the configured search space and scoring function.","status":"source_checked","source_ids":["evidence-official-23e72f7e93a0a633ab2b"],"source_locator":"README.md: introduction, feature list, license and Citations"},{"label":"Outputs","value":"Candidate docking poses and corresponding docking scores.","status":"source_checked","source_ids":["evidence-official-23e72f7e93a0a633ab2b"],"source_locator":"README.md: introduction, feature list, license and Citations"},{"label":"Parameters","value":"Inapplicable as a neural parameter count; scoring/search parameters are separately configured.","status":"inapplicable","source_ids":["evidence-official-23e72f7e93a0a633ab2b"],"source_locator":"README.md: introduction, feature list, license and Citations"},{"label":"Known versions","value":"AutoDock Vina; README cites the 1.2.0 feature expansion separately from the original 2010 method.","status":"source_checked","source_ids":["evidence-official-23e72f7e93a0a633ab2b"],"source_locator":"README.md: introduction, feature list, license and Citations"},{"label":"Training data","value":"Not a pretrained neural model. The selected scoring function and parametrization define the procedural baseline.","status":"source_checked","source_ids":["evidence-official-23e72f7e93a0a633ab2b"],"source_locator":"README.md: introduction, feature list, license and Citations"},{"label":"Training cutoff","value":"Inapplicable to neural pretraining; reference-database and input-data dates must be recorded for each run.","status":"inapplicable","source_ids":["evidence-official-23e72f7e93a0a633ab2b"],"source_locator":"README.md: introduction, feature list, license and Citations"},{"label":"Context limits","value":"Molecular geometry and search-box constraints rather than a sequence context window.","status":"source_checked","source_ids":["evidence-official-23e72f7e93a0a633ab2b"],"source_locator":"README.md: introduction, feature list, license and Citations"},{"label":"Weights licence","value":"Inapplicable: no neural model-weight checkpoint.","status":"inapplicable","source_ids":["evidence-official-23e72f7e93a0a633ab2b"],"source_locator":"README.md: introduction, feature list, license and Citations"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/ccsb-scripps/AutoDock-Vina","status":"source_checked","source_ids":["evidence-official-23e72f7e93a0a633ab2b"],"source_locator":"README.md: introduction, feature list, license and Citations"},{"label":"Code licence","value":"Apache-2.0","status":"source_checked","source_ids":["evidence-official-640665f30e62eed9319b"],"source_locator":"LICENSE: licence text"}],"strengths":[{"text":"Supports Vina and AutoDock4 scoring, multiple-ligand/batch workflows, macrocycles and Python bindings.","source_ids":["evidence-official-23e72f7e93a0a633ab2b"],"source_locator":"README.md: introduction, feature list, license and Citations"}],"limitations":[{"text":"The chosen scoring function, molecular preparation and search configuration form part of the method. A docking score is a computational quantity rather than an experimental affinity measurement.","source_ids":["evidence-official-23e72f7e93a0a633ab2b"],"source_locator":"README.md: introduction, feature list, license and Citations"}],"diagram":{"title":"AutoDock Vina workflow","steps":["Prepared receptor and ligand","Conformational search","Docking score","Ranked candidate poses"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-23e72f7e93a0a633ab2b"],"source_locator":"README.md: introduction, feature list, license and Citations"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}},"description":"Procedural docking and virtual screening","facets":{"areas":["molecular-interactions"]},"id":"discovery-model-autodock-vina","kind":"model","links":[{"relation":"applicable_to","target_id":"discovery-benchmark-posebusters"}],"name":"AutoDock Vina","source_ids":["src-discovery-ccsb-scripps-autodock-vina"],"status":"discovered"}
{"attributes":{"access":"official_source_linked","benchmark_applicability":"candidate; not evidence of a reported evaluation","candidate_benchmark_ids":[],"entity_level":"family","reported_name":"Basenji","version":null,"historical_missing_metadata":{"checkpoint":"unextracted","code_licence":"unextracted","parameters":"unextracted","training_cutoff":"unextracted","training_data":"unextracted","version":"unextracted","weights_licence":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Basenji predicts quantitative regulatory activity along DNA and scores the effects of sequence changes.","summary_source_ids":["evidence-official-dc53e8df56bb7352c014","evidence-official-edce0c06db05117706cb"],"summary_source_locator":"README.md: Basenji and Basset successor","sections":[{"title":"How it works","body":"Basenji predicts quantitative regulatory activity along DNA and scores the effects of sequence changes. Deep convolutional sequence model with binned regression outputs. The documented inputs are DNA sequence and, for training, aligned quantitative regulatory measurements. The output consists of predicted regulatory signal across sequence bins and derived variant-effect scores.","source_ids":["evidence-official-dc53e8df56bb7352c014","evidence-official-edce0c06db05117706cb"],"source_locator":"README.md: Basenji and Basset successor"},{"title":"Versions and reproducibility","body":"Basenji implementation; Basset, Akita and Saluki are described separately. The applicable input limits require configuration-specific checking.","source_ids":["evidence-official-dc53e8df56bb7352c014","evidence-official-edce0c06db05117706cb"],"source_locator":"README.md: Basenji and Basset successor"}],"facts":[{"label":"Model type","value":"Dilated convolutional genomic-track predictor","status":"source_checked","source_ids":["evidence-official-dc53e8df56bb7352c014","evidence-official-edce0c06db05117706cb"],"source_locator":"README.md: Basenji and Basset successor"},{"label":"Architecture","value":"Deep convolutional sequence model with binned regression outputs.","status":"source_checked","source_ids":["evidence-official-dc53e8df56bb7352c014","evidence-official-edce0c06db05117706cb"],"source_locator":"README.md: Basenji and Basset successor"},{"label":"Inputs","value":"DNA sequence and, for training, aligned quantitative regulatory measurements.","status":"source_checked","source_ids":["evidence-official-dc53e8df56bb7352c014","evidence-official-edce0c06db05117706cb"],"source_locator":"README.md: Basenji and Basset successor"},{"label":"Outputs","value":"Predicted regulatory signal across sequence bins and derived variant-effect scores.","status":"source_checked","source_ids":["evidence-official-dc53e8df56bb7352c014","evidence-official-edce0c06db05117706cb"],"source_locator":"README.md: Basenji and Basset successor"},{"label":"Parameters","value":"The repository describes a configurable Basenji model family rather than one checkpoint with a common parameter count; the selected model configuration is required.","status":"unreported","source_ids":["evidence-official-dc53e8df56bb7352c014","evidence-official-edce0c06db05117706cb"],"source_locator":"README.md: Basenji and Basset successor"},{"label":"Known versions","value":"Basenji implementation; Basset, Akita and Saluki are described separately.","status":"source_checked","source_ids":["evidence-official-dc53e8df56bb7352c014","evidence-official-edce0c06db05117706cb"],"source_locator":"README.md: Basenji and Basset successor"},{"label":"Training data","value":"Chosen regulatory-activity datasets; the README points to preprocessing/training tutorials rather than identifying one universal checkpoint.","status":"source_checked","source_ids":["evidence-official-dc53e8df56bb7352c014","evidence-official-edce0c06db05117706cb"],"source_locator":"README.md: Basenji and Basset successor"},{"label":"Training cutoff","value":"No checkpoint is selected by this family record. Training-track accessions and collection dates must be taken from the chosen Basenji release, not inferred from the repository date.","status":"unreported","source_ids":["evidence-official-dc53e8df56bb7352c014","evidence-official-edce0c06db05117706cb"],"source_locator":"README.md: Basenji and Basset successor"},{"label":"Context limits","value":"The inspected family README does not fix one input and output window across Basenji configurations. A run must preserve its sequence length, pooling and output-bin settings.","status":"unreported","source_ids":["evidence-official-dc53e8df56bb7352c014","evidence-official-edce0c06db05117706cb"],"source_locator":"README.md: Basenji and Basset successor"},{"label":"Weights licence","value":"Separate checkpoint-distribution terms are not stated in the inspected release documentation and licence material. The source-code licence alone is not recorded as an explicit weight grant.","status":"unreported","source_ids":["evidence-official-dc53e8df56bb7352c014","evidence-official-edce0c06db05117706cb","evidence-official-fe7d7d8a007d344f589c"],"source_locator":"README.md: Basenji and Basset successor; LICENSE: licence text"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/calico/basenji","status":"source_checked","source_ids":["evidence-official-dc53e8df56bb7352c014","evidence-official-edce0c06db05117706cb"],"source_locator":"README.md: Basenji and Basset successor"},{"label":"Code licence","value":"Apache-2.0","status":"source_checked","source_ids":["evidence-official-fe7d7d8a007d344f589c"],"source_locator":"LICENSE: licence text"}],"strengths":[{"text":"The same framework supports quantitative signal prediction and nucleotide/variant attribution workflows.","source_ids":["evidence-official-dc53e8df56bb7352c014","evidence-official-edce0c06db05117706cb"],"source_locator":"README.md: Basenji and Basset successor"}],"limitations":[{"text":"The repository also contains Akita and Saluki, which are separate models. Their inputs and tasks must not be assigned to Basenji simply because they share a repository.","source_ids":["evidence-official-dc53e8df56bb7352c014","evidence-official-edce0c06db05117706cb"],"source_locator":"README.md: Basenji and Basset successor"}],"diagram":{"title":"Basenji workflow","steps":["DNA sequence","Convolutional sequence model","Binned regulatory signal","Optional variant comparison"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-dc53e8df56bb7352c014","evidence-official-edce0c06db05117706cb"],"source_locator":"README.md: Basenji and Basset successor"},"coverage":"limited","gaps":["Parameters: The repository describes a configurable Basenji model family rather than one checkpoint with a common parameter count; the selected model configuration is required.","Training cutoff: No checkpoint is selected by this family record. Training-track accessions and collection dates must be taken from the chosen Basenji release, not inferred from the repository date.","Context limits: The inspected family README does not fix one input and output window across Basenji configurations. A run must preserve its sequence length, pooling and output-bin settings.","Weights licence: Separate checkpoint-distribution terms are not stated in the inspected release documentation and licence material. The source-code licence alone is not recorded as an explicit weight grant."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}},"description":"Sequence-to-regulatory-profile prediction","facets":{"areas":["genomics"]},"id":"discovery-model-basenji","kind":"model","links":[],"name":"Basenji","source_ids":["src-discovery-calico-basenji"],"status":"discovered"}
{"attributes":{"access":"official_source_linked","benchmark_applicability":"candidate; not evidence of a reported evaluation","candidate_benchmark_ids":["discovery-benchmark-plinder"],"entity_level":"family","reported_name":"Boltz","version":null,"historical_missing_metadata":{"checkpoint":"unextracted","code_licence":"unextracted","parameters":"unextracted","training_cutoff":"unextracted","training_data":"unextracted","version":"unextracted","weights_licence":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Boltz predicts biomolecular complex structures; Boltz-2 also predicts binding affinity.","summary_source_ids":["evidence-official-f780521ccebbb0691f9e","evidence-official-5a9e55abdf288d9dd8da","evidence-official-39a15ed8072aeea14a22","evidence-official-734117ed8501eb26fefc","evidence-official-7cb8cb7f091536b92c11","evidence-official-bd6ec1f819e22f075c32"],"summary_source_locator":"src/boltz/model/models/boltz2.py: Boltz2.__init__, forward, PairformerModule, AtomDiffusion and AffinityModule; docs/prediction.md: affinity outputs; Boltz-2 report Sections 2 Data, 3 Architecture, 4 Training and 6 Limitations","sections":[{"title":"How it works","body":"Boltz-2 first encodes the molecular inputs, alignments and optional templates into token and pair features. A Pairformer trunk updates those features and conditions atom-coordinate diffusion to generate a complex. Separate confidence and affinity modules assess the prediction; affinity classification and regression outputs answer different questions. This describes the Boltz-2 generation; a Boltz-1 result must retain its original checkpoint and prediction procedure.","source_ids":["evidence-official-f780521ccebbb0691f9e","evidence-official-5a9e55abdf288d9dd8da","evidence-official-39a15ed8072aeea14a22","evidence-official-734117ed8501eb26fefc","evidence-official-7cb8cb7f091536b92c11","evidence-official-bd6ec1f819e22f075c32"],"source_locator":"src/boltz/model/models/boltz2.py: Boltz2.__init__, forward, PairformerModule, AtomDiffusion and AffinityModule; docs/prediction.md: affinity outputs; Boltz-2 report Sections 2 Data, 3 Architecture, 4 Training and 6 Limitations"},{"title":"Versions and reproducibility","body":"Boltz-1 and Boltz-2 are distinct released generations; the catalogue does not select an evaluated checkpoint. The Boltz-2 report describes training crops up to 768 tokens. This is a training-crop size rather than a universal inference maximum; affinity additionally uses a pocket crop.","source_ids":["evidence-official-f780521ccebbb0691f9e","evidence-official-5a9e55abdf288d9dd8da","evidence-official-39a15ed8072aeea14a22","evidence-official-734117ed8501eb26fefc","evidence-official-7cb8cb7f091536b92c11","evidence-official-bd6ec1f819e22f075c32"],"source_locator":"src/boltz/model/models/boltz2.py: Boltz2.__init__, forward, PairformerModule, AtomDiffusion and AffinityModule; docs/prediction.md: affinity outputs; Boltz-2 report Sections 2 Data, 3 Architecture, 4 Training and 6 Limitations"}],"facts":[{"label":"Model type","value":"Biomolecular structure predictor family","status":"source_checked","source_ids":["evidence-official-f780521ccebbb0691f9e","evidence-official-5a9e55abdf288d9dd8da","evidence-official-39a15ed8072aeea14a22","evidence-official-734117ed8501eb26fefc","evidence-official-7cb8cb7f091536b92c11","evidence-official-bd6ec1f819e22f075c32"],"source_locator":"src/boltz/model/models/boltz2.py: Boltz2.__init__, forward, PairformerModule, AtomDiffusion and AffinityModule; docs/prediction.md: affinity outputs; Boltz-2 report Sections 2 Data, 3 Architecture, 4 Training and 6 Limitations"},{"label":"Architecture","value":"Boltz-1 and Boltz-2 are separate generations. In the inspected Boltz-2 implementation, molecular/MSA/template embeddings enter a Pairformer trunk, which conditions atom-coordinate diffusion; confidence and affinity are separate output modules.","status":"source_checked","source_ids":["evidence-official-f780521ccebbb0691f9e","evidence-official-5a9e55abdf288d9dd8da","evidence-official-39a15ed8072aeea14a22","evidence-official-734117ed8501eb26fefc","evidence-official-7cb8cb7f091536b92c11","evidence-official-bd6ec1f819e22f075c32"],"source_locator":"src/boltz/model/models/boltz2.py: Boltz2.__init__, forward, PairformerModule, AtomDiffusion and AffinityModule; docs/prediction.md: affinity outputs; Boltz-2 report Sections 2 Data, 3 Architecture, 4 Training and 6 Limitations"},{"label":"Inputs","value":"Protein, nucleic-acid and ligand specifications in prediction input files.","status":"source_checked","source_ids":["evidence-official-f780521ccebbb0691f9e","evidence-official-5a9e55abdf288d9dd8da","evidence-official-39a15ed8072aeea14a22","evidence-official-734117ed8501eb26fefc","evidence-official-7cb8cb7f091536b92c11","evidence-official-bd6ec1f819e22f075c32"],"source_locator":"src/boltz/model/models/boltz2.py: Boltz2.__init__, forward, PairformerModule, AtomDiffusion and AffinityModule; docs/prediction.md: affinity outputs; Boltz-2 report Sections 2 Data, 3 Architecture, 4 Training and 6 Limitations"},{"label":"Outputs","value":"Predicted complex structures and, for supported Boltz-2 inputs, binding-affinity predictions.","status":"source_checked","source_ids":["evidence-official-f780521ccebbb0691f9e","evidence-official-5a9e55abdf288d9dd8da","evidence-official-39a15ed8072aeea14a22","evidence-official-734117ed8501eb26fefc","evidence-official-7cb8cb7f091536b92c11","evidence-official-bd6ec1f819e22f075c32"],"source_locator":"src/boltz/model/models/boltz2.py: Boltz2.__init__, forward, PairformerModule, AtomDiffusion and AffinityModule; docs/prediction.md: affinity outputs; Boltz-2 report Sections 2 Data, 3 Architecture, 4 Training and 6 Limitations"},{"label":"Parameters","value":"A complete parameter total is not stated in the reviewed Boltz-2 architecture report or model constructor; structure, confidence and affinity are separate modules.","status":"unreported","source_ids":["evidence-official-f780521ccebbb0691f9e","evidence-official-5a9e55abdf288d9dd8da","evidence-official-39a15ed8072aeea14a22","evidence-official-734117ed8501eb26fefc","evidence-official-7cb8cb7f091536b92c11","evidence-official-bd6ec1f819e22f075c32"],"source_locator":"src/boltz/model/models/boltz2.py: Boltz2.__init__, forward, PairformerModule, AtomDiffusion and AffinityModule; docs/prediction.md: affinity outputs; Boltz-2 report Sections 2 Data, 3 Architecture, 4 Training and 6 Limitations"},{"label":"Known versions","value":"Boltz-1 and Boltz-2 are distinct released generations; the catalogue does not select an evaluated checkpoint.","status":"source_checked","source_ids":["evidence-official-f780521ccebbb0691f9e","evidence-official-5a9e55abdf288d9dd8da","evidence-official-39a15ed8072aeea14a22","evidence-official-734117ed8501eb26fefc","evidence-official-7cb8cb7f091536b92c11","evidence-official-bd6ec1f819e22f075c32"],"source_locator":"src/boltz/model/models/boltz2.py: Boltz2.__init__, forward, PairformerModule, AtomDiffusion and AffinityModule; docs/prediction.md: affinity outputs; Boltz-2 report Sections 2 Data, 3 Architecture, 4 Training and 6 Limitations"},{"label":"Training data","value":"Boltz-2 structure training combines pre-June-2023 PDB entries, MISATO/ATLAS/mdCATH molecular dynamics, and AlphaFold2/Boltz-1 distillation. Separate affinity training uses curated PubChem, ChEMBL, BindingDB, HTS, CeMM and MIDAS evidence with different regression/classification labels.","status":"source_checked","source_ids":["evidence-official-f780521ccebbb0691f9e","evidence-official-5a9e55abdf288d9dd8da","evidence-official-39a15ed8072aeea14a22","evidence-official-734117ed8501eb26fefc","evidence-official-7cb8cb7f091536b92c11","evidence-official-bd6ec1f819e22f075c32"],"source_locator":"src/boltz/model/models/boltz2.py: Boltz2.__init__, forward, PairformerModule, AtomDiffusion and AffinityModule; docs/prediction.md: affinity outputs; Boltz-2 report Sections 2 Data, 3 Architecture, 4 Training and 6 Limitations"},{"label":"Training cutoff","value":"Boltz-2 experimental PDB structures were released before 2023-06-01. This is not a shared cutoff for every affinity, MD or distilled resource, nor a Boltz-1 training cutoff.","status":"source_checked","source_ids":["evidence-official-f780521ccebbb0691f9e","evidence-official-5a9e55abdf288d9dd8da","evidence-official-39a15ed8072aeea14a22","evidence-official-734117ed8501eb26fefc","evidence-official-7cb8cb7f091536b92c11","evidence-official-bd6ec1f819e22f075c32"],"source_locator":"src/boltz/model/models/boltz2.py: Boltz2.__init__, forward, PairformerModule, AtomDiffusion and AffinityModule; docs/prediction.md: affinity outputs; Boltz-2 report Sections 2 Data, 3 Architecture, 4 Training and 6 Limitations"},{"label":"Context limits","value":"The Boltz-2 report describes training crops up to 768 tokens. This is a training-crop size rather than a universal inference maximum; affinity additionally uses a pocket crop.","status":"source_checked","source_ids":["evidence-official-f780521ccebbb0691f9e","evidence-official-5a9e55abdf288d9dd8da","evidence-official-39a15ed8072aeea14a22","evidence-official-734117ed8501eb26fefc","evidence-official-7cb8cb7f091536b92c11","evidence-official-bd6ec1f819e22f075c32"],"source_locator":"src/boltz/model/models/boltz2.py: Boltz2.__init__, forward, PairformerModule, AtomDiffusion and AffinityModule; docs/prediction.md: affinity outputs; Boltz-2 report Sections 2 Data, 3 Architecture, 4 Training and 6 Limitations"},{"label":"Weights licence","value":"MIT; the README explicitly applies this licence to code and model weights.","status":"source_checked","source_ids":["evidence-official-f780521ccebbb0691f9e","evidence-official-5a9e55abdf288d9dd8da","evidence-official-39a15ed8072aeea14a22","evidence-official-734117ed8501eb26fefc","evidence-official-7cb8cb7f091536b92c11","evidence-official-bd6ec1f819e22f075c32"],"source_locator":"src/boltz/model/models/boltz2.py: Boltz2.__init__, forward, PairformerModule, AtomDiffusion and AffinityModule; docs/prediction.md: affinity outputs; Boltz-2 report Sections 2 Data, 3 Architecture, 4 Training and 6 Limitations"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/jwohlwend/boltz","status":"source_checked","source_ids":["evidence-official-f780521ccebbb0691f9e","evidence-official-5a9e55abdf288d9dd8da","evidence-official-39a15ed8072aeea14a22","evidence-official-734117ed8501eb26fefc","evidence-official-7cb8cb7f091536b92c11","evidence-official-bd6ec1f819e22f075c32"],"source_locator":"src/boltz/model/models/boltz2.py: Boltz2.__init__, forward, PairformerModule, AtomDiffusion and AffinityModule; docs/prediction.md: affinity outputs; Boltz-2 report Sections 2 Data, 3 Architecture, 4 Training and 6 Limitations"},{"label":"Code licence","value":"MIT","status":"source_checked","source_ids":["evidence-official-8a985eabfd054f0dec4d"],"source_locator":"LICENSE: licence text"}],"strengths":[{"text":"The project distributes prediction code, model weights and training instructions.","source_ids":["evidence-official-f780521ccebbb0691f9e","evidence-official-5a9e55abdf288d9dd8da","evidence-official-39a15ed8072aeea14a22","evidence-official-734117ed8501eb26fefc","evidence-official-7cb8cb7f091536b92c11","evidence-official-bd6ec1f819e22f075c32"],"source_locator":"src/boltz/model/models/boltz2.py: Boltz2.__init__, forward, PairformerModule, AtomDiffusion and AffinityModule; docs/prediction.md: affinity outputs; Boltz-2 report Sections 2 Data, 3 Architecture, 4 Training and 6 Limitations"}],"limitations":[{"text":"Affinity predictions depend on a plausible binding pose and mix biochemical endpoint types. The report notes limited handling of cofactors, water and multimeric binding partners, and substantial variation between assays.","source_ids":["evidence-official-f780521ccebbb0691f9e","evidence-official-5a9e55abdf288d9dd8da","evidence-official-39a15ed8072aeea14a22","evidence-official-734117ed8501eb26fefc","evidence-official-7cb8cb7f091536b92c11","evidence-official-bd6ec1f819e22f075c32"],"source_locator":"src/boltz/model/models/boltz2.py: Boltz2.__init__, forward, PairformerModule, AtomDiffusion and AffinityModule; docs/prediction.md: affinity outputs; Boltz-2 report Sections 2 Data, 3 Architecture, 4 Training and 6 Limitations"}],"diagram":{"title":"Boltz workflow","steps":["Molecular inputs, MSA and templates","Token and pair embeddings","Pairformer trunk","Atom-coordinate diffusion","Structure, confidence and affinity"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-f780521ccebbb0691f9e","evidence-official-5a9e55abdf288d9dd8da","evidence-official-39a15ed8072aeea14a22","evidence-official-734117ed8501eb26fefc","evidence-official-7cb8cb7f091536b92c11","evidence-official-bd6ec1f819e22f075c32"],"source_locator":"src/boltz/model/models/boltz2.py: Boltz2.__init__, forward, PairformerModule, AtomDiffusion and AffinityModule; docs/prediction.md: affinity outputs; Boltz-2 report Sections 2 Data, 3 Architecture, 4 Training and 6 Limitations"},"coverage":"limited","gaps":["Parameters: A complete parameter total is not stated in the reviewed Boltz-2 architecture report or model constructor; structure, confidence and affinity are separate modules."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}},"description":"Biomolecular structure and affinity model family","facets":{"areas":["molecular-interactions"]},"id":"discovery-model-boltz","kind":"model","links":[{"relation":"applicable_to","target_id":"discovery-benchmark-plinder"}],"name":"Boltz","source_ids":["src-discovery-jwohlwend-boltz"],"status":"discovered"}
{"attributes":{"access":"official_source_linked","benchmark_applicability":"candidate; not evidence of a reported evaluation","candidate_benchmark_ids":["discovery-benchmark-cami"],"entity_level":"method","reported_name":"CAMISIM","version":null,"historical_missing_metadata":{"checkpoint":"unextracted","code_licence":"unextracted","parameters":"unextracted","training_cutoff":"unextracted","training_data":"unextracted","version":"unextracted","weights_licence":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"CAMISIM creates simulated microbial communities and corresponding shotgun metagenomic datasets.","summary_source_ids":["evidence-official-e604b144dfbd97bf8917"],"summary_source_locator":"README.md: overview and CAMISIM 2.0","sections":[{"title":"How it works","body":"CAMISIM creates simulated microbial communities and corresponding shotgun metagenomic datasets. Community-abundance simulation followed by metagenomic read generation; CAMISIM 2 uses a Nextflow workflow. The documented inputs are chosen genomes, community-abundance settings and simulation configuration. The output consists of simulated shotgun metagenomic datasets with known generating communities.","source_ids":["evidence-official-e604b144dfbd97bf8917"],"source_locator":"README.md: overview and CAMISIM 2.0"},{"title":"Versions and reproducibility","body":"CAMISIM 2.0 Nextflow workflow; legacy Python version retained as 1.31-final. Simulation size and read-generation configuration, not a neural token window.","source_ids":["evidence-official-e604b144dfbd97bf8917"],"source_locator":"README.md: overview and CAMISIM 2.0"}],"facts":[{"label":"Model type","value":"Microbial community simulation software","status":"source_checked","source_ids":["evidence-official-e604b144dfbd97bf8917"],"source_locator":"README.md: overview and CAMISIM 2.0"},{"label":"Architecture","value":"Community-abundance simulation followed by metagenomic read generation; CAMISIM 2 uses a Nextflow workflow.","status":"source_checked","source_ids":["evidence-official-e604b144dfbd97bf8917"],"source_locator":"README.md: overview and CAMISIM 2.0"},{"label":"Inputs","value":"Chosen genomes, community-abundance settings and simulation configuration.","status":"source_checked","source_ids":["evidence-official-e604b144dfbd97bf8917"],"source_locator":"README.md: overview and CAMISIM 2.0"},{"label":"Outputs","value":"Simulated shotgun metagenomic datasets with known generating communities.","status":"source_checked","source_ids":["evidence-official-e604b144dfbd97bf8917"],"source_locator":"README.md: overview and CAMISIM 2.0"},{"label":"Parameters","value":"Inapplicable as a neural parameter count; simulation settings must be pinned.","status":"inapplicable","source_ids":["evidence-official-e604b144dfbd97bf8917"],"source_locator":"README.md: overview and CAMISIM 2.0"},{"label":"Known versions","value":"CAMISIM 2.0 Nextflow workflow; legacy Python version retained as 1.31-final.","status":"source_checked","source_ids":["evidence-official-e604b144dfbd97bf8917"],"source_locator":"README.md: overview and CAMISIM 2.0"},{"label":"Training data","value":"No neural pretraining; selected input genomes determine the simulation source material.","status":"source_checked","source_ids":["evidence-official-e604b144dfbd97bf8917"],"source_locator":"README.md: overview and CAMISIM 2.0"},{"label":"Training cutoff","value":"Inapplicable to neural pretraining; reference-database and input-data dates must be recorded for each run.","status":"inapplicable","source_ids":["evidence-official-e604b144dfbd97bf8917"],"source_locator":"README.md: overview and CAMISIM 2.0"},{"label":"Context limits","value":"Simulation size and read-generation configuration, not a neural token window.","status":"source_checked","source_ids":["evidence-official-e604b144dfbd97bf8917"],"source_locator":"README.md: overview and CAMISIM 2.0"},{"label":"Weights licence","value":"Inapplicable: simulator rather than a pretrained predictor.","status":"inapplicable","source_ids":["evidence-official-e604b144dfbd97bf8917"],"source_locator":"README.md: overview and CAMISIM 2.0"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/CAMI-challenge/CAMISIM","status":"source_checked","source_ids":["evidence-official-e604b144dfbd97bf8917"],"source_locator":"README.md: overview and CAMISIM 2.0"},{"label":"Code licence","value":"Apache-2.0","status":"source_checked","source_ids":["evidence-official-1005311cb2f598fe63de"],"source_locator":"LICENSE.txt: licence text"}],"strengths":[{"text":"Provides controlled synthetic data useful for evaluating metagenomic analysis methods.","source_ids":["evidence-official-e604b144dfbd97bf8917"],"source_locator":"README.md: overview and CAMISIM 2.0"}],"limitations":[{"text":"Simulation realism depends on its inputs and configuration. The authors advise checking converted CAMISIM 1 configurations rather than assuming automatic compatibility.","source_ids":["evidence-official-e604b144dfbd97bf8917"],"source_locator":"README.md: overview and CAMISIM 2.0"}],"diagram":{"title":"CAMISIM workflow","steps":["Reference genomes","Community abundance model","Read simulation","Synthetic metagenome"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-e604b144dfbd97bf8917"],"source_locator":"README.md: overview and CAMISIM 2.0"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}},"description":"Microbial community and metagenome simulation","facets":{"areas":["microbiome"]},"id":"discovery-model-camisim","kind":"model","links":[{"relation":"applicable_to","target_id":"discovery-benchmark-cami"}],"name":"CAMISIM","source_ids":["src-discovery-cami-challenge-camisim"],"status":"discovered"}
{"attributes":{"access":"official_source_linked","benchmark_applicability":"candidate; not evidence of a reported evaluation","candidate_benchmark_ids":["discovery-benchmark-dart-eval"],"entity_level":"family","reported_name":"ChromBPNet","version":null,"historical_missing_metadata":{"checkpoint":"unextracted","code_licence":"unextracted","parameters":"unextracted","training_cutoff":"unextracted","training_data":"unextracted","version":"unextracted","weights_licence":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"ChromBPNet predicts base-resolution chromatin accessibility while modeling assay-specific enzyme bias separately.","summary_source_ids":["evidence-official-a766fb9176a491dc1ec3","evidence-official-bc3f1b0fa5b0999bbfa0"],"summary_source_locator":"README.md: introductory explanation and Bias-factorized ChromBPNet training; chrombpnet/training/models/chrombpnet_with_bias_model.py: bpnet_model and model_params","sections":[{"title":"How it works","body":"ChromBPNet predicts base-resolution chromatin accessibility while modeling assay-specific enzyme bias separately. Residual dilated convolutional network with a frozen bias model and a transcription-factor sequence component. The documented inputs are DNA sequences and ATAC-seq or DNase-seq profiles for the configured assay. The output consists of predicted accessibility profiles and quantities used to study sequence contributions and variants.","source_ids":["evidence-official-a766fb9176a491dc1ec3","evidence-official-bc3f1b0fa5b0999bbfa0"],"source_locator":"README.md: introductory explanation and Bias-factorized ChromBPNet training; chrombpnet/training/models/chrombpnet_with_bias_model.py: bpnet_model and model_params"},{"title":"Versions and reproducibility","body":"Version-dependent trained models; README highlights a motif-discovery note for versions at or below 0.1.3. inputlen and outputlen are explicit model configuration fields. Record both the input sequence window and the central output window for the selected trained model.","source_ids":["evidence-official-a766fb9176a491dc1ec3","evidence-official-bc3f1b0fa5b0999bbfa0"],"source_locator":"README.md: introductory explanation and Bias-factorized ChromBPNet training; chrombpnet/training/models/chrombpnet_with_bias_model.py: bpnet_model and model_params"}],"facts":[{"label":"Model type","value":"Bias-factorized convolutional chromatin-profile predictor","status":"source_checked","source_ids":["evidence-official-a766fb9176a491dc1ec3","evidence-official-bc3f1b0fa5b0999bbfa0"],"source_locator":"README.md: introductory explanation and Bias-factorized ChromBPNet training; chrombpnet/training/models/chrombpnet_with_bias_model.py: bpnet_model and model_params"},{"label":"Architecture","value":"Residual dilated convolutional network with a frozen bias model and a transcription-factor sequence component.","status":"source_checked","source_ids":["evidence-official-a766fb9176a491dc1ec3","evidence-official-bc3f1b0fa5b0999bbfa0"],"source_locator":"README.md: introductory explanation and Bias-factorized ChromBPNet training; chrombpnet/training/models/chrombpnet_with_bias_model.py: bpnet_model and model_params"},{"label":"Inputs","value":"DNA sequences and ATAC-seq or DNase-seq profiles for the configured assay.","status":"source_checked","source_ids":["evidence-official-a766fb9176a491dc1ec3","evidence-official-bc3f1b0fa5b0999bbfa0"],"source_locator":"README.md: introductory explanation and Bias-factorized ChromBPNet training; chrombpnet/training/models/chrombpnet_with_bias_model.py: bpnet_model and model_params"},{"label":"Outputs","value":"Predicted accessibility profiles and quantities used to study sequence contributions and variants.","status":"source_checked","source_ids":["evidence-official-a766fb9176a491dc1ec3","evidence-official-bc3f1b0fa5b0999bbfa0"],"source_locator":"README.md: introductory explanation and Bias-factorized ChromBPNet training; chrombpnet/training/models/chrombpnet_with_bias_model.py: bpnet_model and model_params"},{"label":"Parameters","value":"Configuration-dependent: convolutional filter count and number of dilated layers are supplied in model_params; the bias component is separate.","status":"source_checked","source_ids":["evidence-official-a766fb9176a491dc1ec3","evidence-official-bc3f1b0fa5b0999bbfa0"],"source_locator":"README.md: introductory explanation and Bias-factorized ChromBPNet training; chrombpnet/training/models/chrombpnet_with_bias_model.py: bpnet_model and model_params"},{"label":"Known versions","value":"Version-dependent trained models; README highlights a motif-discovery note for versions at or below 0.1.3.","status":"source_checked","source_ids":["evidence-official-a766fb9176a491dc1ec3","evidence-official-bc3f1b0fa5b0999bbfa0"],"source_locator":"README.md: introductory explanation and Bias-factorized ChromBPNet training; chrombpnet/training/models/chrombpnet_with_bias_model.py: bpnet_model and model_params"},{"label":"Training data","value":"Two-stage fitting: background regions for enzyme bias, then accessibility-profile training for the sequence model.","status":"source_checked","source_ids":["evidence-official-a766fb9176a491dc1ec3","evidence-official-bc3f1b0fa5b0999bbfa0"],"source_locator":"README.md: introductory explanation and Bias-factorized ChromBPNet training; chrombpnet/training/models/chrombpnet_with_bias_model.py: bpnet_model and model_params"},{"label":"Training cutoff","value":"Inapplicable as a universal pretraining date: the workflow fits the supplied chromatin tracks and bias model; their accessions, dates and split belong to the individual evaluation.","status":"inapplicable","source_ids":["evidence-official-a766fb9176a491dc1ec3","evidence-official-bc3f1b0fa5b0999bbfa0"],"source_locator":"README.md: introductory explanation and Bias-factorized ChromBPNet training; chrombpnet/training/models/chrombpnet_with_bias_model.py: bpnet_model and model_params"},{"label":"Context limits","value":"inputlen and outputlen are explicit model configuration fields. Record both the input sequence window and the central output window for the selected trained model.","status":"source_checked","source_ids":["evidence-official-a766fb9176a491dc1ec3","evidence-official-bc3f1b0fa5b0999bbfa0"],"source_locator":"README.md: introductory explanation and Bias-factorized ChromBPNet training; chrombpnet/training/models/chrombpnet_with_bias_model.py: bpnet_model and model_params"},{"label":"Weights licence","value":"Separate checkpoint-distribution terms are not stated in the inspected release documentation and licence material. The source-code licence alone is not recorded as an explicit weight grant.","status":"unreported","source_ids":["evidence-official-a766fb9176a491dc1ec3","evidence-official-bc3f1b0fa5b0999bbfa0","evidence-official-45f05eebcf7a03ac99cf"],"source_locator":"README.md: introductory explanation and Bias-factorized ChromBPNet training; chrombpnet/training/models/chrombpnet_with_bias_model.py: bpnet_model and model_params; LICENSE: licence text"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/kundajelab/chrombpnet","status":"source_checked","source_ids":["evidence-official-a766fb9176a491dc1ec3","evidence-official-bc3f1b0fa5b0999bbfa0"],"source_locator":"README.md: introductory explanation and Bias-factorized ChromBPNet training; chrombpnet/training/models/chrombpnet_with_bias_model.py: bpnet_model and model_params"},{"label":"Code licence","value":"MIT","status":"source_checked","source_ids":["evidence-official-45f05eebcf7a03ac99cf"],"source_locator":"LICENSE: licence text"}],"strengths":[{"text":"The two-stage procedure explicitly separates a learned assay-bias component from the sequence component of interest.","source_ids":["evidence-official-a766fb9176a491dc1ec3","evidence-official-bc3f1b0fa5b0999bbfa0"],"source_locator":"README.md: introductory explanation and Bias-factorized ChromBPNet training; chrombpnet/training/models/chrombpnet_with_bias_model.py: bpnet_model and model_params"}],"limitations":[{"text":"The bias model must correspond to the assay. The model is fitted to a specified context, so a checkpoint cannot be assumed to apply equally across assays or cell types.","source_ids":["evidence-official-a766fb9176a491dc1ec3","evidence-official-bc3f1b0fa5b0999bbfa0"],"source_locator":"README.md: introductory explanation and Bias-factorized ChromBPNet training; chrombpnet/training/models/chrombpnet_with_bias_model.py: bpnet_model and model_params"}],"diagram":{"title":"ChromBPNet workflow","steps":["Background assay data","Learn enzyme-bias model","Fit sequence model with frozen bias","Accessibility predictions"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-a766fb9176a491dc1ec3","evidence-official-bc3f1b0fa5b0999bbfa0"],"source_locator":"README.md: introductory explanation and Bias-factorized ChromBPNet training; chrombpnet/training/models/chrombpnet_with_bias_model.py: bpnet_model and model_params"},"coverage":"limited","gaps":["Weights licence: Separate checkpoint-distribution terms are not stated in the inspected release documentation and licence material. The source-code licence alone is not recorded as an explicit weight grant."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}},"description":"Bias-aware chromatin accessibility prediction","facets":{"areas":["genomics"]},"id":"discovery-model-chrombpnet","kind":"model","links":[{"relation":"applicable_to","target_id":"discovery-benchmark-dart-eval"}],"name":"ChromBPNet","source_ids":["src-discovery-kundajelab-chrombpnet"],"status":"discovered"}
{"attributes":{"access":"official_source_linked","benchmark_applicability":"candidate; not evidence of a reported evaluation","candidate_benchmark_ids":[],"entity_level":"method","reported_name":"COBRApy","version":null,"historical_missing_metadata":{"checkpoint":"unextracted","code_licence":"unextracted","parameters":"unextracted","training_cutoff":"unextracted","training_data":"unextracted","version":"unextracted","weights_licence":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"COBRApy supports constraint-based analysis of metabolic networks, including flux balance and gene-deletion analyses.","summary_source_ids":["evidence-official-eb328c1926fbe500f86b"],"summary_source_locator":"README.rst: What is COBRApy? and License","sections":[{"title":"How it works","body":"COBRApy supports constraint-based analysis of metabolic networks, including flux balance and gene-deletion analyses. Metabolic network representation coupled to constrained optimization through a selected solver. The documented inputs are A metabolic reconstruction, reaction constraints and an analysis objective. The output consists of flux solutions, feasible flux ranges and outputs of configured perturbation/essentiality analyses.","source_ids":["evidence-official-eb328c1926fbe500f86b"],"source_locator":"README.rst: What is COBRApy? and License"},{"title":"Versions and reproducibility","body":"COBRApy software and selected metabolic reconstruction must both be versioned. Network size and solver capacity; no sequence context.","source_ids":["evidence-official-eb328c1926fbe500f86b"],"source_locator":"README.rst: What is COBRApy? and License"}],"facts":[{"label":"Model type","value":"Constraint-based metabolic modelling software","status":"source_checked","source_ids":["evidence-official-eb328c1926fbe500f86b"],"source_locator":"README.rst: What is COBRApy? and License"},{"label":"Architecture","value":"Metabolic network representation coupled to constrained optimization through a selected solver.","status":"source_checked","source_ids":["evidence-official-eb328c1926fbe500f86b"],"source_locator":"README.rst: What is COBRApy? and License"},{"label":"Inputs","value":"A metabolic reconstruction, reaction constraints and an analysis objective.","status":"source_checked","source_ids":["evidence-official-eb328c1926fbe500f86b"],"source_locator":"README.rst: What is COBRApy? and License"},{"label":"Outputs","value":"Flux solutions, feasible flux ranges and outputs of configured perturbation/essentiality analyses.","status":"source_checked","source_ids":["evidence-official-eb328c1926fbe500f86b"],"source_locator":"README.rst: What is COBRApy? and License"},{"label":"Parameters","value":"Inapplicable as a neural parameter count; reaction bounds and model constraints are scientific inputs.","status":"inapplicable","source_ids":["evidence-official-eb328c1926fbe500f86b"],"source_locator":"README.rst: What is COBRApy? and License"},{"label":"Known versions","value":"COBRApy software and selected metabolic reconstruction must both be versioned.","status":"source_checked","source_ids":["evidence-official-eb328c1926fbe500f86b"],"source_locator":"README.rst: What is COBRApy? and License"},{"label":"Training data","value":"No neural pretraining; the metabolic reconstruction and experimental constraints are supplied separately.","status":"source_checked","source_ids":["evidence-official-eb328c1926fbe500f86b"],"source_locator":"README.rst: What is COBRApy? and License"},{"label":"Training cutoff","value":"Inapplicable to neural pretraining; reference-database and input-data dates must be recorded for each run.","status":"inapplicable","source_ids":["evidence-official-eb328c1926fbe500f86b"],"source_locator":"README.rst: What is COBRApy? and License"},{"label":"Context limits","value":"Network size and solver capacity; no sequence context.","status":"source_checked","source_ids":["evidence-official-eb328c1926fbe500f86b"],"source_locator":"README.rst: What is COBRApy? and License"},{"label":"Weights licence","value":"Inapplicable: no universal neural checkpoint.","status":"inapplicable","source_ids":["evidence-official-eb328c1926fbe500f86b"],"source_locator":"README.rst: What is COBRApy? and License"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/opencobra/cobrapy","status":"source_checked","source_ids":["evidence-official-eb328c1926fbe500f86b"],"source_locator":"README.rst: What is COBRApy? and License"},{"label":"Code licence","value":"GPL-2.0-or-later or LGPL-2.0-or-later, at the user’s choice, as stated by the README.","status":"source_checked","source_ids":["evidence-official-eb328c1926fbe500f86b"],"source_locator":"README.rst: License"}],"strengths":[{"text":"One software framework exposes FBA, flux variability analysis and related methods over reusable metabolic models.","source_ids":["evidence-official-eb328c1926fbe500f86b"],"source_locator":"README.rst: What is COBRApy? and License"}],"limitations":[{"text":"Results depend on the metabolic reconstruction, constraints, objective and solver; a library version alone does not define the biological model.","source_ids":["evidence-official-eb328c1926fbe500f86b"],"source_locator":"README.rst: What is COBRApy? and License"}],"diagram":{"title":"COBRApy workflow","steps":["Metabolic network","Constraints and objective","Optimization solver","Flux or perturbation analysis"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-eb328c1926fbe500f86b"],"source_locator":"README.rst: What is COBRApy? and License"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}},"description":"Constraint-based metabolic modelling","facets":{"areas":["mechanistic-biology"]},"id":"discovery-model-cobrapy","kind":"model","links":[],"name":"COBRApy","source_ids":["src-discovery-opencobra-cobrapy"],"status":"discovered"}
{"attributes":{"access":"official_source_linked","benchmark_applicability":"candidate; not evidence of a reported evaluation","candidate_benchmark_ids":["discovery-benchmark-casp"],"entity_level":"method","reported_name":"ColabFold","version":null,"historical_missing_metadata":{"checkpoint":"unextracted","code_licence":"unextracted","parameters":"unextracted","training_cutoff":"unextracted","training_data":"unextracted","version":"unextracted","weights_licence":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"ColabFold packages sequence-search and protein-folding workflows for notebooks and local batch prediction.","summary_source_ids":["evidence-official-efc93d867cbf62e80ebb"],"summary_source_locator":"README.md: notebook table and FAQ","sections":[{"title":"How it works","body":"ColabFold packages sequence-search and protein-folding workflows for notebooks and local batch prediction. A prediction pipeline that can combine MMseqs2 alignments with AlphaFold-family or other folding models; it is not a single neural architecture. The documented inputs are protein sequences, optionally templates and the selected alignment procedure. The output consists of predicted structures and confidence outputs from the chosen underlying folding model.","source_ids":["evidence-official-efc93d867cbf62e80ebb"],"source_locator":"README.md: notebook table and FAQ"},{"title":"Versions and reproducibility","body":"README identifies ColabFold 1.6.3 and separately lists AlphaFold2, OpenFold3, ESMFold and beta notebooks. Hardware- and predictor-dependent. The README gives approximate 16GB-GPU guidance, not a universal model context.","source_ids":["evidence-official-efc93d867cbf62e80ebb"],"source_locator":"README.md: notebook table and FAQ"}],"facts":[{"label":"Model type","value":"Sequence-search and folding pipeline","status":"source_checked","source_ids":["evidence-official-efc93d867cbf62e80ebb"],"source_locator":"README.md: notebook table and FAQ"},{"label":"Architecture","value":"A prediction pipeline that can combine MMseqs2 alignments with AlphaFold-family or other folding models; it is not a single neural architecture.","status":"source_checked","source_ids":["evidence-official-efc93d867cbf62e80ebb"],"source_locator":"README.md: notebook table and FAQ"},{"label":"Inputs","value":"Protein sequences, optionally templates and the selected alignment procedure.","status":"source_checked","source_ids":["evidence-official-efc93d867cbf62e80ebb"],"source_locator":"README.md: notebook table and FAQ"},{"label":"Outputs","value":"Predicted structures and confidence outputs from the chosen underlying folding model.","status":"source_checked","source_ids":["evidence-official-efc93d867cbf62e80ebb"],"source_locator":"README.md: notebook table and FAQ"},{"label":"Parameters","value":"Depends on the underlying folding model; no single ColabFold parameter total.","status":"source_checked","source_ids":["evidence-official-efc93d867cbf62e80ebb"],"source_locator":"README.md: notebook table and FAQ"},{"label":"Known versions","value":"README identifies ColabFold 1.6.3 and separately lists AlphaFold2, OpenFold3, ESMFold and beta notebooks.","status":"source_checked","source_ids":["evidence-official-efc93d867cbf62e80ebb"],"source_locator":"README.md: notebook table and FAQ"},{"label":"Training data","value":"ColabFold assembles existing prediction models and search databases; their provenance must be recorded separately.","status":"source_checked","source_ids":["evidence-official-efc93d867cbf62e80ebb"],"source_locator":"README.md: notebook table and FAQ"},{"label":"Training cutoff","value":"Inapplicable as one pipeline-wide date: the selected folding weights and the MSA/template databases have separate releases and cutoffs.","status":"inapplicable","source_ids":["evidence-official-efc93d867cbf62e80ebb"],"source_locator":"README.md: notebook table and FAQ"},{"label":"Context limits","value":"Hardware- and predictor-dependent. The README gives approximate 16GB-GPU guidance, not a universal model context.","status":"source_checked","source_ids":["evidence-official-efc93d867cbf62e80ebb"],"source_locator":"README.md: notebook table and FAQ"},{"label":"Weights licence","value":"Each underlying model has its own weight licence; repository code terms do not replace those licences.","status":"source_checked","source_ids":["evidence-official-efc93d867cbf62e80ebb"],"source_locator":"README.md: notebook table and FAQ"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/sokrypton/ColabFold","status":"source_checked","source_ids":["evidence-official-efc93d867cbf62e80ebb"],"source_locator":"README.md: notebook table and FAQ"},{"label":"Code licence","value":"MIT","status":"source_checked","source_ids":["evidence-official-f3fd16387ba8459ecf34"],"source_locator":"LICENSE: licence text"}],"strengths":[{"text":"Notebook and batch interfaces make several folding workflows available with documented input choices.","source_ids":["evidence-official-efc93d867cbf62e80ebb"],"source_locator":"README.md: notebook table and FAQ"}],"limitations":[{"text":"The repository distinguishes supported, beta and retired notebooks. GPU memory limits and MSA-server usage policies apply; the underlying predictor and database version must be named in a benchmark.","source_ids":["evidence-official-efc93d867cbf62e80ebb"],"source_locator":"README.md: notebook table and FAQ"}],"diagram":{"title":"ColabFold workflow","steps":["Protein sequences","Selected alignment search","Selected folding model","Structure and confidence"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-efc93d867cbf62e80ebb"],"source_locator":"README.md: notebook table and FAQ"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}},"description":"Protein folding pipeline","facets":{"areas":["protein-structure"]},"id":"discovery-model-colabfold","kind":"model","links":[{"relation":"applicable_to","target_id":"discovery-benchmark-casp"}],"name":"ColabFold","source_ids":["src-discovery-sokrypton-colabfold"],"status":"discovered"}
{"attributes":{"access":"official_source_linked","benchmark_applicability":"candidate; not evidence of a reported evaluation","candidate_benchmark_ids":["discovery-benchmark-gue"],"entity_level":"family","reported_name":"DNABERT-2","version":null,"historical_missing_metadata":{"checkpoint":"unextracted","code_licence":"unextracted","parameters":"unextracted","training_cutoff":"unextracted","training_data":"unextracted","version":"unextracted","weights_licence":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"DNABERT-2 learns DNA representations that can be adapted to genomic prediction tasks.","summary_source_ids":["evidence-official-96b5c3a31a50f7c259d6","evidence-official-25b222d11900e0e88a51","evidence-official-3aa749d3205c0a768b22"],"summary_source_locator":"DNABERT-2 paper Sections 3.2, 4.1 and 5.2 (Further Pre-Training); official model card and implementation README; official DNABERT-2-117M checkpoint LICENSE at revision 7bce263b15377fc15361f52cfab88f8b586abda0","sections":[{"title":"How it works","body":"DNABERT-2 merges recurring DNA substrings into byte-pair tokens, then processes those tokens with a masked-language-model transformer. ALiBi supplies distance-dependent attention biases, while FlashAttention changes how attention is computed. The resulting contextual embeddings need an explicit pooling rule and prediction head for a downstream task.","source_ids":["evidence-official-96b5c3a31a50f7c259d6","evidence-official-25b222d11900e0e88a51","evidence-official-3aa749d3205c0a768b22"],"source_locator":"DNABERT-2 paper Sections 3.2, 4.1 and 5.2 (Further Pre-Training); official model card and implementation README; official DNABERT-2-117M checkpoint LICENSE at revision 7bce263b15377fc15361f52cfab88f8b586abda0"},{"title":"Versions and reproducibility","body":"DNABERT-2-117M model card and official DNABERT_2 implementation. ALiBi permits inference beyond the pretraining sequence length, subject to attention/memory cost; this does not establish unlimited biological context or validated accuracy at arbitrary lengths.","source_ids":["evidence-official-96b5c3a31a50f7c259d6","evidence-official-25b222d11900e0e88a51","evidence-official-3aa749d3205c0a768b22"],"source_locator":"DNABERT-2 paper Sections 3.2, 4.1 and 5.2 (Further Pre-Training); official model card and implementation README; official DNABERT-2-117M checkpoint LICENSE at revision 7bce263b15377fc15361f52cfab88f8b586abda0"}],"facts":[{"label":"Model type","value":"Masked-token DNA transformer encoder","status":"source_checked","source_ids":["evidence-official-96b5c3a31a50f7c259d6","evidence-official-25b222d11900e0e88a51","evidence-official-3aa749d3205c0a768b22"],"source_locator":"DNABERT-2 paper Sections 3.2, 4.1 and 5.2 (Further Pre-Training); official model card and implementation README; official DNABERT-2-117M checkpoint LICENSE at revision 7bce263b15377fc15361f52cfab88f8b586abda0"},{"label":"Architecture","value":"BERT-style DNA encoder with byte-pair tokenization, ALiBi relative attention biases and FlashAttention; task heads and pooling are separately configured.","status":"source_checked","source_ids":["evidence-official-96b5c3a31a50f7c259d6","evidence-official-25b222d11900e0e88a51","evidence-official-3aa749d3205c0a768b22"],"source_locator":"DNABERT-2 paper Sections 3.2, 4.1 and 5.2 (Further Pre-Training); official model card and implementation README; official DNABERT-2-117M checkpoint LICENSE at revision 7bce263b15377fc15361f52cfab88f8b586abda0"},{"label":"Inputs","value":"DNA sequence tokenized with the supplied tokenizer.","status":"source_checked","source_ids":["evidence-official-96b5c3a31a50f7c259d6","evidence-official-25b222d11900e0e88a51","evidence-official-3aa749d3205c0a768b22"],"source_locator":"DNABERT-2 paper Sections 3.2, 4.1 and 5.2 (Further Pre-Training); official model card and implementation README; official DNABERT-2-117M checkpoint LICENSE at revision 7bce263b15377fc15361f52cfab88f8b586abda0"},{"label":"Outputs","value":"Token representations and, after a specified adaptation, task predictions.","status":"source_checked","source_ids":["evidence-official-96b5c3a31a50f7c259d6","evidence-official-25b222d11900e0e88a51","evidence-official-3aa749d3205c0a768b22"],"source_locator":"DNABERT-2 paper Sections 3.2, 4.1 and 5.2 (Further Pre-Training); official model card and implementation README; official DNABERT-2-117M checkpoint LICENSE at revision 7bce263b15377fc15361f52cfab88f8b586abda0"},{"label":"Parameters","value":"117 million for DNABERT-2-117M; family names do not establish a particular checkpoint.","status":"source_checked","source_ids":["evidence-official-96b5c3a31a50f7c259d6","evidence-official-25b222d11900e0e88a51","evidence-official-3aa749d3205c0a768b22"],"source_locator":"DNABERT-2 paper Sections 3.2, 4.1 and 5.2 (Further Pre-Training); official model card and implementation README; official DNABERT-2-117M checkpoint LICENSE at revision 7bce263b15377fc15361f52cfab88f8b586abda0"},{"label":"Known versions","value":"DNABERT-2-117M model card and official DNABERT_2 implementation.","status":"source_checked","source_ids":["evidence-official-96b5c3a31a50f7c259d6","evidence-official-25b222d11900e0e88a51","evidence-official-3aa749d3205c0a768b22"],"source_locator":"DNABERT-2 paper Sections 3.2, 4.1 and 5.2 (Further Pre-Training); official model card and implementation README; official DNABERT-2-117M checkpoint LICENSE at revision 7bce263b15377fc15361f52cfab88f8b586abda0"},{"label":"Training data","value":"The paper describes a 32.49-billion-base corpus covering 135 species in six groups, alongside a 2.75-billion-base human corpus. Further GUE-domain pretraining is a separately reported model variant.","status":"source_checked","source_ids":["evidence-official-96b5c3a31a50f7c259d6","evidence-official-25b222d11900e0e88a51","evidence-official-3aa749d3205c0a768b22"],"source_locator":"DNABERT-2 paper Sections 3.2, 4.1 and 5.2 (Further Pre-Training); official model card and implementation README; official DNABERT-2-117M checkpoint LICENSE at revision 7bce263b15377fc15361f52cfab88f8b586abda0"},{"label":"Training cutoff","value":"The paper identifies the human and multispecies genome corpora but does not state one latest-sequence deposition date in its reviewed pretraining-data sections.","status":"unreported","source_ids":["evidence-official-96b5c3a31a50f7c259d6","evidence-official-25b222d11900e0e88a51","evidence-official-3aa749d3205c0a768b22"],"source_locator":"DNABERT-2 paper Sections 3.2, 4.1 and 5.2 (Further Pre-Training); official model card and implementation README; official DNABERT-2-117M checkpoint LICENSE at revision 7bce263b15377fc15361f52cfab88f8b586abda0"},{"label":"Context limits","value":"ALiBi permits inference beyond the pretraining sequence length, subject to attention/memory cost; this does not establish unlimited biological context or validated accuracy at arbitrary lengths.","status":"source_checked","source_ids":["evidence-official-96b5c3a31a50f7c259d6","evidence-official-25b222d11900e0e88a51","evidence-official-3aa749d3205c0a768b22"],"source_locator":"DNABERT-2 paper Sections 3.2, 4.1 and 5.2 (Further Pre-Training); official model card and implementation README; official DNABERT-2-117M checkpoint LICENSE at revision 7bce263b15377fc15361f52cfab88f8b586abda0"},{"label":"Weights licence","value":"The official zhihan1996/DNABERT-2-117M checkpoint repository carries Apache-2.0 in its pinned LICENSE. This does not assign terms to a separately fitted downstream predictor.","status":"source_checked","source_ids":["evidence-official-96b5c3a31a50f7c259d6","evidence-official-25b222d11900e0e88a51","evidence-official-3aa749d3205c0a768b22"],"source_locator":"DNABERT-2 paper Sections 3.2, 4.1 and 5.2 (Further Pre-Training); official model card and implementation README; official DNABERT-2-117M checkpoint LICENSE at revision 7bce263b15377fc15361f52cfab88f8b586abda0"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/MAGICS-LAB/DNABERT_2","status":"source_checked","source_ids":["evidence-official-96b5c3a31a50f7c259d6","evidence-official-25b222d11900e0e88a51","evidence-official-3aa749d3205c0a768b22"],"source_locator":"DNABERT-2 paper Sections 3.2, 4.1 and 5.2 (Further Pre-Training); official model card and implementation README; official DNABERT-2-117M checkpoint LICENSE at revision 7bce263b15377fc15361f52cfab88f8b586abda0"},{"label":"Code licence","value":"Apache-2.0","status":"source_checked","source_ids":["evidence-official-25b222d11900e0e88a51"],"source_locator":"LICENSE: licence text"}],"strengths":[{"text":"The released model supports embedding extraction and task-specific fine-tuning.","source_ids":["evidence-official-96b5c3a31a50f7c259d6","evidence-official-25b222d11900e0e88a51","evidence-official-3aa749d3205c0a768b22"],"source_locator":"DNABERT-2 paper Sections 3.2, 4.1 and 5.2 (Further Pre-Training); official model card and implementation README; official DNABERT-2-117M checkpoint LICENSE at revision 7bce263b15377fc15361f52cfab88f8b586abda0"}],"limitations":[{"text":"An embedding model alone is not the same evaluated pipeline as frozen embeddings followed by logistic regression. Tokenization and pooling choices must be preserved.","source_ids":["evidence-official-96b5c3a31a50f7c259d6","evidence-official-25b222d11900e0e88a51","evidence-official-3aa749d3205c0a768b22"],"source_locator":"DNABERT-2 paper Sections 3.2, 4.1 and 5.2 (Further Pre-Training); official model card and implementation README; official DNABERT-2-117M checkpoint LICENSE at revision 7bce263b15377fc15361f52cfab88f8b586abda0"}],"diagram":{"title":"DNABERT-2 workflow","steps":["DNA sequence","BPE tokens","Transformer encoder","Representations","Specified task head"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-96b5c3a31a50f7c259d6","evidence-official-25b222d11900e0e88a51","evidence-official-3aa749d3205c0a768b22"],"source_locator":"DNABERT-2 paper Sections 3.2, 4.1 and 5.2 (Further Pre-Training); official model card and implementation README; official DNABERT-2-117M checkpoint LICENSE at revision 7bce263b15377fc15361f52cfab88f8b586abda0"},"coverage":"limited","gaps":["Training cutoff: The paper identifies the human and multispecies genome corpora but does not state one latest-sequence deposition date in its reviewed pretraining-data sections."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}},"description":"Genomic sequence representation model","facets":{"areas":["genomics"]},"id":"discovery-model-dnabert-2","kind":"model","links":[{"relation":"applicable_to","target_id":"discovery-benchmark-gue"}],"name":"DNABERT-2","source_ids":["src-discovery-magics-lab-dnabert-2"],"status":"discovered"}
{"attributes":{"access":"official_source_linked","benchmark_applicability":"candidate; not evidence of a reported evaluation","candidate_benchmark_ids":["discovery-benchmark-massspecgym"],"entity_level":"family","reported_name":"DreaMS","version":null,"historical_missing_metadata":{"checkpoint":"unextracted","code_licence":"unextracted","parameters":"unextracted","training_cutoff":"unextracted","training_data":"unextracted","version":"unextracted","weights_licence":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"DreaMS learns molecular representations from tandem mass spectra using self-supervised learning.","summary_source_ids":["evidence-official-89de5de6fecdda12060d","evidence-official-3076c52b2fa7cb48ba7b","evidence-official-116050b81d03a38e3b6f"],"summary_source_locator":"Paper: DreaMS neural network architecture and Hyperparameters, ablation studies, implementation details and benchmarking; README.md: models/data links; Zenodo record 10997887 metadata.license and files; DreaMS paper Introduction and GeMS mining methods","sections":[{"title":"How it works","body":"DreaMS learns molecular representations from tandem mass spectra using self-supervised learning. PeakEncoder maps spectral peaks to continuous features; SpectrumEncoder uses transformer blocks; a task-specific PeakDecoder maps the contextual features to predictions. The documented inputs are MS/MS spectra with the required peak and acquisition information. The output consists of spectrum embeddings or predictions from separately fine-tuned spectral tasks.","source_ids":["evidence-official-89de5de6fecdda12060d","evidence-official-3076c52b2fa7cb48ba7b","evidence-official-116050b81d03a38e3b6f"],"source_locator":"Paper: DreaMS neural network architecture and Hyperparameters, ablation studies, implementation details and benchmarking; README.md: models/data links; Zenodo record 10997887 metadata.license and files; DreaMS paper Introduction and GeMS mining methods"},{"title":"Versions and reproducibility","body":"Pretrained model and task-specific fine-tunes distributed separately via the linked Zenodo release. The paper describes retaining 60 spectral peaks for the transformer; this is peak-count preprocessing, not a nucleotide or protein context.","source_ids":["evidence-official-89de5de6fecdda12060d","evidence-official-3076c52b2fa7cb48ba7b","evidence-official-116050b81d03a38e3b6f"],"source_locator":"Paper: DreaMS neural network architecture and Hyperparameters, ablation studies, implementation details and benchmarking; README.md: models/data links; Zenodo record 10997887 metadata.license and files; DreaMS paper Introduction and GeMS mining methods"}],"facts":[{"label":"Model type","value":"Tandem mass-spectral transformer","status":"source_checked","source_ids":["evidence-official-89de5de6fecdda12060d","evidence-official-3076c52b2fa7cb48ba7b","evidence-official-116050b81d03a38e3b6f"],"source_locator":"Paper: DreaMS neural network architecture and Hyperparameters, ablation studies, implementation details and benchmarking; README.md: models/data links; Zenodo record 10997887 metadata.license and files; DreaMS paper Introduction and GeMS mining methods"},{"label":"Architecture","value":"PeakEncoder maps spectral peaks to continuous features; SpectrumEncoder uses transformer blocks; a task-specific PeakDecoder maps the contextual features to predictions.","status":"source_checked","source_ids":["evidence-official-89de5de6fecdda12060d","evidence-official-3076c52b2fa7cb48ba7b","evidence-official-116050b81d03a38e3b6f"],"source_locator":"Paper: DreaMS neural network architecture and Hyperparameters, ablation studies, implementation details and benchmarking; README.md: models/data links; Zenodo record 10997887 metadata.license and files; DreaMS paper Introduction and GeMS mining methods"},{"label":"Inputs","value":"MS/MS spectra with the required peak and acquisition information.","status":"source_checked","source_ids":["evidence-official-89de5de6fecdda12060d","evidence-official-3076c52b2fa7cb48ba7b","evidence-official-116050b81d03a38e3b6f"],"source_locator":"Paper: DreaMS neural network architecture and Hyperparameters, ablation studies, implementation details and benchmarking; README.md: models/data links; Zenodo record 10997887 metadata.license and files; DreaMS paper Introduction and GeMS mining methods"},{"label":"Outputs","value":"Spectrum embeddings or predictions from separately fine-tuned spectral tasks.","status":"source_checked","source_ids":["evidence-official-89de5de6fecdda12060d","evidence-official-3076c52b2fa7cb48ba7b","evidence-official-116050b81d03a38e3b6f"],"source_locator":"Paper: DreaMS neural network architecture and Hyperparameters, ablation studies, implementation details and benchmarking; README.md: models/data links; Zenodo record 10997887 metadata.license and files; DreaMS paper Introduction and GeMS mining methods"},{"label":"Parameters","value":"116M for the complete self-supervised network reported in the DreaMS paper; downstream embedding-only configurations may have fewer parameters.","status":"source_checked","source_ids":["evidence-official-89de5de6fecdda12060d","evidence-official-3076c52b2fa7cb48ba7b","evidence-official-116050b81d03a38e3b6f"],"source_locator":"Paper: DreaMS neural network architecture and Hyperparameters, ablation studies, implementation details and benchmarking; README.md: models/data links; Zenodo record 10997887 metadata.license and files; DreaMS paper Introduction and GeMS mining methods"},{"label":"Known versions","value":"Pretrained model and task-specific fine-tunes distributed separately via the linked Zenodo release.","status":"source_checked","source_ids":["evidence-official-89de5de6fecdda12060d","evidence-official-3076c52b2fa7cb48ba7b","evidence-official-116050b81d03a38e3b6f"],"source_locator":"Paper: DreaMS neural network architecture and Hyperparameters, ablation studies, implementation details and benchmarking; README.md: models/data links; Zenodo record 10997887 metadata.license and files; DreaMS paper Introduction and GeMS mining methods"},{"label":"Training data","value":"GeMS mined from MassIVE/GNPS; the paper identifies GeMS-A10, approximately 24M spectra, as the high-quality pretraining subset.","status":"source_checked","source_ids":["evidence-official-89de5de6fecdda12060d","evidence-official-3076c52b2fa7cb48ba7b","evidence-official-116050b81d03a38e3b6f"],"source_locator":"Paper: DreaMS neural network architecture and Hyperparameters, ablation studies, implementation details and benchmarking; README.md: models/data links; Zenodo record 10997887 metadata.license and files; DreaMS paper Introduction and GeMS mining methods"},{"label":"Training cutoff","value":"GeMS mining selected GNPS-tagged MassIVE studies available as of November 2022; subsequent task-specific datasets have separate provenance.","status":"source_checked","source_ids":["evidence-official-89de5de6fecdda12060d","evidence-official-3076c52b2fa7cb48ba7b","evidence-official-116050b81d03a38e3b6f"],"source_locator":"Paper: DreaMS neural network architecture and Hyperparameters, ablation studies, implementation details and benchmarking; README.md: models/data links; Zenodo record 10997887 metadata.license and files; DreaMS paper Introduction and GeMS mining methods"},{"label":"Context limits","value":"The paper describes retaining 60 spectral peaks for the transformer; this is peak-count preprocessing, not a nucleotide or protein context.","status":"source_checked","source_ids":["evidence-official-89de5de6fecdda12060d","evidence-official-3076c52b2fa7cb48ba7b","evidence-official-116050b81d03a38e3b6f"],"source_locator":"Paper: DreaMS neural network architecture and Hyperparameters, ablation studies, implementation details and benchmarking; README.md: models/data links; Zenodo record 10997887 metadata.license and files; DreaMS paper Introduction and GeMS mining methods"},{"label":"Weights licence","value":"CC-BY-4.0 for the embedding_model.ckpt and ssl_model.ckpt files in author-linked Zenodo record 10997887; separate from the MIT code licence.","status":"source_checked","source_ids":["evidence-official-89de5de6fecdda12060d","evidence-official-3076c52b2fa7cb48ba7b","evidence-official-116050b81d03a38e3b6f"],"source_locator":"Paper: DreaMS neural network architecture and Hyperparameters, ablation studies, implementation details and benchmarking; README.md: models/data links; Zenodo record 10997887 metadata.license and files; DreaMS paper Introduction and GeMS mining methods"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/pluskal-lab/DreaMS","status":"source_checked","source_ids":["evidence-official-89de5de6fecdda12060d","evidence-official-3076c52b2fa7cb48ba7b","evidence-official-116050b81d03a38e3b6f"],"source_locator":"Paper: DreaMS neural network architecture and Hyperparameters, ablation studies, implementation details and benchmarking; README.md: models/data links; Zenodo record 10997887 metadata.license and files; DreaMS paper Introduction and GeMS mining methods"},{"label":"Code licence","value":"MIT","status":"source_checked","source_ids":["evidence-official-f6c94a04c51af984a008"],"source_locator":"LICENSE: licence text"}],"strengths":[{"text":"The same pretrained representation can be adapted to spectral similarity and molecular annotation tasks.","source_ids":["evidence-official-89de5de6fecdda12060d","evidence-official-3076c52b2fa7cb48ba7b","evidence-official-116050b81d03a38e3b6f"],"source_locator":"Paper: DreaMS neural network architecture and Hyperparameters, ablation studies, implementation details and benchmarking; README.md: models/data links; Zenodo record 10997887 metadata.license and files; DreaMS paper Introduction and GeMS mining methods"}],"limitations":[{"text":"Pretrained representations and fine-tuned similarity/property predictors are distinct models. Spectrum quality and preprocessing remain essential parts of an evaluation.","source_ids":["evidence-official-89de5de6fecdda12060d","evidence-official-3076c52b2fa7cb48ba7b","evidence-official-116050b81d03a38e3b6f"],"source_locator":"Paper: DreaMS neural network architecture and Hyperparameters, ablation studies, implementation details and benchmarking; README.md: models/data links; Zenodo record 10997887 metadata.license and files; DreaMS paper Introduction and GeMS mining methods"}],"diagram":{"title":"DreaMS workflow","steps":["MS/MS spectrum","Peak representation","Pretrained transformer","Embedding or fine-tuned prediction"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-89de5de6fecdda12060d","evidence-official-3076c52b2fa7cb48ba7b","evidence-official-116050b81d03a38e3b6f"],"source_locator":"Paper: DreaMS neural network architecture and Hyperparameters, ablation studies, implementation details and benchmarking; README.md: models/data links; Zenodo record 10997887 metadata.license and files; DreaMS paper Introduction and GeMS mining methods"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}},"description":"Tandem mass spectrum representation model","facets":{"areas":["metabolomics"]},"id":"discovery-model-dreams","kind":"model","links":[{"relation":"applicable_to","target_id":"discovery-benchmark-massspecgym"}],"name":"DreaMS","source_ids":["src-discovery-pluskal-lab-dreams"],"status":"discovered"}
{"attributes":{"access":"official_source_linked","benchmark_applicability":"candidate; not evidence of a reported evaluation","candidate_benchmark_ids":["discovery-benchmark-proteingym"],"entity_level":"family","reported_name":"ESM-1v","version":null,"historical_missing_metadata":{"checkpoint":"unextracted","code_licence":"unextracted","parameters":"unextracted","training_cutoff":"unextracted","training_data":"unextracted","version":"unextracted","weights_licence":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"ESM-1v provides protein language models intended for zero-shot variant-effect scoring.","summary_source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-137a3a936ffa4673b3d3"],"summary_source_locator":"README.md: Main models, Zero-shot variant prediction and Pre-trained Models","sections":[{"title":"How it works","body":"ESM-1v provides protein language models intended for zero-shot variant-effect scoring. Transformer architecture shared with ESM-1b, trained on UniRef90; multiple released model instances support the variant-scoring workflow. The documented inputs are protein amino-acid sequence and the specified sequence variation. The output consists of token probabilities used to calculate sequence-variant scores.","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-137a3a936ffa4673b3d3"],"source_locator":"README.md: Main models, Zero-shot variant prediction and Pre-trained Models"},{"title":"Versions and reproducibility","body":"esm1v_t33_650M_UR90S_1 through _5. The applicable input limits require configuration-specific checking.","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-137a3a936ffa4673b3d3"],"source_locator":"README.md: Main models, Zero-shot variant prediction and Pre-trained Models"}],"facts":[{"label":"Model type","value":"Protein transformer for variant-effect scoring","status":"source_checked","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-137a3a936ffa4673b3d3"],"source_locator":"README.md: Main models, Zero-shot variant prediction and Pre-trained Models"},{"label":"Architecture","value":"Transformer architecture shared with ESM-1b, trained on UniRef90; multiple released model instances support the variant-scoring workflow.","status":"source_checked","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-137a3a936ffa4673b3d3"],"source_locator":"README.md: Main models, Zero-shot variant prediction and Pre-trained Models"},{"label":"Inputs","value":"Protein amino-acid sequence and the specified sequence variation.","status":"source_checked","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-137a3a936ffa4673b3d3"],"source_locator":"README.md: Main models, Zero-shot variant prediction and Pre-trained Models"},{"label":"Outputs","value":"Token probabilities used to calculate sequence-variant scores.","status":"source_checked","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-137a3a936ffa4673b3d3"],"source_locator":"README.md: Main models, Zero-shot variant prediction and Pre-trained Models"},{"label":"Parameters","value":"650 million per model; 33 transformer layers.","status":"source_checked","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-137a3a936ffa4673b3d3"],"source_locator":"README.md: Main models, Zero-shot variant prediction and Pre-trained Models"},{"label":"Known versions","value":"esm1v_t33_650M_UR90S_1 through _5.","status":"source_checked","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-137a3a936ffa4673b3d3"],"source_locator":"README.md: Main models, Zero-shot variant prediction and Pre-trained Models"},{"label":"Training data","value":"UniRef90/S 2020_03, as reported in the pretrained-model table.","status":"source_checked","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-137a3a936ffa4673b3d3"],"source_locator":"README.md: Main models, Zero-shot variant prediction and Pre-trained Models"},{"label":"Training cutoff","value":"The official model table identifies the UniRef90/S2020_03 training release; this is a corpus version rather than a verified latest-sequence deposition date.","status":"source_checked","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-137a3a936ffa4673b3d3"],"source_locator":"README.md: Main models, Zero-shot variant prediction and Pre-trained Models"},{"label":"Context limits","value":"The inspected official esm1v_t33_650M_UR90S_1 config reserves 1,026 absolute positions. This includes model positions and is not a claim of 1,026-residue validated biological context.","status":"source_checked","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-137a3a936ffa4673b3d3"],"source_locator":"README.md: Main models, Zero-shot variant prediction and Pre-trained Models"},{"label":"Weights licence","value":"Separate checkpoint-distribution terms are not stated in the inspected release documentation and licence material. The source-code licence alone is not recorded as an explicit weight grant.","status":"unreported","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-137a3a936ffa4673b3d3","evidence-official-2e7c7649620407f50f6b"],"source_locator":"README.md: Main models, Zero-shot variant prediction and Pre-trained Models; LICENSE: licence text"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/facebookresearch/esm","status":"source_checked","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-137a3a936ffa4673b3d3"],"source_locator":"README.md: Main models, Zero-shot variant prediction and Pre-trained Models"},{"label":"Code licence","value":"MIT","status":"source_checked","source_ids":["evidence-official-2e7c7649620407f50f6b"],"source_locator":"LICENSE: licence text"}],"strengths":[{"text":"The package provides a dedicated variant-prediction example without fitting a task-specific supervised classifier.","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-137a3a936ffa4673b3d3"],"source_locator":"README.md: Main models, Zero-shot variant prediction and Pre-trained Models"}],"limitations":[{"text":"The five released instances are distinct models. A language-model score is not a calibrated clinical pathogenicity probability.","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-137a3a936ffa4673b3d3"],"source_locator":"README.md: Main models, Zero-shot variant prediction and Pre-trained Models"}],"diagram":{"title":"ESM-1v workflow","steps":["Reference protein sequence","ESM-1v token predictions","Specified variant scoring rule","Variant score"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-137a3a936ffa4673b3d3"],"source_locator":"README.md: Main models, Zero-shot variant prediction and Pre-trained Models"},"coverage":"limited","gaps":["Weights licence: Separate checkpoint-distribution terms are not stated in the inspected release documentation and licence material. The source-code licence alone is not recorded as an explicit weight grant."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}},"description":"Protein variant effect model family","facets":{"areas":["protein-function"]},"id":"discovery-model-esm-1v","kind":"model","links":[{"relation":"applicable_to","target_id":"discovery-benchmark-proteingym"}],"name":"ESM-1v","source_ids":["src-discovery-facebookresearch-esm"],"status":"discovered"}
{"attributes":{"access":"official_source_linked","benchmark_applicability":"candidate; not evidence of a reported evaluation","candidate_benchmark_ids":["discovery-benchmark-proteingym"],"entity_level":"family","reported_name":"ESM-2","version":null,"historical_missing_metadata":{"checkpoint":"unextracted","code_licence":"unextracted","parameters":"unextracted","training_cutoff":"unextracted","training_data":"unextracted","version":"unextracted","weights_licence":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"ESM-2 is a family of protein sequence encoders that produce representations for downstream protein analyses.","summary_source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-d7e5c63feb0e0e627625","evidence-official-28fc17fbe0c1d99da219"],"summary_source_locator":"ESM README: Pre-trained Models and Main models; official facebook/esm2_t33_650M_UR50D README licence metadata and config.json","sections":[{"title":"How it works","body":"ESM-2 tokenizes an amino-acid sequence and uses a transformer encoder trained to recover masked residues. Self-attention lets each residue representation depend on its sequence context. The released model returns token probabilities and embeddings; a specified pooling rule, task head or complete folding pipeline is needed for a particular biological prediction.","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-d7e5c63feb0e0e627625","evidence-official-28fc17fbe0c1d99da219"],"source_locator":"ESM README: Pre-trained Models and Main models; official facebook/esm2_t33_650M_UR50D README licence metadata and config.json"},{"title":"Versions and reproducibility","body":"ESM-2 checkpoint identifiers encode layer count, parameter scale and training-data tag. The checked esm2_t33_650M_UR50D configuration lists max_position_embeddings=1,026. This configuration field includes model positions and is not a claim of training or validated inference on 1,026 amino acids.","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-d7e5c63feb0e0e627625","evidence-official-28fc17fbe0c1d99da219"],"source_locator":"ESM README: Pre-trained Models and Main models; official facebook/esm2_t33_650M_UR50D README licence metadata and config.json"}],"facts":[{"label":"Model type","value":"Masked-token protein transformer encoder","status":"source_checked","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-d7e5c63feb0e0e627625","evidence-official-28fc17fbe0c1d99da219"],"source_locator":"ESM README: Pre-trained Models and Main models; official facebook/esm2_t33_650M_UR50D README licence metadata and config.json"},{"label":"Architecture","value":"Masked-token protein transformer encoder; the checked 650M checkpoint has 33 layers, hidden width 1,280, 20 attention heads and rotary positional encoding.","status":"source_checked","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-d7e5c63feb0e0e627625","evidence-official-28fc17fbe0c1d99da219"],"source_locator":"ESM README: Pre-trained Models and Main models; official facebook/esm2_t33_650M_UR50D README licence metadata and config.json"},{"label":"Inputs","value":"Single amino-acid sequences.","status":"source_checked","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-d7e5c63feb0e0e627625","evidence-official-28fc17fbe0c1d99da219"],"source_locator":"ESM README: Pre-trained Models and Main models; official facebook/esm2_t33_650M_UR50D README licence metadata and config.json"},{"label":"Outputs","value":"Residue embeddings, sequence representations and masked-token predictions.","status":"source_checked","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-d7e5c63feb0e0e627625","evidence-official-28fc17fbe0c1d99da219"],"source_locator":"ESM README: Pre-trained Models and Main models; official facebook/esm2_t33_650M_UR50D README licence metadata and config.json"},{"label":"Parameters","value":"Released scales: 8M, 35M, 150M, 650M, 3B and 15B.","status":"source_checked","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-d7e5c63feb0e0e627625","evidence-official-28fc17fbe0c1d99da219"],"source_locator":"ESM README: Pre-trained Models and Main models; official facebook/esm2_t33_650M_UR50D README licence metadata and config.json"},{"label":"Known versions","value":"ESM-2 checkpoint identifiers encode layer count, parameter scale and training-data tag.","status":"source_checked","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-d7e5c63feb0e0e627625","evidence-official-28fc17fbe0c1d99da219"],"source_locator":"ESM README: Pre-trained Models and Main models; official facebook/esm2_t33_650M_UR50D README licence metadata and config.json"},{"label":"Training data","value":"UniRef50 clusters with UniRef90 sampling; the pretrained-model table labels UR50/D 2021_04.","status":"source_checked","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-d7e5c63feb0e0e627625","evidence-official-28fc17fbe0c1d99da219"],"source_locator":"ESM README: Pre-trained Models and Main models; official facebook/esm2_t33_650M_UR50D README licence metadata and config.json"},{"label":"Training cutoff","value":"The pretrained-model table identifies training-data release UR50/D 2021_04; a corpus release date is not necessarily a last-deposited-sequence cutoff.","status":"source_checked","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-d7e5c63feb0e0e627625","evidence-official-28fc17fbe0c1d99da219"],"source_locator":"ESM README: Pre-trained Models and Main models; official facebook/esm2_t33_650M_UR50D README licence metadata and config.json"},{"label":"Context limits","value":"The checked esm2_t33_650M_UR50D configuration lists max_position_embeddings=1,026. This configuration field includes model positions and is not a claim of training or validated inference on 1,026 amino acids.","status":"source_checked","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-d7e5c63feb0e0e627625","evidence-official-28fc17fbe0c1d99da219"],"source_locator":"ESM README: Pre-trained Models and Main models; official facebook/esm2_t33_650M_UR50D README licence metadata and config.json"},{"label":"Weights licence","value":"The official facebook/esm2_t33_650M_UR50D model card declares MIT; this is the inspected checkpoint, not a licence inference from source code.","status":"source_checked","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-d7e5c63feb0e0e627625","evidence-official-28fc17fbe0c1d99da219"],"source_locator":"ESM README: Pre-trained Models and Main models; official facebook/esm2_t33_650M_UR50D README licence metadata and config.json"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/facebookresearch/esm","status":"source_checked","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-d7e5c63feb0e0e627625","evidence-official-28fc17fbe0c1d99da219"],"source_locator":"ESM README: Pre-trained Models and Main models; official facebook/esm2_t33_650M_UR50D README licence metadata and config.json"},{"label":"Code licence","value":"MIT","status":"source_checked","source_ids":["evidence-official-2e7c7649620407f50f6b"],"source_locator":"LICENSE: licence text"}],"strengths":[{"text":"Released checkpoints span several sizes and can be used without constructing a multiple sequence alignment.","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-d7e5c63feb0e0e627625","evidence-official-28fc17fbe0c1d99da219"],"source_locator":"ESM README: Pre-trained Models and Main models; official facebook/esm2_t33_650M_UR50D README licence metadata and config.json"}],"limitations":[{"text":"A general embedding is not a directly measured function or structure. Fine-tuning, pooling and downstream heads remain part of each evaluated configuration.","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-d7e5c63feb0e0e627625","evidence-official-28fc17fbe0c1d99da219"],"source_locator":"ESM README: Pre-trained Models and Main models; official facebook/esm2_t33_650M_UR50D README licence metadata and config.json"}],"diagram":{"title":"ESM-2 workflow","steps":["Protein sequence","Transformer layers","Residue embeddings","Specified downstream analysis"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-d7e5c63feb0e0e627625","evidence-official-28fc17fbe0c1d99da219"],"source_locator":"ESM README: Pre-trained Models and Main models; official facebook/esm2_t33_650M_UR50D README licence metadata and config.json"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}},"description":"Protein sequence representation family","facets":{"areas":["protein-function"]},"id":"discovery-model-esm-2","kind":"model","links":[{"relation":"applicable_to","target_id":"discovery-benchmark-proteingym"}],"name":"ESM-2","source_ids":["src-discovery-facebookresearch-esm"],"status":"discovered"}
{"attributes":{"access":"official_source_linked","benchmark_applicability":"candidate; not evidence of a reported evaluation","candidate_benchmark_ids":["discovery-benchmark-proteinbench"],"entity_level":"family","reported_name":"ESM-IF1","version":null,"historical_missing_metadata":{"checkpoint":"unextracted","code_licence":"unextracted","parameters":"unextracted","training_cutoff":"unextracted","training_data":"unextracted","version":"unextracted","weights_licence":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"ESM-IF1 designs protein sequences conditioned on backbone coordinates.","summary_source_ids":["evidence-official-29d4b229a3aa426a6dfb"],"summary_source_locator":"README.md: Inverse folding and Pre-trained Models","sections":[{"title":"How it works","body":"ESM-IF1 designs protein sequences conditioned on backbone coordinates. Geometric-vector-perceptron input processing followed by a sequence-to-sequence transformer. The documented inputs are protein backbone atom coordinates; the model supports missing backbone spans. The output consists of sampled protein sequences or conditional sequence likelihoods.","source_ids":["evidence-official-29d4b229a3aa426a6dfb"],"source_locator":"README.md: Inverse folding and Pre-trained Models"},{"title":"Versions and reproducibility","body":"esm_if1_gvp4_t16_142M_UR50. The applicable input limits require configuration-specific checking.","source_ids":["evidence-official-29d4b229a3aa426a6dfb"],"source_locator":"README.md: Inverse folding and Pre-trained Models"}],"facts":[{"label":"Model type","value":"Geometric encoder and inverse-folding transformer","status":"source_checked","source_ids":["evidence-official-29d4b229a3aa426a6dfb"],"source_locator":"README.md: Inverse folding and Pre-trained Models"},{"label":"Architecture","value":"Geometric-vector-perceptron input processing followed by a sequence-to-sequence transformer.","status":"source_checked","source_ids":["evidence-official-29d4b229a3aa426a6dfb"],"source_locator":"README.md: Inverse folding and Pre-trained Models"},{"label":"Inputs","value":"Protein backbone atom coordinates; the model supports missing backbone spans.","status":"source_checked","source_ids":["evidence-official-29d4b229a3aa426a6dfb"],"source_locator":"README.md: Inverse folding and Pre-trained Models"},{"label":"Outputs","value":"Sampled protein sequences or conditional sequence likelihoods.","status":"source_checked","source_ids":["evidence-official-29d4b229a3aa426a6dfb"],"source_locator":"README.md: Inverse folding and Pre-trained Models"},{"label":"Parameters","value":"The official checkpoint identifier contains 142M but the same repository model table states 124M. Both values are retained as a source discrepancy, without choosing a total.","status":"source_checked","source_ids":["evidence-official-29d4b229a3aa426a6dfb"],"source_locator":"README.md: Inverse folding and Pre-trained Models"},{"label":"Known versions","value":"esm_if1_gvp4_t16_142M_UR50.","status":"source_checked","source_ids":["evidence-official-29d4b229a3aa426a6dfb"],"source_locator":"README.md: Inverse folding and Pre-trained Models"},{"label":"Training data","value":"CATH 4.3 and predicted UniRef50 structures; README reports 12M structures predicted by AlphaFold2.","status":"source_checked","source_ids":["evidence-official-29d4b229a3aa426a6dfb"],"source_locator":"README.md: Inverse folding and Pre-trained Models"},{"label":"Training cutoff","value":"The official model table identifies CATH 4.3 and predicted UniRef50 structures, but does not state a common latest-structure or sequence date.","status":"unreported","source_ids":["evidence-official-29d4b229a3aa426a6dfb"],"source_locator":"README.md: Inverse folding and Pre-trained Models"},{"label":"Context limits","value":"The reviewed inverse-folding usage and model table do not establish a universal maximum backbone length; the structural graph and selected inference configuration determine resource use.","status":"unreported","source_ids":["evidence-official-29d4b229a3aa426a6dfb"],"source_locator":"README.md: Inverse folding and Pre-trained Models"},{"label":"Weights licence","value":"Separate checkpoint-distribution terms are not stated in the inspected release documentation and licence material. The source-code licence alone is not recorded as an explicit weight grant.","status":"unreported","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-2e7c7649620407f50f6b"],"source_locator":"README.md: Inverse folding and Pre-trained Models; LICENSE: licence text"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/facebookresearch/esm","status":"source_checked","source_ids":["evidence-official-29d4b229a3aa426a6dfb"],"source_locator":"README.md: Inverse folding and Pre-trained Models"},{"label":"Code licence","value":"MIT","status":"source_checked","source_ids":["evidence-official-2e7c7649620407f50f6b"],"source_locator":"LICENSE: licence text"}],"strengths":[{"text":"Structure-conditioned design can use partially missing backbones because training included span masking.","source_ids":["evidence-official-29d4b229a3aa426a6dfb"],"source_locator":"README.md: Inverse folding and Pre-trained Models"}],"limitations":[{"text":"The public checkpoint name contains 142M but the README parameter table reports 124M. This discrepancy is retained rather than silently selecting a total.","source_ids":["evidence-official-29d4b229a3aa426a6dfb"],"source_locator":"README.md: Inverse folding and Pre-trained Models"}],"diagram":{"title":"ESM-IF1 workflow","steps":["Backbone coordinates","Geometric processing","Sequence transformer","Designed sequence or likelihood"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-29d4b229a3aa426a6dfb"],"source_locator":"README.md: Inverse folding and Pre-trained Models"},"coverage":"limited","gaps":["Training cutoff: The official model table identifies CATH 4.3 and predicted UniRef50 structures, but does not state a common latest-structure or sequence date.","Context limits: The reviewed inverse-folding usage and model table do not establish a universal maximum backbone length; the structural graph and selected inference configuration determine resource use.","Weights licence: Separate checkpoint-distribution terms are not stated in the inspected release documentation and licence material. The source-code licence alone is not recorded as an explicit weight grant."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}},"description":"Protein inverse folding model","facets":{"areas":["protein-structure"]},"id":"discovery-model-esm-if1","kind":"model","links":[{"relation":"applicable_to","target_id":"discovery-benchmark-proteinbench"}],"name":"ESM-IF1","source_ids":["src-discovery-facebookresearch-esm"],"status":"discovered"}
{"attributes":{"access":"official_source_linked","benchmark_applicability":"candidate; not evidence of a reported evaluation","candidate_benchmark_ids":["discovery-benchmark-proteinbench"],"entity_level":"family","reported_name":"ESM3","version":null,"historical_missing_metadata":{"checkpoint":"unextracted","code_licence":"unextracted","parameters":"unextracted","training_cutoff":"unextracted","training_data":"unextracted","version":"unextracted","weights_licence":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"ESM3 generates and completes protein sequence, structure and functional annotations using a shared multimodal representation.","summary_source_ids":["evidence-official-4bf005200ebfb2586b78","evidence-official-0437be6150a3da9f8efc","evidence-official-1f027f8ecc53decdeec8"],"summary_source_locator":"_assets/ESM3_README.md: overview, ESM3 Family, Running Locally and Licenses","sections":[{"title":"How it works","body":"ESM3 generates and completes protein sequence, structure and functional annotations using a shared multimodal representation. Transformer generative masked-language model with discrete sequence, structure and function tracks; generation iteratively fills masked positions. The documented inputs are complete or partial protein sequence, structure and function-keyword tracks. The output consists of completed sequence, structure and functional tracks.","source_ids":["evidence-official-4bf005200ebfb2586b78","evidence-official-0437be6150a3da9f8efc","evidence-official-1f027f8ecc53decdeec8"],"source_locator":"_assets/ESM3_README.md: overview, ESM3 Family, Running Locally and Licenses"},{"title":"Versions and reproducibility","body":"Published March 2024 small/medium/large models, August 2024 small/medium models, and esm3-sm-open-v1 local weights. The applicable input limits require configuration-specific checking.","source_ids":["evidence-official-4bf005200ebfb2586b78","evidence-official-0437be6150a3da9f8efc","evidence-official-1f027f8ecc53decdeec8"],"source_locator":"_assets/ESM3_README.md: overview, ESM3 Family, Running Locally and Licenses"}],"facts":[{"label":"Model type","value":"Multitrack generative protein transformer","status":"source_checked","source_ids":["evidence-official-4bf005200ebfb2586b78","evidence-official-0437be6150a3da9f8efc","evidence-official-1f027f8ecc53decdeec8"],"source_locator":"_assets/ESM3_README.md: overview, ESM3 Family, Running Locally and Licenses"},{"label":"Architecture","value":"Transformer generative masked-language model with discrete sequence, structure and function tracks; generation iteratively fills masked positions.","status":"source_checked","source_ids":["evidence-official-4bf005200ebfb2586b78","evidence-official-0437be6150a3da9f8efc","evidence-official-1f027f8ecc53decdeec8"],"source_locator":"_assets/ESM3_README.md: overview, ESM3 Family, Running Locally and Licenses"},{"label":"Inputs","value":"Complete or partial protein sequence, structure and function-keyword tracks.","status":"source_checked","source_ids":["evidence-official-4bf005200ebfb2586b78","evidence-official-0437be6150a3da9f8efc","evidence-official-1f027f8ecc53decdeec8"],"source_locator":"_assets/ESM3_README.md: overview, ESM3 Family, Running Locally and Licenses"},{"label":"Outputs","value":"Completed sequence, structure and functional tracks.","status":"source_checked","source_ids":["evidence-official-4bf005200ebfb2586b78","evidence-official-0437be6150a3da9f8efc","evidence-official-1f027f8ecc53decdeec8"],"source_locator":"_assets/ESM3_README.md: overview, ESM3 Family, Running Locally and Licenses"},{"label":"Parameters","value":"1.4B small, 7B medium and 98B large variants.","status":"source_checked","source_ids":["evidence-official-4bf005200ebfb2586b78","evidence-official-0437be6150a3da9f8efc","evidence-official-1f027f8ecc53decdeec8"],"source_locator":"_assets/ESM3_README.md: overview, ESM3 Family, Running Locally and Licenses"},{"label":"Known versions","value":"Published March 2024 small/medium/large models, August 2024 small/medium models, and esm3-sm-open-v1 local weights.","status":"source_checked","source_ids":["evidence-official-4bf005200ebfb2586b78","evidence-official-0437be6150a3da9f8efc","evidence-official-1f027f8ecc53decdeec8"],"source_locator":"_assets/ESM3_README.md: overview, ESM3 Family, Running Locally and Licenses"},{"label":"Training data","value":"ESM3 overview reports 2.78 billion proteins and 771 billion unique tokens for the largest model; this is not a verified per-checkpoint manifest.","status":"source_checked","source_ids":["evidence-official-4bf005200ebfb2586b78","evidence-official-0437be6150a3da9f8efc","evidence-official-1f027f8ecc53decdeec8"],"source_locator":"_assets/ESM3_README.md: overview, ESM3 Family, Running Locally and Licenses"},{"label":"Training cutoff","value":"The inspected ESM3 release documentation does not supply one latest-data date shared across its sequence, structure and function tracks.","status":"unreported","source_ids":["evidence-official-4bf005200ebfb2586b78","evidence-official-0437be6150a3da9f8efc","evidence-official-1f027f8ecc53decdeec8"],"source_locator":"_assets/ESM3_README.md: overview, ESM3 Family, Running Locally and Licenses"},{"label":"Context limits","value":"The inspected ESM3 family documentation does not establish one context limit for local small, hosted medium and hosted large models; a specific service/checkpoint is required.","status":"unreported","source_ids":["evidence-official-4bf005200ebfb2586b78","evidence-official-0437be6150a3da9f8efc","evidence-official-1f027f8ecc53decdeec8"],"source_locator":"_assets/ESM3_README.md: overview, ESM3 Family, Running Locally and Licenses"},{"label":"Weights licence","value":"MIT stated in the inspected ESM3 README; model and API access conditions remain separate.","status":"source_checked","source_ids":["evidence-official-4bf005200ebfb2586b78","evidence-official-0437be6150a3da9f8efc","evidence-official-1f027f8ecc53decdeec8"],"source_locator":"_assets/ESM3_README.md: overview, ESM3 Family, Running Locally and Licenses"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/evolutionaryscale/esm","status":"source_checked","source_ids":["evidence-official-4bf005200ebfb2586b78","evidence-official-0437be6150a3da9f8efc","evidence-official-1f027f8ecc53decdeec8"],"source_locator":"_assets/ESM3_README.md: overview, ESM3 Family, Running Locally and Licenses"},{"label":"Code licence","value":"MIT","status":"source_checked","source_ids":["evidence-official-2f4711ecd64b0162e85b"],"source_locator":"LICENSE.md: licence text"}],"strengths":[{"text":"Partial prompts can constrain more than one biological property in the same generation workflow.","source_ids":["evidence-official-4bf005200ebfb2586b78","evidence-official-0437be6150a3da9f8efc","evidence-official-1f027f8ecc53decdeec8"],"source_locator":"_assets/ESM3_README.md: overview, ESM3 Family, Running Locally and Licenses"}],"limitations":[{"text":"Open local weights and API model names refer to different released sizes and dates; results require the actual configuration, not only the ESM3 family name.","source_ids":["evidence-official-4bf005200ebfb2586b78","evidence-official-0437be6150a3da9f8efc","evidence-official-1f027f8ecc53decdeec8"],"source_locator":"_assets/ESM3_README.md: overview, ESM3 Family, Running Locally and Licenses"}],"diagram":{"title":"ESM3 workflow","steps":["Partial protein tracks","Multitrack transformer","Iterative unmasking","Completed sequence or structure"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-4bf005200ebfb2586b78","evidence-official-0437be6150a3da9f8efc","evidence-official-1f027f8ecc53decdeec8"],"source_locator":"_assets/ESM3_README.md: overview, ESM3 Family, Running Locally and Licenses"},"coverage":"limited","gaps":["Training cutoff: The inspected ESM3 release documentation does not supply one latest-data date shared across its sequence, structure and function tracks.","Context limits: The inspected ESM3 family documentation does not establish one context limit for local small, hosted medium and hosted large models; a specific service/checkpoint is required."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}},"description":"Multimodal protein model family","facets":{"areas":["protein-structure"]},"id":"discovery-model-esm3","kind":"model","links":[{"relation":"applicable_to","target_id":"discovery-benchmark-proteinbench"}],"name":"ESM3","source_ids":["src-discovery-evolutionaryscale-esm"],"status":"discovered"}
{"attributes":{"access":"official_source_linked","benchmark_applicability":"candidate; not evidence of a reported evaluation","candidate_benchmark_ids":[],"entity_level":"family","reported_name":"ESMC","version":null,"historical_missing_metadata":{"checkpoint":"unextracted","code_licence":"unextracted","parameters":"unextracted","training_cutoff":"unextracted","training_data":"unextracted","version":"unextracted","weights_licence":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"ESM C learns protein sequence representations for downstream analysis.","summary_source_ids":["evidence-official-4bf005200ebfb2586b78","evidence-official-0437be6150a3da9f8efc","evidence-official-1f027f8ecc53decdeec8","evidence-official-1d81a48c6dcc0e6dc14d","evidence-official-e69877a2a54345e162b3"],"summary_source_locator":"Official biohub/ESMC-6B card: Model Architecture, Parameters, Training Data, Training Procedure and Limitations; THIRD_PARTY_NOTICE.md","sections":[{"title":"How it works","body":"ESM C learns protein sequence representations for downstream analysis. Pre-normalized transformer with rotary embeddings, SwiGLU feed-forward activations and no linear/layer-norm biases. The documented inputs are protein amino-acid sequences. The output consists of final-layer or all-layer protein representations and masked-token outputs.","source_ids":["evidence-official-4bf005200ebfb2586b78","evidence-official-0437be6150a3da9f8efc","evidence-official-1f027f8ecc53decdeec8","evidence-official-1d81a48c6dcc0e6dc14d","evidence-official-e69877a2a54345e162b3"],"source_locator":"Official biohub/ESMC-6B card: Model Architecture, Parameters, Training Data, Training Procedure and Limitations; THIRD_PARTY_NOTICE.md"},{"title":"Versions and reproducibility","body":"esmc-600m-2024-12 API identifier and biohub/ESMC-6B local example. The card specifies a2,048-token window after an initial 512-token training phase.","source_ids":["evidence-official-4bf005200ebfb2586b78","evidence-official-0437be6150a3da9f8efc","evidence-official-1f027f8ecc53decdeec8","evidence-official-1d81a48c6dcc0e6dc14d","evidence-official-e69877a2a54345e162b3"],"source_locator":"Official biohub/ESMC-6B card: Model Architecture, Parameters, Training Data, Training Procedure and Limitations; THIRD_PARTY_NOTICE.md"}],"facts":[{"label":"Model type","value":"Protein representation transformer","status":"source_checked","source_ids":["evidence-official-4bf005200ebfb2586b78","evidence-official-0437be6150a3da9f8efc","evidence-official-1f027f8ecc53decdeec8","evidence-official-1d81a48c6dcc0e6dc14d","evidence-official-e69877a2a54345e162b3"],"source_locator":"Official biohub/ESMC-6B card: Model Architecture, Parameters, Training Data, Training Procedure and Limitations; THIRD_PARTY_NOTICE.md"},{"label":"Architecture","value":"Pre-normalized transformer with rotary embeddings, SwiGLU feed-forward activations and no linear/layer-norm biases.","status":"source_checked","source_ids":["evidence-official-4bf005200ebfb2586b78","evidence-official-0437be6150a3da9f8efc","evidence-official-1f027f8ecc53decdeec8","evidence-official-1d81a48c6dcc0e6dc14d","evidence-official-e69877a2a54345e162b3"],"source_locator":"Official biohub/ESMC-6B card: Model Architecture, Parameters, Training Data, Training Procedure and Limitations; THIRD_PARTY_NOTICE.md"},{"label":"Inputs","value":"Protein amino-acid sequences.","status":"source_checked","source_ids":["evidence-official-4bf005200ebfb2586b78","evidence-official-0437be6150a3da9f8efc","evidence-official-1f027f8ecc53decdeec8","evidence-official-1d81a48c6dcc0e6dc14d","evidence-official-e69877a2a54345e162b3"],"source_locator":"Official biohub/ESMC-6B card: Model Architecture, Parameters, Training Data, Training Procedure and Limitations; THIRD_PARTY_NOTICE.md"},{"label":"Outputs","value":"Final-layer or all-layer protein representations and masked-token outputs.","status":"source_checked","source_ids":["evidence-official-4bf005200ebfb2586b78","evidence-official-0437be6150a3da9f8efc","evidence-official-1f027f8ecc53decdeec8","evidence-official-1d81a48c6dcc0e6dc14d","evidence-official-e69877a2a54345e162b3"],"source_locator":"Official biohub/ESMC-6B card: Model Architecture, Parameters, Training Data, Training Procedure and Limitations; THIRD_PARTY_NOTICE.md"},{"label":"Parameters","value":"300M/30 layers, 600M/36 layers and 6B/80 layers in the checked model card.","status":"source_checked","source_ids":["evidence-official-4bf005200ebfb2586b78","evidence-official-0437be6150a3da9f8efc","evidence-official-1f027f8ecc53decdeec8","evidence-official-1d81a48c6dcc0e6dc14d","evidence-official-e69877a2a54345e162b3"],"source_locator":"Official biohub/ESMC-6B card: Model Architecture, Parameters, Training Data, Training Procedure and Limitations; THIRD_PARTY_NOTICE.md"},{"label":"Known versions","value":"esmc-600m-2024-12 API identifier and biohub/ESMC-6B local example.","status":"source_checked","source_ids":["evidence-official-4bf005200ebfb2586b78","evidence-official-0437be6150a3da9f8efc","evidence-official-1f027f8ecc53decdeec8","evidence-official-1d81a48c6dcc0e6dc14d","evidence-official-e69877a2a54345e162b3"],"source_locator":"Official biohub/ESMC-6B card: Model Architecture, Parameters, Training Data, Training Procedure and Limitations; THIRD_PARTY_NOTICE.md"},{"label":"Training data","value":"UniRef, MGnify and JGI protein sequences clustered at 70% identity. The card distinguishes83M, 372M and 2B clusters respectively from the number of repeated training tokens.","status":"source_checked","source_ids":["evidence-official-4bf005200ebfb2586b78","evidence-official-0437be6150a3da9f8efc","evidence-official-1f027f8ecc53decdeec8","evidence-official-1d81a48c6dcc0e6dc14d","evidence-official-e69877a2a54345e162b3"],"source_locator":"Official biohub/ESMC-6B card: Model Architecture, Parameters, Training Data, Training Procedure and Limitations; THIRD_PARTY_NOTICE.md"},{"label":"Training cutoff","value":"The official ESMC-6B card identifies UniRef, MGnify and JGI clusters and training stages, but does not give a latest-sequence date shared across those corpora.","status":"unreported","source_ids":["evidence-official-4bf005200ebfb2586b78","evidence-official-0437be6150a3da9f8efc","evidence-official-1f027f8ecc53decdeec8","evidence-official-1d81a48c6dcc0e6dc14d","evidence-official-e69877a2a54345e162b3"],"source_locator":"Official biohub/ESMC-6B card: Model Architecture, Parameters, Training Data, Training Procedure and Limitations; THIRD_PARTY_NOTICE.md"},{"label":"Context limits","value":"The card specifies a2,048-token window after an initial 512-token training phase.","status":"source_checked","source_ids":["evidence-official-4bf005200ebfb2586b78","evidence-official-0437be6150a3da9f8efc","evidence-official-1f027f8ecc53decdeec8","evidence-official-1d81a48c6dcc0e6dc14d","evidence-official-e69877a2a54345e162b3"],"source_locator":"Official biohub/ESMC-6B card: Model Architecture, Parameters, Training Data, Training Procedure and Limitations; THIRD_PARTY_NOTICE.md"},{"label":"Weights licence","value":"MIT is declared alongside third-party notices in the checked6B card; THIRD_PARTY_NOTICE.md lists dependency licences.","status":"source_checked","source_ids":["evidence-official-4bf005200ebfb2586b78","evidence-official-0437be6150a3da9f8efc","evidence-official-1f027f8ecc53decdeec8","evidence-official-1d81a48c6dcc0e6dc14d","evidence-official-e69877a2a54345e162b3"],"source_locator":"Official biohub/ESMC-6B card: Model Architecture, Parameters, Training Data, Training Procedure and Limitations; THIRD_PARTY_NOTICE.md"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/evolutionaryscale/esm","status":"source_checked","source_ids":["evidence-official-4bf005200ebfb2586b78","evidence-official-0437be6150a3da9f8efc","evidence-official-1f027f8ecc53decdeec8","evidence-official-1d81a48c6dcc0e6dc14d","evidence-official-e69877a2a54345e162b3"],"source_locator":"Official biohub/ESMC-6B card: Model Architecture, Parameters, Training Data, Training Procedure and Limitations; THIRD_PARTY_NOTICE.md"},{"label":"Code licence","value":"MIT","status":"source_checked","source_ids":["evidence-official-2f4711ecd64b0162e85b"],"source_locator":"LICENSE.md: licence text"}],"strengths":[{"text":"The implementation supports local inference and a hosted interface, with downloadable model variants.","source_ids":["evidence-official-4bf005200ebfb2586b78","evidence-official-0437be6150a3da9f8efc","evidence-official-1f027f8ecc53decdeec8","evidence-official-1d81a48c6dcc0e6dc14d","evidence-official-e69877a2a54345e162b3"],"source_locator":"Official biohub/ESMC-6B card: Model Architecture, Parameters, Training Data, Training Procedure and Limitations; THIRD_PARTY_NOTICE.md"}],"limitations":[{"text":"Hosted API names and downloadable weights identify different versions. The main repository does not by itself provide a complete per-checkpoint training manifest.","source_ids":["evidence-official-4bf005200ebfb2586b78","evidence-official-0437be6150a3da9f8efc","evidence-official-1f027f8ecc53decdeec8","evidence-official-1d81a48c6dcc0e6dc14d","evidence-official-e69877a2a54345e162b3"],"source_locator":"Official biohub/ESMC-6B card: Model Architecture, Parameters, Training Data, Training Procedure and Limitations; THIRD_PARTY_NOTICE.md"}],"diagram":{"title":"ESMC workflow","steps":["Protein sequence","ESM C transformer","Layer representations","Downstream analysis"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-4bf005200ebfb2586b78","evidence-official-0437be6150a3da9f8efc","evidence-official-1f027f8ecc53decdeec8","evidence-official-1d81a48c6dcc0e6dc14d","evidence-official-e69877a2a54345e162b3"],"source_locator":"Official biohub/ESMC-6B card: Model Architecture, Parameters, Training Data, Training Procedure and Limitations; THIRD_PARTY_NOTICE.md"},"coverage":"limited","gaps":["Training cutoff: The official ESMC-6B card identifies UniRef, MGnify and JGI clusters and training stages, but does not give a latest-sequence date shared across those corpora."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}},"description":"Protein representation family","facets":{"areas":["protein-function"]},"id":"discovery-model-esmc","kind":"model","links":[],"name":"ESMC","source_ids":["src-discovery-evolutionaryscale-esm"],"status":"discovered"}
{"attributes":{"access":"official_source_linked","benchmark_applicability":"candidate; not evidence of a reported evaluation","candidate_benchmark_ids":["discovery-benchmark-proteinbench"],"entity_level":"family","reported_name":"ESMFold","version":null,"historical_missing_metadata":{"checkpoint":"unextracted","code_licence":"unextracted","parameters":"unextracted","training_cutoff":"unextracted","training_data":"unextracted","version":"unextracted","weights_licence":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"ESMFold predicts protein structures directly from amino-acid sequence using ESM-2 representations.","summary_source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-fa53ab621979fb5f37c8","evidence-official-867f2d1dcb4084248daa"],"summary_source_locator":"ESM README: ESMFold Structure Prediction; official facebook/esmfold_v1 model card and config.json esmfold_config","sections":[{"title":"How it works","body":"ESMFold predicts protein structures directly from amino-acid sequence using ESM-2 representations. ESM-2 sequence representations feed a folding trunk and structure module. The checked v1 configuration has 48 trunk blocks, eight structure-module blocks and up to four recycles. The documented inputs are protein amino-acid sequence; the ESMFold interface also accepts chains separated by a colon. The output consists of predicted PDB structure and confidence values.","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-fa53ab621979fb5f37c8","evidence-official-867f2d1dcb4084248daa"],"source_locator":"ESM README: ESMFold Structure Prediction; official facebook/esmfold_v1 model card and config.json esmfold_config"},{"title":"Versions and reproducibility","body":"esmfold_v0 and esmfold_v1; v1 is the repository recommendation. Inference length is constrained by memory; the repository documents chunking and CPU offload. Backbone position settings do not alone establish the full folding pipeline limit.","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-fa53ab621979fb5f37c8","evidence-official-867f2d1dcb4084248daa"],"source_locator":"ESM README: ESMFold Structure Prediction; official facebook/esmfold_v1 model card and config.json esmfold_config"}],"facts":[{"label":"Model type","value":"Sequence-to-structure protein prediction pipeline","status":"source_checked","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-fa53ab621979fb5f37c8","evidence-official-867f2d1dcb4084248daa"],"source_locator":"ESM README: ESMFold Structure Prediction; official facebook/esmfold_v1 model card and config.json esmfold_config"},{"label":"Architecture","value":"ESM-2 sequence representations feed a folding trunk and structure module. The checked v1 configuration has 48 trunk blocks, eight structure-module blocks and up to four recycles.","status":"source_checked","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-fa53ab621979fb5f37c8","evidence-official-867f2d1dcb4084248daa"],"source_locator":"ESM README: ESMFold Structure Prediction; official facebook/esmfold_v1 model card and config.json esmfold_config"},{"label":"Inputs","value":"Protein amino-acid sequence; the ESMFold interface also accepts chains separated by a colon.","status":"source_checked","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-fa53ab621979fb5f37c8","evidence-official-867f2d1dcb4084248daa"],"source_locator":"ESM README: ESMFold Structure Prediction; official facebook/esmfold_v1 model card and config.json esmfold_config"},{"label":"Outputs","value":"Predicted PDB structure and confidence values.","status":"source_checked","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-fa53ab621979fb5f37c8","evidence-official-867f2d1dcb4084248daa"],"source_locator":"ESM README: ESMFold Structure Prediction; official facebook/esmfold_v1 model card and config.json esmfold_config"},{"label":"Parameters","value":"The released v1 configuration identifies an ESM-2 3B backbone plus a folding trunk and structure module. The 3B figure is not the total size of the complete predictor.","status":"source_checked","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-fa53ab621979fb5f37c8","evidence-official-867f2d1dcb4084248daa"],"source_locator":"ESM README: ESMFold Structure Prediction; official facebook/esmfold_v1 model card and config.json esmfold_config"},{"label":"Known versions","value":"esmfold_v0 and esmfold_v1; v1 is the repository recommendation.","status":"source_checked","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-fa53ab621979fb5f37c8","evidence-official-867f2d1dcb4084248daa"],"source_locator":"ESM README: ESMFold Structure Prediction; official facebook/esmfold_v1 model card and config.json esmfold_config"},{"label":"Training data","value":"PDB and UniRef50 are listed for ESMFold. Full structural training-cutoff verification remains outstanding.","status":"source_checked","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-fa53ab621979fb5f37c8","evidence-official-867f2d1dcb4084248daa"],"source_locator":"ESM README: ESMFold Structure Prediction; official facebook/esmfold_v1 model card and config.json esmfold_config"},{"label":"Training cutoff","value":"PDB and UniRef50 are identified in the official model table; the inspected ESMFold-v1 card and configuration do not supply a shared latest-data date for both components.","status":"unreported","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-fa53ab621979fb5f37c8","evidence-official-867f2d1dcb4084248daa"],"source_locator":"ESM README: ESMFold Structure Prediction; official facebook/esmfold_v1 model card and config.json esmfold_config"},{"label":"Context limits","value":"Inference length is constrained by memory; the repository documents chunking and CPU offload. Backbone position settings do not alone establish the full folding pipeline limit.","status":"source_checked","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-fa53ab621979fb5f37c8","evidence-official-867f2d1dcb4084248daa"],"source_locator":"ESM README: ESMFold Structure Prediction; official facebook/esmfold_v1 model card and config.json esmfold_config"},{"label":"Weights licence","value":"MIT declared by the official facebook/esmfold_v1 model card.","status":"source_checked","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-fa53ab621979fb5f37c8","evidence-official-867f2d1dcb4084248daa"],"source_locator":"ESM README: ESMFold Structure Prediction; official facebook/esmfold_v1 model card and config.json esmfold_config"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/facebookresearch/esm","status":"source_checked","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-fa53ab621979fb5f37c8","evidence-official-867f2d1dcb4084248daa"],"source_locator":"ESM README: ESMFold Structure Prediction; official facebook/esmfold_v1 model card and config.json esmfold_config"},{"label":"Code licence","value":"MIT","status":"source_checked","source_ids":["evidence-official-2e7c7649620407f50f6b"],"source_locator":"LICENSE: licence text"}],"strengths":[{"text":"Supports sequence-only structure prediction without an MSA search.","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-fa53ab621979fb5f37c8","evidence-official-867f2d1dcb4084248daa"],"source_locator":"ESM README: ESMFold Structure Prediction; official facebook/esmfold_v1 model card and config.json esmfold_config"}],"limitations":[{"text":"ESMFold v0 and v1 are different releases. The repository discourages using structure-module-only ablation models as the standard predictor.","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-fa53ab621979fb5f37c8","evidence-official-867f2d1dcb4084248daa"],"source_locator":"ESM README: ESMFold Structure Prediction; official facebook/esmfold_v1 model card and config.json esmfold_config"}],"diagram":{"title":"ESMFold workflow","steps":["Protein sequence","ESM-2 representations","Folding module","Predicted structure and confidence"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-29d4b229a3aa426a6dfb","evidence-official-fa53ab621979fb5f37c8","evidence-official-867f2d1dcb4084248daa"],"source_locator":"ESM README: ESMFold Structure Prediction; official facebook/esmfold_v1 model card and config.json esmfold_config"},"coverage":"limited","gaps":["Training cutoff: PDB and UniRef50 are identified in the official model table; the inspected ESMFold-v1 card and configuration do not supply a shared latest-data date for both components."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}},"description":"Protein structure prediction model","facets":{"areas":["protein-structure"]},"id":"discovery-model-esmfold","kind":"model","links":[{"relation":"applicable_to","target_id":"discovery-benchmark-proteinbench"}],"name":"ESMFold","source_ids":["src-discovery-facebookresearch-esm"],"status":"discovered"}
{"attributes":{"access":"official_source_linked","benchmark_applicability":"candidate; not evidence of a reported evaluation","candidate_benchmark_ids":[],"entity_level":"family","reported_name":"ESMFold2","version":null,"historical_missing_metadata":{"checkpoint":"unextracted","code_licence":"unextracted","parameters":"unextracted","training_cutoff":"unextracted","training_data":"unextracted","version":"unextracted","weights_licence":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"ESMFold2 predicts biomolecular structures from protein, DNA, RNA and ligand inputs, optionally using protein alignments.","summary_source_ids":["evidence-official-4bf005200ebfb2586b78","evidence-official-0437be6150a3da9f8efc","evidence-official-1f027f8ecc53decdeec8","evidence-official-d65bf8ba587b6841e0d9","evidence-official-6a331ce74fa7d11e0bfe"],"summary_source_locator":"Official biohub/ESMFold2 card: Model Details, Model Variants, Training Data and Biases and Limitations; official source README and THIRD_PARTY_NOTICE.md","sections":[{"title":"How it works","body":"ESMFold2 predicts biomolecular structures from protein, DNA, RNA and ligand inputs, optionally using protein alignments. ESM C 6B embeddings coupled to a diffusion-based structure prediction architecture. The documented inputs are protein, DNA/RNA, modified-residue and small-molecule specifications; optional protein MSAs for the full variant. The output consists of all-atom complex coordinates with confidence estimates and optional distogram predictions.","source_ids":["evidence-official-4bf005200ebfb2586b78","evidence-official-0437be6150a3da9f8efc","evidence-official-1f027f8ecc53decdeec8","evidence-official-d65bf8ba587b6841e0d9","evidence-official-6a331ce74fa7d11e0bfe"],"source_locator":"Official biohub/ESMFold2 card: Model Details, Model Variants, Training Data and Biases and Limitations; official source README and THIRD_PARTY_NOTICE.md"},{"title":"Versions and reproducibility","body":"ESMFold2 supports optional MSA conditioning; ESMFold2-Fast is a distinct single-sequence inference variant. The applicable input limits require configuration-specific checking.","source_ids":["evidence-official-4bf005200ebfb2586b78","evidence-official-0437be6150a3da9f8efc","evidence-official-1f027f8ecc53decdeec8","evidence-official-d65bf8ba587b6841e0d9","evidence-official-6a331ce74fa7d11e0bfe"],"source_locator":"Official biohub/ESMFold2 card: Model Details, Model Variants, Training Data and Biases and Limitations; official source README and THIRD_PARTY_NOTICE.md"}],"facts":[{"label":"Model type","value":"Biomolecular diffusion structure predictor with protein language-model features","status":"source_checked","source_ids":["evidence-official-4bf005200ebfb2586b78","evidence-official-0437be6150a3da9f8efc","evidence-official-1f027f8ecc53decdeec8","evidence-official-d65bf8ba587b6841e0d9","evidence-official-6a331ce74fa7d11e0bfe"],"source_locator":"Official biohub/ESMFold2 card: Model Details, Model Variants, Training Data and Biases and Limitations; official source README and THIRD_PARTY_NOTICE.md"},{"label":"Architecture","value":"ESM C 6B embeddings coupled to a diffusion-based structure prediction architecture.","status":"source_checked","source_ids":["evidence-official-4bf005200ebfb2586b78","evidence-official-0437be6150a3da9f8efc","evidence-official-1f027f8ecc53decdeec8","evidence-official-d65bf8ba587b6841e0d9","evidence-official-6a331ce74fa7d11e0bfe"],"source_locator":"Official biohub/ESMFold2 card: Model Details, Model Variants, Training Data and Biases and Limitations; official source README and THIRD_PARTY_NOTICE.md"},{"label":"Inputs","value":"Protein, DNA/RNA, modified-residue and small-molecule specifications; optional protein MSAs for the full variant.","status":"source_checked","source_ids":["evidence-official-4bf005200ebfb2586b78","evidence-official-0437be6150a3da9f8efc","evidence-official-1f027f8ecc53decdeec8","evidence-official-d65bf8ba587b6841e0d9","evidence-official-6a331ce74fa7d11e0bfe"],"source_locator":"Official biohub/ESMFold2 card: Model Details, Model Variants, Training Data and Biases and Limitations; official source README and THIRD_PARTY_NOTICE.md"},{"label":"Outputs","value":"All-atom complex coordinates with confidence estimates and optional distogram predictions.","status":"source_checked","source_ids":["evidence-official-4bf005200ebfb2586b78","evidence-official-0437be6150a3da9f8efc","evidence-official-1f027f8ecc53decdeec8","evidence-official-d65bf8ba587b6841e0d9","evidence-official-6a331ce74fa7d11e0bfe"],"source_locator":"Official biohub/ESMFold2 card: Model Details, Model Variants, Training Data and Biases and Limitations; official source README and THIRD_PARTY_NOTICE.md"},{"label":"Parameters","value":"The official card identifies a 6B ESM C backbone. It does not state a total including the complete molecular structure and confidence modules.","status":"unreported","source_ids":["evidence-official-4bf005200ebfb2586b78","evidence-official-0437be6150a3da9f8efc","evidence-official-1f027f8ecc53decdeec8","evidence-official-d65bf8ba587b6841e0d9","evidence-official-6a331ce74fa7d11e0bfe"],"source_locator":"Official biohub/ESMFold2 card: Model Details, Model Variants, Training Data and Biases and Limitations; official source README and THIRD_PARTY_NOTICE.md"},{"label":"Known versions","value":"ESMFold2 supports optional MSA conditioning; ESMFold2-Fast is a distinct single-sequence inference variant.","status":"source_checked","source_ids":["evidence-official-4bf005200ebfb2586b78","evidence-official-0437be6150a3da9f8efc","evidence-official-1f027f8ecc53decdeec8","evidence-official-d65bf8ba587b6841e0d9","evidence-official-6a331ce74fa7d11e0bfe"],"source_locator":"Official biohub/ESMFold2 card: Model Details, Model Variants, Training Data and Biases and Limitations; official source README and THIRD_PARTY_NOTICE.md"},{"label":"Training data","value":"The checked model card identifies PDB structures and AlphaFoldDB-derived training data.","status":"source_checked","source_ids":["evidence-official-4bf005200ebfb2586b78","evidence-official-0437be6150a3da9f8efc","evidence-official-1f027f8ecc53decdeec8","evidence-official-d65bf8ba587b6841e0d9","evidence-official-6a331ce74fa7d11e0bfe"],"source_locator":"Official biohub/ESMFold2 card: Model Details, Model Variants, Training Data and Biases and Limitations; official source README and THIRD_PARTY_NOTICE.md"},{"label":"Training cutoff","value":"September 2021 is the model-card data cutoff for both ESMFold2 and ESMFold2-Fast; the separate ESMC backbone corpus has its own provenance.","status":"source_checked","source_ids":["evidence-official-4bf005200ebfb2586b78","evidence-official-0437be6150a3da9f8efc","evidence-official-1f027f8ecc53decdeec8","evidence-official-d65bf8ba587b6841e0d9","evidence-official-6a331ce74fa7d11e0bfe"],"source_locator":"Official biohub/ESMFold2 card: Model Details, Model Variants, Training Data and Biases and Limitations; official source README and THIRD_PARTY_NOTICE.md"},{"label":"Context limits","value":"The inspected card describes Full and Fast variants without giving a single maximum token/atom budget that applies to every supported molecular composition.","status":"unreported","source_ids":["evidence-official-4bf005200ebfb2586b78","evidence-official-0437be6150a3da9f8efc","evidence-official-1f027f8ecc53decdeec8","evidence-official-d65bf8ba587b6841e0d9","evidence-official-6a331ce74fa7d11e0bfe"],"source_locator":"Official biohub/ESMFold2 card: Model Details, Model Variants, Training Data and Biases and Limitations; official source README and THIRD_PARTY_NOTICE.md"},{"label":"Weights licence","value":"MIT declared in the official ESMFold2 card, with linked third-party dependency notices.","status":"source_checked","source_ids":["evidence-official-4bf005200ebfb2586b78","evidence-official-0437be6150a3da9f8efc","evidence-official-1f027f8ecc53decdeec8","evidence-official-d65bf8ba587b6841e0d9","evidence-official-6a331ce74fa7d11e0bfe"],"source_locator":"Official biohub/ESMFold2 card: Model Details, Model Variants, Training Data and Biases and Limitations; official source README and THIRD_PARTY_NOTICE.md"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/evolutionaryscale/esm","status":"source_checked","source_ids":["evidence-official-4bf005200ebfb2586b78","evidence-official-0437be6150a3da9f8efc","evidence-official-1f027f8ecc53decdeec8","evidence-official-d65bf8ba587b6841e0d9","evidence-official-6a331ce74fa7d11e0bfe"],"source_locator":"Official biohub/ESMFold2 card: Model Details, Model Variants, Training Data and Biases and Limitations; official source README and THIRD_PARTY_NOTICE.md"},{"label":"Code licence","value":"MIT","status":"source_checked","source_ids":["evidence-official-2f4711ecd64b0162e85b"],"source_locator":"LICENSE.md: licence text"}],"strengths":[{"text":"Supports both sequence-only prediction and an MSA-assisted mode.","source_ids":["evidence-official-4bf005200ebfb2586b78","evidence-official-0437be6150a3da9f8efc","evidence-official-1f027f8ecc53decdeec8","evidence-official-d65bf8ba587b6841e0d9","evidence-official-6a331ce74fa7d11e0bfe"],"source_locator":"Official biohub/ESMFold2 card: Model Details, Model Variants, Training Data and Biases and Limitations; official source README and THIRD_PARTY_NOTICE.md"}],"limitations":[{"text":"The 6B figure describes the ESM C backbone, not a verified parameter total for the complete folding system. Sequence-only and MSA-assisted runs need separate evaluation identities.","source_ids":["evidence-official-4bf005200ebfb2586b78","evidence-official-0437be6150a3da9f8efc","evidence-official-1f027f8ecc53decdeec8","evidence-official-d65bf8ba587b6841e0d9","evidence-official-6a331ce74fa7d11e0bfe"],"source_locator":"Official biohub/ESMFold2 card: Model Details, Model Variants, Training Data and Biases and Limitations; official source README and THIRD_PARTY_NOTICE.md"}],"diagram":{"title":"ESMFold2 workflow","steps":["Protein, DNA, RNA and ligands","Molecular features; ESM C for proteins","Optional protein MSA conditioning","Diffusion structure prediction","Complex coordinates and confidence"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-4bf005200ebfb2586b78","evidence-official-0437be6150a3da9f8efc","evidence-official-1f027f8ecc53decdeec8","evidence-official-d65bf8ba587b6841e0d9","evidence-official-6a331ce74fa7d11e0bfe"],"source_locator":"Official biohub/ESMFold2 card: Model Details, Model Variants, Training Data and Biases and Limitations; official source README and THIRD_PARTY_NOTICE.md"},"coverage":"limited","gaps":["Parameters: The official card identifies a 6B ESM C backbone. It does not state a total including the complete molecular structure and confidence modules.","Context limits: The inspected card describes Full and Fast variants without giving a single maximum token/atom budget that applies to every supported molecular composition."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}},"description":"Protein structure and complex prediction","facets":{"areas":["protein-structure"]},"id":"discovery-model-esmfold2","kind":"model","links":[],"name":"ESMFold2","source_ids":["src-discovery-evolutionaryscale-esm"],"status":"discovered"}
{"attributes":{"access":"official_source_linked","benchmark_applicability":"candidate; not evidence of a reported evaluation","candidate_benchmark_ids":["discovery-benchmark-geneb"],"entity_level":"family","reported_name":"Evo 2","version":null,"historical_missing_metadata":{"checkpoint":"unextracted","code_licence":"unextracted","parameters":"unextracted","training_cutoff":"unextracted","training_data":"unextracted","version":"unextracted","weights_licence":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Evo 2 models and generates DNA over long contexts at single-nucleotide resolution.","summary_source_ids":["evidence-official-a248990b44d481814f47","evidence-official-849b751ed4bf88622749","evidence-official-b8df803e8f55324a4eb5","evidence-official-8717ffb993cc8f7eddec"],"summary_source_locator":"Evo 2 paper: architecture, training, and data, Figure 1 and Discussion; official checkpoints table and evo2_7b licence metadata","sections":[{"title":"How it works","body":"Evo 2 models and generates DNA over long contexts at single-nucleotide resolution. StripedHyena 2 hybrid architecture combining short, medium and long convolution operators with attention, trained autoregressively at single-base resolution. The documented inputs are DNA sequences represented at single-base resolution. The output consists of next-token outputs, embeddings and generated DNA sequences.","source_ids":["evidence-official-a248990b44d481814f47","evidence-official-849b751ed4bf88622749","evidence-official-b8df803e8f55324a4eb5","evidence-official-8717ffb993cc8f7eddec"],"source_locator":"Evo 2 paper: architecture, training, and data, Figure 1 and Discussion; official checkpoints table and evo2_7b licence metadata"},{"title":"Versions and reproducibility","body":"Base 8K models, long-context 1M models, 7B 262K model and separately fine-tuned Microviridae model. Checkpoint-dependent: 8K, 262K or 1M bases as listed in the Checkpoints table.","source_ids":["evidence-official-a248990b44d481814f47","evidence-official-849b751ed4bf88622749","evidence-official-b8df803e8f55324a4eb5","evidence-official-8717ffb993cc8f7eddec"],"source_locator":"Evo 2 paper: architecture, training, and data, Figure 1 and Discussion; official checkpoints table and evo2_7b licence metadata"}],"facts":[{"label":"Model type","value":"Autoregressive DNA model with StripedHyena 2","status":"source_checked","source_ids":["evidence-official-a248990b44d481814f47","evidence-official-849b751ed4bf88622749","evidence-official-b8df803e8f55324a4eb5","evidence-official-8717ffb993cc8f7eddec"],"source_locator":"Evo 2 paper: architecture, training, and data, Figure 1 and Discussion; official checkpoints table and evo2_7b licence metadata"},{"label":"Architecture","value":"StripedHyena 2 hybrid architecture combining short, medium and long convolution operators with attention, trained autoregressively at single-base resolution.","status":"source_checked","source_ids":["evidence-official-a248990b44d481814f47","evidence-official-849b751ed4bf88622749","evidence-official-b8df803e8f55324a4eb5","evidence-official-8717ffb993cc8f7eddec"],"source_locator":"Evo 2 paper: architecture, training, and data, Figure 1 and Discussion; official checkpoints table and evo2_7b licence metadata"},{"label":"Inputs","value":"DNA sequences represented at single-base resolution.","status":"source_checked","source_ids":["evidence-official-a248990b44d481814f47","evidence-official-849b751ed4bf88622749","evidence-official-b8df803e8f55324a4eb5","evidence-official-8717ffb993cc8f7eddec"],"source_locator":"Evo 2 paper: architecture, training, and data, Figure 1 and Discussion; official checkpoints table and evo2_7b licence metadata"},{"label":"Outputs","value":"Next-token outputs, embeddings and generated DNA sequences.","status":"source_checked","source_ids":["evidence-official-a248990b44d481814f47","evidence-official-849b751ed4bf88622749","evidence-official-b8df803e8f55324a4eb5","evidence-official-8717ffb993cc8f7eddec"],"source_locator":"Evo 2 paper: architecture, training, and data, Figure 1 and Discussion; official checkpoints table and evo2_7b licence metadata"},{"label":"Parameters","value":"1B, 7B, 20B and 40B checkpoints are listed.","status":"source_checked","source_ids":["evidence-official-a248990b44d481814f47","evidence-official-849b751ed4bf88622749","evidence-official-b8df803e8f55324a4eb5","evidence-official-8717ffb993cc8f7eddec"],"source_locator":"Evo 2 paper: architecture, training, and data, Figure 1 and Discussion; official checkpoints table and evo2_7b licence metadata"},{"label":"Known versions","value":"Base 8K models, long-context 1M models, 7B 262K model and separately fine-tuned Microviridae model.","status":"source_checked","source_ids":["evidence-official-a248990b44d481814f47","evidence-official-849b751ed4bf88622749","evidence-official-b8df803e8f55324a4eb5","evidence-official-8717ffb993cc8f7eddec"],"source_locator":"Evo 2 paper: architecture, training, and data, Figure 1 and Discussion; official checkpoints table and evo2_7b licence metadata"},{"label":"Training data","value":"OpenGenome2 contains more than 8.8T curated nucleotides across bacteria, archaea, eukaryotes and bacteriophage. The paper separates 2.4T tokens of training exposure for 7B from 9.3T for 40B; eukaryotic-host viral sequences were excluded.","status":"source_checked","source_ids":["evidence-official-a248990b44d481814f47","evidence-official-849b751ed4bf88622749","evidence-official-b8df803e8f55324a4eb5","evidence-official-8717ffb993cc8f7eddec"],"source_locator":"Evo 2 paper: architecture, training, and data, Figure 1 and Discussion; official checkpoints table and evo2_7b licence metadata"},{"label":"Training cutoff","value":"OpenGenome2 combines multiple nucleotide collections. The inspected paper and released checkpoint documentation do not provide one latest-deposition date that covers every component.","status":"unreported","source_ids":["evidence-official-a248990b44d481814f47","evidence-official-849b751ed4bf88622749","evidence-official-b8df803e8f55324a4eb5","evidence-official-8717ffb993cc8f7eddec"],"source_locator":"Evo 2 paper: architecture, training, and data, Figure 1 and Discussion; official checkpoints table and evo2_7b licence metadata"},{"label":"Context limits","value":"Checkpoint-dependent: 8K, 262K or 1M bases as listed in the Checkpoints table.","status":"source_checked","source_ids":["evidence-official-a248990b44d481814f47","evidence-official-849b751ed4bf88622749","evidence-official-b8df803e8f55324a4eb5","evidence-official-8717ffb993cc8f7eddec"],"source_locator":"Evo 2 paper: architecture, training, and data, Figure 1 and Discussion; official checkpoints table and evo2_7b licence metadata"},{"label":"Weights licence","value":"Apache-2.0 is declared in the inspected ArcInstitute/evo2_7b model card; other checkpoints require their own pinned terms.","status":"source_checked","source_ids":["evidence-official-a248990b44d481814f47","evidence-official-849b751ed4bf88622749","evidence-official-b8df803e8f55324a4eb5","evidence-official-8717ffb993cc8f7eddec"],"source_locator":"Evo 2 paper: architecture, training, and data, Figure 1 and Discussion; official checkpoints table and evo2_7b licence metadata"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/ArcInstitute/evo2","status":"source_checked","source_ids":["evidence-official-a248990b44d481814f47","evidence-official-849b751ed4bf88622749","evidence-official-b8df803e8f55324a4eb5","evidence-official-8717ffb993cc8f7eddec"],"source_locator":"Evo 2 paper: architecture, training, and data, Figure 1 and Discussion; official checkpoints table and evo2_7b licence metadata"},{"label":"Code licence","value":"Apache-2.0","status":"source_checked","source_ids":["evidence-official-67ee1cc31060ba8c9569"],"source_locator":"LICENSE: licence text"}],"strengths":[{"text":"Different released context lengths and model scales support a range of sequence modeling workflows.","source_ids":["evidence-official-a248990b44d481814f47","evidence-official-849b751ed4bf88622749","evidence-official-b8df803e8f55324a4eb5","evidence-official-8717ffb993cc8f7eddec"],"source_locator":"Evo 2 paper: architecture, training, and data, Figure 1 and Discussion; official checkpoints table and evo2_7b licence metadata"}],"limitations":[{"text":"Hardware requirements differ by checkpoint: the README requires FP8/Transformer Engine and Hopper GPUs for some scales, while 7B supports bfloat 16 on a wider set of GPUs.","source_ids":["evidence-official-a248990b44d481814f47","evidence-official-849b751ed4bf88622749","evidence-official-b8df803e8f55324a4eb5","evidence-official-8717ffb993cc8f7eddec"],"source_locator":"Evo 2 paper: architecture, training, and data, Figure 1 and Discussion; official checkpoints table and evo2_7b licence metadata"}],"diagram":{"title":"Evo 2 workflow","steps":["DNA bases","StripedHyena 2","Autoregressive outputs","Sequence scoring or generation"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-a248990b44d481814f47","evidence-official-849b751ed4bf88622749","evidence-official-b8df803e8f55324a4eb5","evidence-official-8717ffb993cc8f7eddec"],"source_locator":"Evo 2 paper: architecture, training, and data, Figure 1 and Discussion; official checkpoints table and evo2_7b licence metadata"},"coverage":"limited","gaps":["Training cutoff: OpenGenome2 combines multiple nucleotide collections. The inspected paper and released checkpoint documentation do not provide one latest-deposition date that covers every component."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}},"description":"Genome sequence modelling family","facets":{"areas":["genomics"]},"id":"discovery-model-evo-2","kind":"model","links":[{"relation":"applicable_to","target_id":"discovery-benchmark-geneb"}],"name":"Evo 2","source_ids":["src-discovery-arcinstitute-evo2"],"status":"discovered"}
{"attributes":{"access":"official_source_linked","benchmark_applicability":"candidate; not evidence of a reported evaluation","candidate_benchmark_ids":["discovery-benchmark-dart-eval"],"entity_level":"method","reported_name":"FIMO","version":null,"historical_missing_metadata":{"checkpoint":"unextracted","code_licence":"unextracted","parameters":"unextracted","training_cutoff":"unextracted","training_data":"unextracted","version":"unextracted","weights_licence":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"FIMO scans sequences for occurrences of supplied sequence motifs.","summary_source_ids":["evidence-official-f0b17e30e69338b7762f","evidence-official-94c61d963d28fa3a8993"],"summary_source_locator":"Official FIMO page: motif input, sequence input, p-value threshold and strand options","sections":[{"title":"How it works","body":"FIMO scans sequences for occurrences of supplied sequence motifs. Position-specific motif scanning with statistical match thresholds. The documented inputs are DNA, RNA or protein sequences and motif matrices in the supported alphabet/formats. The output consists of motif matches at sequence positions subject to the configured statistical threshold.","source_ids":["evidence-official-f0b17e30e69338b7762f","evidence-official-94c61d963d28fa3a8993"],"source_locator":"Official FIMO page: motif input, sequence input, p-value threshold and strand options"},{"title":"Versions and reproducibility","body":"MEME Suite FIMO; site snapshot pinned by retrieval time and hash. Input sequences and motif lengths, not a fixed language-model context.","source_ids":["evidence-official-f0b17e30e69338b7762f","evidence-official-94c61d963d28fa3a8993"],"source_locator":"Official FIMO page: motif input, sequence input, p-value threshold and strand options"}],"facts":[{"label":"Model type","value":"Statistical motif scanning procedure","status":"source_checked","source_ids":["evidence-official-f0b17e30e69338b7762f","evidence-official-94c61d963d28fa3a8993"],"source_locator":"Official FIMO page: motif input, sequence input, p-value threshold and strand options"},{"label":"Architecture","value":"Position-specific motif scanning with statistical match thresholds.","status":"source_checked","source_ids":["evidence-official-f0b17e30e69338b7762f","evidence-official-94c61d963d28fa3a8993"],"source_locator":"Official FIMO page: motif input, sequence input, p-value threshold and strand options"},{"label":"Inputs","value":"DNA, RNA or protein sequences and motif matrices in the supported alphabet/formats.","status":"source_checked","source_ids":["evidence-official-f0b17e30e69338b7762f","evidence-official-94c61d963d28fa3a8993"],"source_locator":"Official FIMO page: motif input, sequence input, p-value threshold and strand options"},{"label":"Outputs","value":"Motif matches at sequence positions subject to the configured statistical threshold.","status":"source_checked","source_ids":["evidence-official-f0b17e30e69338b7762f","evidence-official-94c61d963d28fa3a8993"],"source_locator":"Official FIMO page: motif input, sequence input, p-value threshold and strand options"},{"label":"Parameters","value":"Motif probabilities and scan settings; not a neural parameter count.","status":"source_checked","source_ids":["evidence-official-f0b17e30e69338b7762f","evidence-official-94c61d963d28fa3a8993"],"source_locator":"Official FIMO page: motif input, sequence input, p-value threshold and strand options"},{"label":"Known versions","value":"MEME Suite FIMO; site snapshot pinned by retrieval time and hash.","status":"source_checked","source_ids":["evidence-official-f0b17e30e69338b7762f","evidence-official-94c61d963d28fa3a8993"],"source_locator":"Official FIMO page: motif input, sequence input, p-value threshold and strand options"},{"label":"Training data","value":"Motifs are supplied by the user or a selected database. FIMO is the scanning procedure, not the motif-training method.","status":"source_checked","source_ids":["evidence-official-f0b17e30e69338b7762f","evidence-official-94c61d963d28fa3a8993"],"source_locator":"Official FIMO page: motif input, sequence input, p-value threshold and strand options"},{"label":"Training cutoff","value":"Inapplicable to neural pretraining; reference-database and input-data dates must be recorded for each run.","status":"inapplicable","source_ids":["evidence-official-f0b17e30e69338b7762f","evidence-official-94c61d963d28fa3a8993"],"source_locator":"Official FIMO page: motif input, sequence input, p-value threshold and strand options"},{"label":"Context limits","value":"Input sequences and motif lengths, not a fixed language-model context.","status":"source_checked","source_ids":["evidence-official-f0b17e30e69338b7762f","evidence-official-94c61d963d28fa3a8993"],"source_locator":"Official FIMO page: motif input, sequence input, p-value threshold and strand options"},{"label":"Weights licence","value":"Inapplicable: supplied motif matrices replace neural weights.","status":"inapplicable","source_ids":["evidence-official-f0b17e30e69338b7762f","evidence-official-94c61d963d28fa3a8993"],"source_locator":"Official FIMO page: motif input, sequence input, p-value threshold and strand options"},{"label":"Access","value":"Official project documentation and implementation: https://meme-suite.org/meme/tools/fimo","status":"source_checked","source_ids":["evidence-official-f0b17e30e69338b7762f","evidence-official-94c61d963d28fa3a8993"],"source_locator":"Official FIMO page: motif input, sequence input, p-value threshold and strand options"},{"label":"Code licence","value":"MEME Suite custom terms permit educational, research and non-profit use with notices retained; commercial use requires contacting UC San Diego for terms.","status":"source_checked","source_ids":["evidence-official-94c61d963d28fa3a8993"],"source_locator":"MEME Suite copyright notice: permission grant and commercial-use paragraph"}],"strengths":[{"text":"Provides a simple motif-based reference whose supplied motif and background assumptions can be reported explicitly.","source_ids":["evidence-official-f0b17e30e69338b7762f","evidence-official-94c61d963d28fa3a8993"],"source_locator":"Official FIMO page: motif input, sequence input, p-value threshold and strand options"}],"limitations":[{"text":"Motif matches do not establish molecular binding or regulation by themselves. Alphabet, strand and threshold choices affect the scan.","source_ids":["evidence-official-f0b17e30e69338b7762f","evidence-official-94c61d963d28fa3a8993"],"source_locator":"Official FIMO page: motif input, sequence input, p-value threshold and strand options"}],"diagram":{"title":"FIMO workflow","steps":["Motif matrices and sequence","Scan motif positions","Apply statistical threshold","Reported motif occurrences"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-f0b17e30e69338b7762f","evidence-official-94c61d963d28fa3a8993"],"source_locator":"Official FIMO page: motif input, sequence input, p-value threshold and strand options"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}},"description":"Sequence motif scanning","facets":{"areas":["genomics"]},"id":"discovery-model-fimo","kind":"model","links":[{"relation":"applicable_to","target_id":"discovery-benchmark-dart-eval"}],"name":"FIMO","source_ids":["src-discovery-meme"],"status":"discovered"}
{"attributes":{"access":"official_source_linked","benchmark_applicability":"candidate; not evidence of a reported evaluation","candidate_benchmark_ids":["discovery-benchmark-perturbench"],"entity_level":"family","reported_name":"GEARS","version":null,"historical_missing_metadata":{"checkpoint":"unextracted","code_licence":"unextracted","parameters":"unextracted","training_cutoff":"unextracted","training_data":"unextracted","version":"unextracted","weights_licence":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"GEARS predicts transcriptional responses to single- and multi-gene perturbations from single-cell perturbation screens.","summary_source_ids":["evidence-official-c037e3419c04936262a0","evidence-official-b5e81b2c497214b18f8a","evidence-official-f464cd407a6a9d031720"],"summary_source_locator":"GEARS paper Methods: Overview, Gene coexpression graph encoder, GO graph, Cross-gene effects and gene-specific decoder; gears/model.py: GEARS_Model","sections":[{"title":"How it works","body":"GEARS predicts transcriptional responses to single- and multi-gene perturbations from single-cell perturbation screens. Two graph encoders represent gene coexpression and Gene Ontology perturbation similarity. Perturbation embeddings are composed with gene embeddings, then a cross-gene network and gene-specific decoders predict expression changes. The documented inputs are single-cell expression data, perturbation labels and the graph resources used by the configured model. The output consists of predicted gene-expression responses and genetic-interaction analyses.","source_ids":["evidence-official-c037e3419c04936262a0","evidence-official-b5e81b2c497214b18f8a","evidence-official-f464cd407a6a9d031720"],"source_locator":"GEARS paper Methods: Overview, Gene coexpression graph encoder, GO graph, Cross-gene effects and gene-specific decoder; gears/model.py: GEARS_Model"},{"title":"Versions and reproducibility","body":"The README describes v0.1.1 updates; a specific trained checkpoint must be recorded separately. A gene-expression vector and perturbation set over the configured gene inventory; no fixed nucleotide or amino-acid token window.","source_ids":["evidence-official-c037e3419c04936262a0","evidence-official-b5e81b2c497214b18f8a","evidence-official-f464cd407a6a9d031720"],"source_locator":"GEARS paper Methods: Overview, Gene coexpression graph encoder, GO graph, Cross-gene effects and gene-specific decoder; gears/model.py: GEARS_Model"}],"facts":[{"label":"Model type","value":"Graph-based perturbation-response predictor","status":"source_checked","source_ids":["evidence-official-c037e3419c04936262a0","evidence-official-b5e81b2c497214b18f8a","evidence-official-f464cd407a6a9d031720"],"source_locator":"GEARS paper Methods: Overview, Gene coexpression graph encoder, GO graph, Cross-gene effects and gene-specific decoder; gears/model.py: GEARS_Model"},{"label":"Architecture","value":"Two graph encoders represent gene coexpression and Gene Ontology perturbation similarity. Perturbation embeddings are composed with gene embeddings, then a cross-gene network and gene-specific decoders predict expression changes.","status":"source_checked","source_ids":["evidence-official-c037e3419c04936262a0","evidence-official-b5e81b2c497214b18f8a","evidence-official-f464cd407a6a9d031720"],"source_locator":"GEARS paper Methods: Overview, Gene coexpression graph encoder, GO graph, Cross-gene effects and gene-specific decoder; gears/model.py: GEARS_Model"},{"label":"Inputs","value":"Single-cell expression data, perturbation labels and the graph resources used by the configured model.","status":"source_checked","source_ids":["evidence-official-c037e3419c04936262a0","evidence-official-b5e81b2c497214b18f8a","evidence-official-f464cd407a6a9d031720"],"source_locator":"GEARS paper Methods: Overview, Gene coexpression graph encoder, GO graph, Cross-gene effects and gene-specific decoder; gears/model.py: GEARS_Model"},{"label":"Outputs","value":"Predicted gene-expression responses and genetic-interaction analyses.","status":"source_checked","source_ids":["evidence-official-c037e3419c04936262a0","evidence-official-b5e81b2c497214b18f8a","evidence-official-f464cd407a6a9d031720"],"source_locator":"GEARS paper Methods: Overview, Gene coexpression graph encoder, GO graph, Cross-gene effects and gene-specific decoder; gears/model.py: GEARS_Model"},{"label":"Parameters","value":"Configuration-dependent: gene and perturbation embedding tables grow with the selected gene/perturbation inventory, alongside graph and decoder parameters.","status":"source_checked","source_ids":["evidence-official-c037e3419c04936262a0","evidence-official-b5e81b2c497214b18f8a","evidence-official-f464cd407a6a9d031720"],"source_locator":"GEARS paper Methods: Overview, Gene coexpression graph encoder, GO graph, Cross-gene effects and gene-specific decoder; gears/model.py: GEARS_Model"},{"label":"Known versions","value":"The README describes v0.1.1 updates; a specific trained checkpoint must be recorded separately.","status":"source_checked","source_ids":["evidence-official-c037e3419c04936262a0","evidence-official-b5e81b2c497214b18f8a","evidence-official-f464cd407a6a9d031720"],"source_locator":"GEARS paper Methods: Overview, Gene coexpression graph encoder, GO graph, Cross-gene effects and gene-specific decoder; gears/model.py: GEARS_Model"},{"label":"Training data","value":"Fitted to the selected perturbation screen. Examples include Norman, Adamson and Dixit; the repository also lists Replogle RPE1/K562 loaders.","status":"source_checked","source_ids":["evidence-official-c037e3419c04936262a0","evidence-official-b5e81b2c497214b18f8a","evidence-official-f464cd407a6a9d031720"],"source_locator":"GEARS paper Methods: Overview, Gene coexpression graph encoder, GO graph, Cross-gene effects and gene-specific decoder; gears/model.py: GEARS_Model"},{"label":"Training cutoff","value":"Inapplicable as a universal pretrained-model cutoff: GEARS fits the provided perturbation training set and builds its coexpression graph from that set.","status":"inapplicable","source_ids":["evidence-official-c037e3419c04936262a0","evidence-official-b5e81b2c497214b18f8a","evidence-official-f464cd407a6a9d031720"],"source_locator":"GEARS paper Methods: Overview, Gene coexpression graph encoder, GO graph, Cross-gene effects and gene-specific decoder; gears/model.py: GEARS_Model"},{"label":"Context limits","value":"A gene-expression vector and perturbation set over the configured gene inventory; no fixed nucleotide or amino-acid token window.","status":"source_checked","source_ids":["evidence-official-c037e3419c04936262a0","evidence-official-b5e81b2c497214b18f8a","evidence-official-f464cd407a6a9d031720"],"source_locator":"GEARS paper Methods: Overview, Gene coexpression graph encoder, GO graph, Cross-gene effects and gene-specific decoder; gears/model.py: GEARS_Model"},{"label":"Weights licence","value":"Separate checkpoint-distribution terms are not stated in the inspected release documentation and licence material. The source-code licence alone is not recorded as an explicit weight grant.","status":"unreported","source_ids":["evidence-official-c037e3419c04936262a0","evidence-official-b5e81b2c497214b18f8a","evidence-official-f464cd407a6a9d031720","evidence-official-f06a8695b2915f86a45a"],"source_locator":"GEARS paper Methods: Overview, Gene coexpression graph encoder, GO graph, Cross-gene effects and gene-specific decoder; gears/model.py: GEARS_Model; LICENSE: licence text"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/snap-stanford/GEARS","status":"source_checked","source_ids":["evidence-official-c037e3419c04936262a0","evidence-official-b5e81b2c497214b18f8a","evidence-official-f464cd407a6a9d031720"],"source_locator":"GEARS paper Methods: Overview, Gene coexpression graph encoder, GO graph, Cross-gene effects and gene-specific decoder; gears/model.py: GEARS_Model"},{"label":"Code licence","value":"MIT","status":"source_checked","source_ids":["evidence-official-f06a8695b2915f86a45a"],"source_locator":"LICENSE: licence text"}],"strengths":[{"text":"Provides documented dataset/split handling and training interfaces for single and combinatorial perturbations.","source_ids":["evidence-official-c037e3419c04936262a0","evidence-official-b5e81b2c497214b18f8a","evidence-official-f464cd407a6a9d031720"],"source_locator":"GEARS paper Methods: Overview, Gene coexpression graph encoder, GO graph, Cross-gene effects and gene-specific decoder; gears/model.py: GEARS_Model"}],"limitations":[{"text":"The authors explicitly warn against cross-cell-type transfer, bulk-RNA assumptions and predicting combinations after training only on single perturbations.","source_ids":["evidence-official-c037e3419c04936262a0","evidence-official-b5e81b2c497214b18f8a","evidence-official-f464cd407a6a9d031720"],"source_locator":"GEARS paper Methods: Overview, Gene coexpression graph encoder, GO graph, Cross-gene effects and gene-specific decoder; gears/model.py: GEARS_Model"}],"diagram":{"title":"GEARS workflow","steps":["Perturbation screen","Gene and perturbation graph representations","GEARS prediction","Expression response"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-c037e3419c04936262a0","evidence-official-b5e81b2c497214b18f8a","evidence-official-f464cd407a6a9d031720"],"source_locator":"GEARS paper Methods: Overview, Gene coexpression graph encoder, GO graph, Cross-gene effects and gene-specific decoder; gears/model.py: GEARS_Model"},"coverage":"limited","gaps":["Weights licence: Separate checkpoint-distribution terms are not stated in the inspected release documentation and licence material. The source-code licence alone is not recorded as an explicit weight grant."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}},"description":"Genetic perturbation response prediction","facets":{"areas":["single-cell"]},"id":"discovery-model-gears","kind":"model","links":[{"relation":"applicable_to","target_id":"discovery-benchmark-perturbench"}],"name":"GEARS","source_ids":["src-discovery-snap-stanford-gears"],"status":"discovered"}
{"attributes":{"access":"official_source_linked","benchmark_applicability":"candidate; not evidence of a reported evaluation","candidate_benchmark_ids":[],"entity_level":"family","reported_name":"Genie 3","version":null,"historical_missing_metadata":{"checkpoint":"unextracted","code_licence":"unextracted","parameters":"unextracted","training_cutoff":"unextracted","training_data":"unextracted","version":"unextracted","weights_licence":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Genie 3 generates protein designs through all-atom equivariant diffusion.","summary_source_ids":["evidence-official-082d60e1af5af7ed12ca","evidence-official-a07254268b1d1de1f7f8"],"summary_source_locator":"README.md: overview, Download model weights and training data, Training and Codebase Architecture; author-linked yeqinglin/genie3 licence metadata; README.md: Training / Dataset manifest table","sections":[{"title":"How it works","body":"Genie 3 generates protein designs through all-atom equivariant diffusion. All-atom SE(3)-equivariant diffusion model, with separate generation and evaluation workflows. The documented inputs are unconditional design specification, motif constraints or binder-design target context. The output consists of sampled protein designs and outputs from the selected downstream evaluation workflow.","source_ids":["evidence-official-082d60e1af5af7ed12ca","evidence-official-a07254268b1d1de1f7f8"],"source_locator":"README.md: overview, Download model weights and training data, Training and Codebase Architecture; author-linked yeqinglin/genie3 licence metadata; README.md: Training / Dataset manifest table"},{"title":"Versions and reproducibility","body":"Genie 3; repository includes compatibility guidance for Genie 2. The applicable input limits require configuration-specific checking.","source_ids":["evidence-official-082d60e1af5af7ed12ca","evidence-official-a07254268b1d1de1f7f8"],"source_locator":"README.md: overview, Download model weights and training data, Training and Codebase Architecture; author-linked yeqinglin/genie3 licence metadata; README.md: Training / Dataset manifest table"}],"facts":[{"label":"Model type","value":"Diffusion-based protein backbone generator","status":"source_checked","source_ids":["evidence-official-082d60e1af5af7ed12ca","evidence-official-a07254268b1d1de1f7f8"],"source_locator":"README.md: overview, Download model weights and training data, Training and Codebase Architecture; author-linked yeqinglin/genie3 licence metadata; README.md: Training / Dataset manifest table"},{"label":"Architecture","value":"All-atom SE(3)-equivariant diffusion model, with separate generation and evaluation workflows.","status":"source_checked","source_ids":["evidence-official-082d60e1af5af7ed12ca","evidence-official-a07254268b1d1de1f7f8"],"source_locator":"README.md: overview, Download model weights and training data, Training and Codebase Architecture; author-linked yeqinglin/genie3 licence metadata; README.md: Training / Dataset manifest table"},{"label":"Inputs","value":"Unconditional design specification, motif constraints or binder-design target context.","status":"source_checked","source_ids":["evidence-official-082d60e1af5af7ed12ca","evidence-official-a07254268b1d1de1f7f8"],"source_locator":"README.md: overview, Download model weights and training data, Training and Codebase Architecture; author-linked yeqinglin/genie3 licence metadata; README.md: Training / Dataset manifest table"},{"label":"Outputs","value":"Sampled protein designs and outputs from the selected downstream evaluation workflow.","status":"source_checked","source_ids":["evidence-official-082d60e1af5af7ed12ca","evidence-official-a07254268b1d1de1f7f8"],"source_locator":"README.md: overview, Download model weights and training data, Training and Codebase Architecture; author-linked yeqinglin/genie3 licence metadata; README.md: Training / Dataset manifest table"},{"label":"Parameters","value":"The inspected release README and model card do not state a complete parameter total for the released all-atom diffusion checkpoint.","status":"unreported","source_ids":["evidence-official-082d60e1af5af7ed12ca","evidence-official-a07254268b1d1de1f7f8"],"source_locator":"README.md: overview, Download model weights and training data, Training and Codebase Architecture; author-linked yeqinglin/genie3 licence metadata; README.md: Training / Dataset manifest table"},{"label":"Known versions","value":"Genie 3; repository includes compatibility guidance for Genie 2.","status":"source_checked","source_ids":["evidence-official-082d60e1af5af7ed12ca","evidence-official-a07254268b1d1de1f7f8"],"source_locator":"README.md: overview, Download model weights and training data, Training and Codebase Architecture; author-linked yeqinglin/genie3 licence metadata; README.md: Training / Dataset manifest table"},{"label":"Training data","value":"The released training manifests cover AlphaFoldDB representatives of at most 512 residues with pLDDT at least 70, and PiNDER 2024-02. These filters describe training components, not inference limits.","status":"source_checked","source_ids":["evidence-official-082d60e1af5af7ed12ca","evidence-official-a07254268b1d1de1f7f8"],"source_locator":"README.md: overview, Download model weights and training data, Training and Codebase Architecture; author-linked yeqinglin/genie3 licence metadata; README.md: Training / Dataset manifest table"},{"label":"Training cutoff","value":"The documented training inputs include PiNDER 2024-02 and filtered AlphaFoldDB representatives. This identifies a PiNDER release, not a universal latest-deposition cutoff for all training data.","status":"source_checked","source_ids":["evidence-official-082d60e1af5af7ed12ca","evidence-official-a07254268b1d1de1f7f8"],"source_locator":"README.md: overview, Download model weights and training data, Training and Codebase Architecture; author-linked yeqinglin/genie3 licence metadata; README.md: Training / Dataset manifest table"},{"label":"Context limits","value":"The README permits configured design-length ranges and distinguishes sampling settings above and below 300 residues; it does not state one validated maximum for all monomer, motif and binder tasks.","status":"unreported","source_ids":["evidence-official-082d60e1af5af7ed12ca","evidence-official-a07254268b1d1de1f7f8"],"source_locator":"README.md: overview, Download model weights and training data, Training and Codebase Architecture; author-linked yeqinglin/genie3 licence metadata; README.md: Training / Dataset manifest table"},{"label":"Weights licence","value":"Apache-2.0 declared in the author-linked yeqinglin/genie3 model card.","status":"source_checked","source_ids":["evidence-official-082d60e1af5af7ed12ca","evidence-official-a07254268b1d1de1f7f8"],"source_locator":"README.md: overview, Download model weights and training data, Training and Codebase Architecture; author-linked yeqinglin/genie3 licence metadata; README.md: Training / Dataset manifest table"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/aqlaboratory/genie3","status":"source_checked","source_ids":["evidence-official-082d60e1af5af7ed12ca","evidence-official-a07254268b1d1de1f7f8"],"source_locator":"README.md: overview, Download model weights and training data, Training and Codebase Architecture; author-linked yeqinglin/genie3 licence metadata; README.md: Training / Dataset manifest table"},{"label":"Code licence","value":"Apache-2.0","status":"source_checked","source_ids":["evidence-official-bbfa62298e3bb7c5494d"],"source_locator":"LICENSE: licence text"}],"strengths":[{"text":"One documented interface supports unconditional generation, motif scaffolding and binder design.","source_ids":["evidence-official-082d60e1af5af7ed12ca","evidence-official-a07254268b1d1de1f7f8"],"source_locator":"README.md: overview, Download model weights and training data, Training and Codebase Architecture; author-linked yeqinglin/genie3 licence metadata; README.md: Training / Dataset manifest table"}],"limitations":[{"text":"Genie 3 protein design is unrelated to GENIE3 gene-regulatory-network inference. A computationally generated design needs separate experimental validation.","source_ids":["evidence-official-082d60e1af5af7ed12ca","evidence-official-a07254268b1d1de1f7f8"],"source_locator":"README.md: overview, Download model weights and training data, Training and Codebase Architecture; author-linked yeqinglin/genie3 licence metadata; README.md: Training / Dataset manifest table"}],"diagram":{"title":"Genie 3 workflow","steps":["Design constraints","Equivariant diffusion","All-atom design","Configured evaluation"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-082d60e1af5af7ed12ca","evidence-official-a07254268b1d1de1f7f8"],"source_locator":"README.md: overview, Download model weights and training data, Training and Codebase Architecture; author-linked yeqinglin/genie3 licence metadata; README.md: Training / Dataset manifest table"},"coverage":"limited","gaps":["Parameters: The inspected release README and model card do not state a complete parameter total for the released all-atom diffusion checkpoint.","Context limits: The README permits configured design-length ranges and distinguishes sampling settings above and below 300 residues; it does not state one validated maximum for all monomer, motif and binder tasks."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}},"description":"Equivariant all-atom protein design","facets":{"areas":["protein-structure"]},"id":"discovery-model-genie-3","kind":"model","links":[],"name":"Genie 3","source_ids":["src-discovery-aqlaboratory-genie3"],"status":"discovered"}
{"attributes":{"access":"official_source_linked","benchmark_applicability":"candidate; not evidence of a reported evaluation","candidate_benchmark_ids":["discovery-benchmark-beeline"],"entity_level":"method","reported_name":"GENIE3","version":null,"historical_missing_metadata":{"checkpoint":"unextracted","code_licence":"unextracted","parameters":"unextracted","training_cutoff":"unextracted","training_data":"unextracted","version":"unextracted","weights_licence":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"GENIE3 infers candidate gene-regulatory links from expression measurements using ensembles of trees.","summary_source_ids":["evidence-official-f078ec10fa45b5d77ad7","evidence-official-ad7f5c1eb802d9895413","evidence-official-099d109417c7fae96e96"],"summary_source_locator":"README.md: GENIE3 definition; GRNBoost README Introduction describes the GENIE3 target-wise regression blueprint","sections":[{"title":"How it works","body":"GENIE3 infers candidate gene-regulatory links from expression measurements using ensembles of trees. Tree-ensemble regression decomposes network inference into prediction of each target gene from candidate regulators. The documented inputs are gene-expression matrix and the selected candidate-regulator set. The output consists of ranked candidate regulator-to-target links.","source_ids":["evidence-official-f078ec10fa45b5d77ad7","evidence-official-ad7f5c1eb802d9895413","evidence-official-099d109417c7fae96e96"],"source_locator":"README.md: GENIE3 definition; GRNBoost README Introduction describes the GENIE3 target-wise regression blueprint"},{"title":"Versions and reproducibility","body":"GENIE3 implementation and Bioconductor package; preserve the configured tree method and settings. Genes and observations in the input matrix; no token context.","source_ids":["evidence-official-f078ec10fa45b5d77ad7","evidence-official-ad7f5c1eb802d9895413","evidence-official-099d109417c7fae96e96"],"source_locator":"README.md: GENIE3 definition; GRNBoost README Introduction describes the GENIE3 target-wise regression blueprint"}],"facts":[{"label":"Model type","value":"Tree-ensemble gene-regulatory-network inference","status":"source_checked","source_ids":["evidence-official-f078ec10fa45b5d77ad7","evidence-official-ad7f5c1eb802d9895413","evidence-official-099d109417c7fae96e96"],"source_locator":"README.md: GENIE3 definition; GRNBoost README Introduction describes the GENIE3 target-wise regression blueprint"},{"label":"Architecture","value":"Tree-ensemble regression decomposes network inference into prediction of each target gene from candidate regulators.","status":"source_checked","source_ids":["evidence-official-f078ec10fa45b5d77ad7","evidence-official-ad7f5c1eb802d9895413","evidence-official-099d109417c7fae96e96"],"source_locator":"README.md: GENIE3 definition; GRNBoost README Introduction describes the GENIE3 target-wise regression blueprint"},{"label":"Inputs","value":"Gene-expression matrix and the selected candidate-regulator set.","status":"source_checked","source_ids":["evidence-official-f078ec10fa45b5d77ad7","evidence-official-ad7f5c1eb802d9895413","evidence-official-099d109417c7fae96e96"],"source_locator":"README.md: GENIE3 definition; GRNBoost README Introduction describes the GENIE3 target-wise regression blueprint"},{"label":"Outputs","value":"Ranked candidate regulator-to-target links.","status":"source_checked","source_ids":["evidence-official-f078ec10fa45b5d77ad7","evidence-official-ad7f5c1eb802d9895413","evidence-official-099d109417c7fae96e96"],"source_locator":"README.md: GENIE3 definition; GRNBoost README Introduction describes the GENIE3 target-wise regression blueprint"},{"label":"Parameters","value":"Fitted tree ensembles; count depends on genes, candidate regulators and tree settings.","status":"source_checked","source_ids":["evidence-official-f078ec10fa45b5d77ad7","evidence-official-ad7f5c1eb802d9895413","evidence-official-099d109417c7fae96e96"],"source_locator":"README.md: GENIE3 definition; GRNBoost README Introduction describes the GENIE3 target-wise regression blueprint"},{"label":"Known versions","value":"GENIE3 implementation and Bioconductor package; preserve the configured tree method and settings.","status":"source_checked","source_ids":["evidence-official-f078ec10fa45b5d77ad7","evidence-official-ad7f5c1eb802d9895413","evidence-official-099d109417c7fae96e96"],"source_locator":"README.md: GENIE3 definition; GRNBoost README Introduction describes the GENIE3 target-wise regression blueprint"},{"label":"Training data","value":"Fitted to the supplied gene-expression dataset, not a universal pretrained corpus.","status":"source_checked","source_ids":["evidence-official-f078ec10fa45b5d77ad7","evidence-official-ad7f5c1eb802d9895413","evidence-official-099d109417c7fae96e96"],"source_locator":"README.md: GENIE3 definition; GRNBoost README Introduction describes the GENIE3 target-wise regression blueprint"},{"label":"Training cutoff","value":"Inapplicable to neural pretraining; reference-database and input-data dates must be recorded for each run.","status":"inapplicable","source_ids":["evidence-official-f078ec10fa45b5d77ad7","evidence-official-ad7f5c1eb802d9895413","evidence-official-099d109417c7fae96e96"],"source_locator":"README.md: GENIE3 definition; GRNBoost README Introduction describes the GENIE3 target-wise regression blueprint"},{"label":"Context limits","value":"Genes and observations in the input matrix; no token context.","status":"source_checked","source_ids":["evidence-official-f078ec10fa45b5d77ad7","evidence-official-ad7f5c1eb802d9895413","evidence-official-099d109417c7fae96e96"],"source_locator":"README.md: GENIE3 definition; GRNBoost README Introduction describes the GENIE3 target-wise regression blueprint"},{"label":"Weights licence","value":"No universal pretrained checkpoint; each fitted network is dataset-specific.","status":"source_checked","source_ids":["evidence-official-f078ec10fa45b5d77ad7","evidence-official-ad7f5c1eb802d9895413","evidence-official-099d109417c7fae96e96"],"source_locator":"README.md: GENIE3 definition; GRNBoost README Introduction describes the GENIE3 target-wise regression blueprint"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/aertslab/GENIE3","status":"source_checked","source_ids":["evidence-official-f078ec10fa45b5d77ad7","evidence-official-ad7f5c1eb802d9895413","evidence-official-099d109417c7fae96e96"],"source_locator":"README.md: GENIE3 definition; GRNBoost README Introduction describes the GENIE3 target-wise regression blueprint"},{"label":"Code licence","value":"GPL-2.0-or-later, as declared in the R package DESCRIPTION.","status":"source_checked","source_ids":["evidence-official-ad7f5c1eb802d9895413"],"source_locator":"DESCRIPTION: License"}],"strengths":[{"text":"A conventional machine-learning comparator whose target-wise regression and regulator importance are explicit.","source_ids":["evidence-official-f078ec10fa45b5d77ad7","evidence-official-ad7f5c1eb802d9895413","evidence-official-099d109417c7fae96e96"],"source_locator":"README.md: GENIE3 definition; GRNBoost README Introduction describes the GENIE3 target-wise regression blueprint"}],"limitations":[{"text":"Predictive association in expression data does not by itself establish causal regulation. This GENIE3 algorithm is unrelated to Genie 3 protein design.","source_ids":["evidence-official-f078ec10fa45b5d77ad7","evidence-official-ad7f5c1eb802d9895413","evidence-official-099d109417c7fae96e96"],"source_locator":"README.md: GENIE3 definition; GRNBoost README Introduction describes the GENIE3 target-wise regression blueprint"}],"diagram":{"title":"GENIE3 workflow","steps":["Expression matrix","Target-wise tree ensembles","Regulator importance","Candidate network links"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-f078ec10fa45b5d77ad7","evidence-official-ad7f5c1eb802d9895413","evidence-official-099d109417c7fae96e96"],"source_locator":"README.md: GENIE3 definition; GRNBoost README Introduction describes the GENIE3 target-wise regression blueprint"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}},"description":"Tree-ensemble gene regulatory network inference","facets":{"areas":["biological-networks"]},"id":"discovery-model-genie3","kind":"model","links":[{"relation":"applicable_to","target_id":"discovery-benchmark-beeline"}],"name":"GENIE3","source_ids":["src-discovery-aertslab-genie3"],"status":"discovered"}
{"attributes":{"access":"official_source_linked","benchmark_applicability":"candidate; not evidence of a reported evaluation","candidate_benchmark_ids":["discovery-benchmark-glycanml"],"entity_level":"family","reported_name":"GlycanGT","version":null,"historical_missing_metadata":{"checkpoint":"unextracted","code_licence":"unextracted","parameters":"unextracted","training_cutoff":"unextracted","training_data":"unextracted","version":"unextracted","weights_licence":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"GlycanGT learns glycan representations by treating monosaccharides and linkages as graph tokens.","summary_source_ids":["evidence-official-10a6550ca70fdeaf33cb","evidence-official-790ab6283e5bb91474ed","evidence-official-dacb012b22c0a7de4880","evidence-official-87c70bddba97a72a5e5d","evidence-official-ba12a2d759aba6800ae6"],"summary_source_locator":"GlycanGT paper Sections 2.1–2.4,2.9 and 4; official model-card licence metadata; primary paper Sections 2.1 and 3.1 versus README Model architecture / Training details; model/config_tokengt.py: size presets; publisher supplementary Table S2, Total Parameters row","sections":[{"title":"How it works","body":"GlycanGT turns each monosaccharide and glycosidic linkage into a token. Content embeddings, orthogonal node identifiers and token-type embeddings preserve graph structure before transformer attention. A graph token provides the whole-glycan representation; masked-token heads can suggest missing components, while downstream classifiers are fitted separately.","source_ids":["evidence-official-10a6550ca70fdeaf33cb","evidence-official-790ab6283e5bb91474ed","evidence-official-dacb012b22c0a7de4880","evidence-official-87c70bddba97a72a5e5d","evidence-official-ba12a2d759aba6800ae6"],"source_locator":"GlycanGT paper Sections 2.1–2.4,2.9 and 4; official model-card licence metadata; primary paper Sections 2.1 and 3.1 versus README Model architecture / Training details; model/config_tokengt.py: size presets; publisher supplementary Table S2, Total Parameters row"},{"title":"Versions and reproducibility","body":"ss, small, medium and large scales; the documented downstream work uses the large model with 35% masking. The applicable input limits require configuration-specific checking.","source_ids":["evidence-official-10a6550ca70fdeaf33cb","evidence-official-790ab6283e5bb91474ed","evidence-official-dacb012b22c0a7de4880","evidence-official-87c70bddba97a72a5e5d","evidence-official-ba12a2d759aba6800ae6"],"source_locator":"GlycanGT paper Sections 2.1–2.4,2.9 and 4; official model-card licence metadata; primary paper Sections 2.1 and 3.1 versus README Model architecture / Training details; model/config_tokengt.py: size presets; publisher supplementary Table S2, Total Parameters row"}],"facts":[{"label":"Model type","value":"Glycan graph transformer","status":"source_checked","source_ids":["evidence-official-10a6550ca70fdeaf33cb","evidence-official-790ab6283e5bb91474ed","evidence-official-dacb012b22c0a7de4880","evidence-official-87c70bddba97a72a5e5d","evidence-official-ba12a2d759aba6800ae6"],"source_locator":"GlycanGT paper Sections 2.1–2.4,2.9 and 4; official model-card licence metadata; primary paper Sections 2.1 and 3.1 versus README Model architecture / Training details; model/config_tokengt.py: size presets; publisher supplementary Table S2, Total Parameters row"},{"label":"Architecture","value":"TokenGT-derived graph transformer with content features, orthogonal node identifiers, node/edge type embeddings and a graph token.","status":"source_checked","source_ids":["evidence-official-10a6550ca70fdeaf33cb","evidence-official-790ab6283e5bb91474ed","evidence-official-dacb012b22c0a7de4880","evidence-official-87c70bddba97a72a5e5d","evidence-official-ba12a2d759aba6800ae6"],"source_locator":"GlycanGT paper Sections 2.1–2.4,2.9 and 4; official model-card licence metadata; primary paper Sections 2.1 and 3.1 versus README Model architecture / Training details; model/config_tokengt.py: size presets; publisher supplementary Table S2, Total Parameters row"},{"label":"Inputs","value":"Glycan graphs with monosaccharide nodes and glycosidic-linkage edges.","status":"source_checked","source_ids":["evidence-official-10a6550ca70fdeaf33cb","evidence-official-790ab6283e5bb91474ed","evidence-official-dacb012b22c0a7de4880","evidence-official-87c70bddba97a72a5e5d","evidence-official-ba12a2d759aba6800ae6"],"source_locator":"GlycanGT paper Sections 2.1–2.4,2.9 and 4; official model-card licence metadata; primary paper Sections 2.1 and 3.1 versus README Model architecture / Training details; model/config_tokengt.py: size presets; publisher supplementary Table S2, Total Parameters row"},{"label":"Outputs","value":"Graph embeddings and masked-component predictions for incomplete glycans.","status":"source_checked","source_ids":["evidence-official-10a6550ca70fdeaf33cb","evidence-official-790ab6283e5bb91474ed","evidence-official-dacb012b22c0a7de4880","evidence-official-87c70bddba97a72a5e5d","evidence-official-ba12a2d759aba6800ae6"],"source_locator":"GlycanGT paper Sections 2.1–2.4,2.9 and 4; official model-card licence metadata; primary paper Sections 2.1 and 3.1 versus README Model architecture / Training details; model/config_tokengt.py: size presets; publisher supplementary Table S2, Total Parameters row"},{"label":"Parameters","value":"ss: 1,922,002; small: 6,208,722; medium: 29,731,026; large: 91,820,242 total parameters, as printed in supplementary Table S2.","status":"source_checked","source_ids":["evidence-official-10a6550ca70fdeaf33cb","evidence-official-790ab6283e5bb91474ed","evidence-official-dacb012b22c0a7de4880","evidence-official-87c70bddba97a72a5e5d","evidence-official-ba12a2d759aba6800ae6"],"source_locator":"GlycanGT paper Sections 2.1–2.4,2.9 and 4; official model-card licence metadata; primary paper Sections 2.1 and 3.1 versus README Model architecture / Training details; model/config_tokengt.py: size presets; publisher supplementary Table S2, Total Parameters row"},{"label":"Known versions","value":"ss, small, medium and large scales; the documented downstream work uses the large model with 35% masking.","status":"source_checked","source_ids":["evidence-official-10a6550ca70fdeaf33cb","evidence-official-790ab6283e5bb91474ed","evidence-official-dacb012b22c0a7de4880","evidence-official-87c70bddba97a72a5e5d","evidence-official-ba12a2d759aba6800ae6"],"source_locator":"GlycanGT paper Sections 2.1–2.4,2.9 and 4; official model-card licence metadata; primary paper Sections 2.1 and 3.1 versus README Model architecture / Training details; model/config_tokengt.py: size presets; publisher supplementary Table S2, Total Parameters row"},{"label":"Training data","value":"The primary paper reports 83,739 curated glycans after ambiguity and downstream-overlap exclusions; the repository README reports 83,740. Both source values are preserved pending an author correction.","status":"source_checked","source_ids":["evidence-official-10a6550ca70fdeaf33cb","evidence-official-790ab6283e5bb91474ed","evidence-official-dacb012b22c0a7de4880","evidence-official-87c70bddba97a72a5e5d","evidence-official-ba12a2d759aba6800ae6"],"source_locator":"GlycanGT paper Sections 2.1–2.4,2.9 and 4; official model-card licence metadata; primary paper Sections 2.1 and 3.1 versus README Model architecture / Training details; model/config_tokengt.py: size presets; publisher supplementary Table S2, Total Parameters row"},{"label":"Training cutoff","value":"GlyCosmos/GlyTouCan source retrieval: 17 June 2025. The paper describes removal of downstream-benchmark overlaps from pretraining.","status":"source_checked","source_ids":["evidence-official-10a6550ca70fdeaf33cb","evidence-official-790ab6283e5bb91474ed","evidence-official-dacb012b22c0a7de4880","evidence-official-87c70bddba97a72a5e5d","evidence-official-ba12a2d759aba6800ae6"],"source_locator":"GlycanGT paper Sections 2.1–2.4,2.9 and 4; official model-card licence metadata; primary paper Sections 2.1 and 3.1 versus README Model architecture / Training details; model/config_tokengt.py: size presets; publisher supplementary Table S2, Total Parameters row"},{"label":"Context limits","value":"The inspected model presets set graph-identifier dimensions and attention sizes, not a validated maximum glycan node count. Token length depends on both monosaccharides and linkages.","status":"unreported","source_ids":["evidence-official-10a6550ca70fdeaf33cb","evidence-official-790ab6283e5bb91474ed","evidence-official-dacb012b22c0a7de4880","evidence-official-87c70bddba97a72a5e5d","evidence-official-ba12a2d759aba6800ae6"],"source_locator":"GlycanGT paper Sections 2.1–2.4,2.9 and 4; official model-card licence metadata; primary paper Sections 2.1 and 3.1 versus README Model architecture / Training details; model/config_tokengt.py: size presets; publisher supplementary Table S2, Total Parameters row"},{"label":"Weights licence","value":"Apache-2.0 declared in the author-linked Akikitani295/GlycanGT model-card metadata.","status":"source_checked","source_ids":["evidence-official-10a6550ca70fdeaf33cb","evidence-official-790ab6283e5bb91474ed","evidence-official-dacb012b22c0a7de4880","evidence-official-87c70bddba97a72a5e5d","evidence-official-ba12a2d759aba6800ae6"],"source_locator":"GlycanGT paper Sections 2.1–2.4,2.9 and 4; official model-card licence metadata; primary paper Sections 2.1 and 3.1 versus README Model architecture / Training details; model/config_tokengt.py: size presets; publisher supplementary Table S2, Total Parameters row"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/matsui-lab/GlycanGT","status":"source_checked","source_ids":["evidence-official-10a6550ca70fdeaf33cb","evidence-official-790ab6283e5bb91474ed","evidence-official-dacb012b22c0a7de4880","evidence-official-87c70bddba97a72a5e5d","evidence-official-ba12a2d759aba6800ae6"],"source_locator":"GlycanGT paper Sections 2.1–2.4,2.9 and 4; official model-card licence metadata; primary paper Sections 2.1 and 3.1 versus README Model architecture / Training details; model/config_tokengt.py: size presets; publisher supplementary Table S2, Total Parameters row"},{"label":"Code licence","value":"Apache-2.0","status":"source_checked","source_ids":["evidence-official-9576e3936b5a16a999cf"],"source_locator":"LICENSE: licence text"}],"strengths":[{"text":"Represents linkages explicitly and supports prediction of ambiguous glycan components.","source_ids":["evidence-official-10a6550ca70fdeaf33cb","evidence-official-790ab6283e5bb91474ed","evidence-official-dacb012b22c0a7de4880","evidence-official-87c70bddba97a72a5e5d","evidence-official-ba12a2d759aba6800ae6"],"source_locator":"GlycanGT paper Sections 2.1–2.4,2.9 and 4; official model-card licence metadata; primary paper Sections 2.1 and 3.1 versus README Model architecture / Training details; model/config_tokengt.py: size presets; publisher supplementary Table S2, Total Parameters row"}],"limitations":[{"text":"The training corpus excluded ambiguous entries. Downstream taxonomy, glycosylation and immunogenicity tasks require their own evaluation data and heads.","source_ids":["evidence-official-10a6550ca70fdeaf33cb","evidence-official-790ab6283e5bb91474ed","evidence-official-dacb012b22c0a7de4880","evidence-official-87c70bddba97a72a5e5d","evidence-official-ba12a2d759aba6800ae6"],"source_locator":"GlycanGT paper Sections 2.1–2.4,2.9 and 4; official model-card licence metadata; primary paper Sections 2.1 and 3.1 versus README Model architecture / Training details; model/config_tokengt.py: size presets; publisher supplementary Table S2, Total Parameters row"}],"diagram":{"title":"GlycanGT workflow","steps":["Glycan graph","Node and edge tokens","Graph transformer","Embedding or masked component"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-10a6550ca70fdeaf33cb","evidence-official-790ab6283e5bb91474ed","evidence-official-dacb012b22c0a7de4880","evidence-official-87c70bddba97a72a5e5d","evidence-official-ba12a2d759aba6800ae6"],"source_locator":"GlycanGT paper Sections 2.1–2.4,2.9 and 4; official model-card licence metadata; primary paper Sections 2.1 and 3.1 versus README Model architecture / Training details; model/config_tokengt.py: size presets; publisher supplementary Table S2, Total Parameters row"},"coverage":"limited","gaps":["Context limits: The inspected model presets set graph-identifier dimensions and attention sizes, not a validated maximum glycan node count. Token length depends on both monosaccharides and linkages."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}},"description":"Glycan graph transformer","facets":{"areas":["glycomics"]},"id":"discovery-model-glycangt","kind":"model","links":[{"relation":"applicable_to","target_id":"discovery-benchmark-glycanml"}],"name":"GlycanGT","source_ids":["src-discovery-matsui-lab-glycangt"],"status":"discovered"}
{"attributes":{"access":"official_source_linked","benchmark_applicability":"candidate; not evidence of a reported evaluation","candidate_benchmark_ids":["discovery-benchmark-beeline"],"entity_level":"method","reported_name":"GRNBoost","version":null,"historical_missing_metadata":{"checkpoint":"unextracted","code_licence":"unextracted","parameters":"unextracted","training_cutoff":"unextracted","training_data":"unextracted","version":"unextracted","weights_licence":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"GRNBoost infers gene-regulatory networks with distributed gradient-boosted regression.","summary_source_ids":["evidence-official-099d109417c7fae96e96"],"summary_source_locator":"README.md: Introduction and License","sections":[{"title":"How it works","body":"GRNBoost infers gene-regulatory networks with distributed gradient-boosted regression. Spark pipeline implementing target-wise regression with XGBoost, following the GENIE3 inference blueprint. The documented inputs are gene-expression data and candidate regulatory genes. The output consists of predictive regulator-to-target links derived from the fitted regressions.","source_ids":["evidence-official-099d109417c7fae96e96"],"source_locator":"README.md: Introduction and License"},{"title":"Versions and reproducibility","body":"Spark/XGBoost GRNBoost implementation; do not substitute GRNBoost2 silently. Input gene matrix and distributed memory; no sequence token window.","source_ids":["evidence-official-099d109417c7fae96e96"],"source_locator":"README.md: Introduction and License"}],"facts":[{"label":"Model type","value":"Gradient-boosted regulatory-network inference","status":"source_checked","source_ids":["evidence-official-099d109417c7fae96e96"],"source_locator":"README.md: Introduction and License"},{"label":"Architecture","value":"Spark pipeline implementing target-wise regression with XGBoost, following the GENIE3 inference blueprint.","status":"source_checked","source_ids":["evidence-official-099d109417c7fae96e96"],"source_locator":"README.md: Introduction and License"},{"label":"Inputs","value":"Gene-expression data and candidate regulatory genes.","status":"source_checked","source_ids":["evidence-official-099d109417c7fae96e96"],"source_locator":"README.md: Introduction and License"},{"label":"Outputs","value":"Predictive regulator-to-target links derived from the fitted regressions.","status":"source_checked","source_ids":["evidence-official-099d109417c7fae96e96"],"source_locator":"README.md: Introduction and License"},{"label":"Parameters","value":"Dataset- and boosting-configuration-dependent.","status":"source_checked","source_ids":["evidence-official-099d109417c7fae96e96"],"source_locator":"README.md: Introduction and License"},{"label":"Known versions","value":"Spark/XGBoost GRNBoost implementation; do not substitute GRNBoost2 silently.","status":"source_checked","source_ids":["evidence-official-099d109417c7fae96e96"],"source_locator":"README.md: Introduction and License"},{"label":"Training data","value":"Fitted to the supplied expression data.","status":"source_checked","source_ids":["evidence-official-099d109417c7fae96e96"],"source_locator":"README.md: Introduction and License"},{"label":"Training cutoff","value":"Inapplicable to neural pretraining; reference-database and input-data dates must be recorded for each run.","status":"inapplicable","source_ids":["evidence-official-099d109417c7fae96e96"],"source_locator":"README.md: Introduction and License"},{"label":"Context limits","value":"Input gene matrix and distributed memory; no sequence token window.","status":"source_checked","source_ids":["evidence-official-099d109417c7fae96e96"],"source_locator":"README.md: Introduction and License"},{"label":"Weights licence","value":"No universal neural checkpoint; fitted regressors depend on the input dataset.","status":"source_checked","source_ids":["evidence-official-099d109417c7fae96e96"],"source_locator":"README.md: Introduction and License"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/aertslab/GRNBoost","status":"source_checked","source_ids":["evidence-official-099d109417c7fae96e96"],"source_locator":"README.md: Introduction and License"},{"label":"Code licence","value":"BSD-3-Clause","status":"source_checked","source_ids":["evidence-official-af4cadc1f0b4c69528d4"],"source_locator":"LICENSE.txt: licence text"}],"strengths":[{"text":"Distributes target-wise regression using Spark and replaces random forests with gradient boosting.","source_ids":["evidence-official-099d109417c7fae96e96"],"source_locator":"README.md: Introduction and License"}],"limitations":[{"text":"This repository describes GRNBoost, not every later GRNBoost2 implementation. Predictive links remain hypotheses about regulation.","source_ids":["evidence-official-099d109417c7fae96e96"],"source_locator":"README.md: Introduction and License"}],"diagram":{"title":"GRNBoost workflow","steps":["Expression and regulator set","Distributed target-wise boosting","Feature importance","Candidate regulatory network"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-099d109417c7fae96e96"],"source_locator":"README.md: Introduction and License"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}},"description":"Boosted-tree regulatory network inference","facets":{"areas":["biological-networks"]},"id":"discovery-model-grnboost","kind":"model","links":[{"relation":"applicable_to","target_id":"discovery-benchmark-beeline"}],"name":"GRNBoost","source_ids":["src-discovery-aertslab-grnboost"],"status":"discovered"}
{"attributes":{"access":"official_source_linked","benchmark_applicability":"candidate; not evidence of a reported evaluation","candidate_benchmark_ids":["discovery-benchmark-cafa"],"entity_level":"method","reported_name":"HH-suite","version":null,"historical_missing_metadata":{"checkpoint":"unextracted","code_licence":"unextracted","parameters":"unextracted","training_cutoff":"unextracted","training_data":"unextracted","version":"unextracted","weights_licence":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"HH-suite searches for remote protein relationships using profile hidden Markov models.","summary_source_ids":["evidence-official-863f6c2e6e18152c2f8b"],"summary_source_locator":"README.md: opening, Available Databases and Usage","sections":[{"title":"How it works","body":"HH-suite searches for remote protein relationships using profile hidden Markov models. Pairwise profile-HMM alignment with iterative homolog search utilities such as HHblits. The documented inputs are protein sequence or alignment/profile and a selected reference profile database. The output consists of homolog hits, alignments and related search scores.","source_ids":["evidence-official-863f6c2e6e18152c2f8b"],"source_locator":"README.md: opening, Available Databases and Usage"},{"title":"Versions and reproducibility","body":"HH-suite3; README includes v3.3.0 binaries. Query/alignment size and database settings, not a learned token context.","source_ids":["evidence-official-863f6c2e6e18152c2f8b"],"source_locator":"README.md: opening, Available Databases and Usage"}],"facts":[{"label":"Model type","value":"Profile-HMM sequence search software","status":"source_checked","source_ids":["evidence-official-863f6c2e6e18152c2f8b"],"source_locator":"README.md: opening, Available Databases and Usage"},{"label":"Architecture","value":"Pairwise profile-HMM alignment with iterative homolog search utilities such as HHblits.","status":"source_checked","source_ids":["evidence-official-863f6c2e6e18152c2f8b"],"source_locator":"README.md: opening, Available Databases and Usage"},{"label":"Inputs","value":"Protein sequence or alignment/profile and a selected reference profile database.","status":"source_checked","source_ids":["evidence-official-863f6c2e6e18152c2f8b"],"source_locator":"README.md: opening, Available Databases and Usage"},{"label":"Outputs","value":"Homolog hits, alignments and related search scores.","status":"source_checked","source_ids":["evidence-official-863f6c2e6e18152c2f8b"],"source_locator":"README.md: opening, Available Databases and Usage"},{"label":"Parameters","value":"Profile probabilities and search settings; not a fixed neural parameter count.","status":"source_checked","source_ids":["evidence-official-863f6c2e6e18152c2f8b"],"source_locator":"README.md: opening, Available Databases and Usage"},{"label":"Known versions","value":"HH-suite3; README includes v3.3.0 binaries.","status":"source_checked","source_ids":["evidence-official-863f6c2e6e18152c2f8b"],"source_locator":"README.md: opening, Available Databases and Usage"},{"label":"Training data","value":"Profiles/reference databases rather than neural pretraining; examples include Uniclust30, BFD and PDB70.","status":"source_checked","source_ids":["evidence-official-863f6c2e6e18152c2f8b"],"source_locator":"README.md: opening, Available Databases and Usage"},{"label":"Training cutoff","value":"Inapplicable to neural pretraining; reference-database and input-data dates must be recorded for each run.","status":"inapplicable","source_ids":["evidence-official-863f6c2e6e18152c2f8b"],"source_locator":"README.md: opening, Available Databases and Usage"},{"label":"Context limits","value":"Query/alignment size and database settings, not a learned token context.","status":"source_checked","source_ids":["evidence-official-863f6c2e6e18152c2f8b"],"source_locator":"README.md: opening, Available Databases and Usage"},{"label":"Weights licence","value":"Inapplicable: reference profiles/databases are the relevant artifacts.","status":"inapplicable","source_ids":["evidence-official-863f6c2e6e18152c2f8b"],"source_locator":"README.md: opening, Available Databases and Usage"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/soedinglab/hh-suite","status":"source_checked","source_ids":["evidence-official-863f6c2e6e18152c2f8b"],"source_locator":"README.md: opening, Available Databases and Usage"},{"label":"Code licence","value":"GPL-3.0; inspect the pinned licence and any file-specific terms.","status":"source_checked","source_ids":["evidence-official-00ee97a5586392ae2348"],"source_locator":"LICENSE: licence text"}],"strengths":[{"text":"Can search at the profile level and construct alignments of homologous sequences.","source_ids":["evidence-official-863f6c2e6e18152c2f8b"],"source_locator":"README.md: opening, Available Databases and Usage"}],"limitations":[{"text":"Search results depend on database coverage and iteration/settings. Profile similarity is evidence of sequence relationships rather than a direct functional assay.","source_ids":["evidence-official-863f6c2e6e18152c2f8b"],"source_locator":"README.md: opening, Available Databases and Usage"}],"diagram":{"title":"HH-suite workflow","steps":["Query sequence or alignment","Profile HMM","Reference profile search","Homologs and alignment"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-863f6c2e6e18152c2f8b"],"source_locator":"README.md: opening, Available Databases and Usage"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}},"description":"Profile hidden Markov sequence search","facets":{"areas":["protein-function"]},"id":"discovery-model-hh-suite","kind":"model","links":[{"relation":"applicable_to","target_id":"discovery-benchmark-cafa"}],"name":"HH-suite","source_ids":["src-discovery-soedinglab-hh-suite"],"status":"discovered"}
{"attributes":{"access":"official_source_linked","benchmark_applicability":"candidate; not evidence of a reported evaluation","candidate_benchmark_ids":[],"entity_level":"method","reported_name":"HUMAnN","version":null,"historical_missing_metadata":{"checkpoint":"unextracted","code_licence":"unextracted","parameters":"unextracted","training_cutoff":"unextracted","training_data":"unextracted","version":"unextracted","weights_licence":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"HUMAnN profiles microbial gene functions and pathways from metagenomic or metatranscriptomic reads.","summary_source_ids":["evidence-official-10f58e320879def8b102"],"summary_source_locator":"readme.md: description, Main workflow, Download the databases and normalization documentation","sections":[{"title":"How it works","body":"HUMAnN profiles microbial gene functions and pathways from metagenomic or metatranscriptomic reads. Reference-based functional profiling pipeline with nucleotide and translated search against configured databases. The documented inputs are short DNA/RNA reads or supported preprocessed alignment/profile inputs. The output consists of gene-family and pathway abundance/coverage tables, including organism-stratified outputs.","source_ids":["evidence-official-10f58e320879def8b102"],"source_locator":"readme.md: description, Main workflow, Download the databases and normalization documentation"},{"title":"Versions and reproducibility","body":"HUMAnN 3.0 paper and implementation; nucleotide/protein database versions must be preserved separately. Read files and search-database size; no fixed model context.","source_ids":["evidence-official-10f58e320879def8b102"],"source_locator":"readme.md: description, Main workflow, Download the databases and normalization documentation"}],"facts":[{"label":"Model type","value":"Metagenomic functional profiling pipeline","status":"source_checked","source_ids":["evidence-official-10f58e320879def8b102"],"source_locator":"readme.md: description, Main workflow, Download the databases and normalization documentation"},{"label":"Architecture","value":"Reference-based functional profiling pipeline with nucleotide and translated search against configured databases.","status":"source_checked","source_ids":["evidence-official-10f58e320879def8b102"],"source_locator":"readme.md: description, Main workflow, Download the databases and normalization documentation"},{"label":"Inputs","value":"Short DNA/RNA reads or supported preprocessed alignment/profile inputs.","status":"source_checked","source_ids":["evidence-official-10f58e320879def8b102"],"source_locator":"readme.md: description, Main workflow, Download the databases and normalization documentation"},{"label":"Outputs","value":"Gene-family and pathway abundance/coverage tables, including organism-stratified outputs.","status":"source_checked","source_ids":["evidence-official-10f58e320879def8b102"],"source_locator":"readme.md: description, Main workflow, Download the databases and normalization documentation"},{"label":"Parameters","value":"Inapplicable as a neural parameter count.","status":"inapplicable","source_ids":["evidence-official-10f58e320879def8b102"],"source_locator":"readme.md: description, Main workflow, Download the databases and normalization documentation"},{"label":"Known versions","value":"HUMAnN 3.0 paper and implementation; nucleotide/protein database versions must be preserved separately.","status":"source_checked","source_ids":["evidence-official-10f58e320879def8b102"],"source_locator":"readme.md: description, Main workflow, Download the databases and normalization documentation"},{"label":"Training data","value":"ChocoPhlAn and translated protein-search databases are analysis resources, not a universal neural training corpus.","status":"source_checked","source_ids":["evidence-official-10f58e320879def8b102"],"source_locator":"readme.md: description, Main workflow, Download the databases and normalization documentation"},{"label":"Training cutoff","value":"Inapplicable to neural pretraining; reference-database and input-data dates must be recorded for each run.","status":"inapplicable","source_ids":["evidence-official-10f58e320879def8b102"],"source_locator":"readme.md: description, Main workflow, Download the databases and normalization documentation"},{"label":"Context limits","value":"Read files and search-database size; no fixed model context.","status":"source_checked","source_ids":["evidence-official-10f58e320879def8b102"],"source_locator":"readme.md: description, Main workflow, Download the databases and normalization documentation"},{"label":"Weights licence","value":"Inapplicable: no neural checkpoint for the core profiling pipeline.","status":"inapplicable","source_ids":["evidence-official-10f58e320879def8b102"],"source_locator":"readme.md: description, Main workflow, Download the databases and normalization documentation"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/biobakery/humann","status":"source_checked","source_ids":["evidence-official-10f58e320879def8b102"],"source_locator":"readme.md: description, Main workflow, Download the databases and normalization documentation"},{"label":"Code licence","value":"MIT","status":"source_checked","source_ids":["evidence-official-f29710879cf7a4438b12"],"source_locator":"LICENSE: licence text"}],"strengths":[{"text":"Connects community sequencing data to interpretable functional and pathway profiles.","source_ids":["evidence-official-10f58e320879def8b102"],"source_locator":"readme.md: description, Main workflow, Download the databases and normalization documentation"}],"limitations":[{"text":"Reference coverage and normalization affect interpretation. The manual clarifies that CPM means copies per million rather than unnormalized counts per million.","source_ids":["evidence-official-10f58e320879def8b102"],"source_locator":"readme.md: description, Main workflow, Download the databases and normalization documentation"}],"diagram":{"title":"HUMAnN workflow","steps":["Community reads","Reference searches","Gene-family quantification","Pathway profiles"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-10f58e320879def8b102"],"source_locator":"readme.md: description, Main workflow, Download the databases and normalization documentation"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}},"description":"Microbial functional profiling","facets":{"areas":["microbiome"]},"id":"discovery-model-humann","kind":"model","links":[],"name":"HUMAnN","source_ids":["src-discovery-biobakery-humann"],"status":"discovered"}
{"attributes":{"access":"official_source_linked","benchmark_applicability":"candidate; not evidence of a reported evaluation","candidate_benchmark_ids":["discovery-benchmark-cami"],"entity_level":"method","reported_name":"Kraken 2","version":null,"historical_missing_metadata":{"checkpoint":"unextracted","code_licence":"unextracted","parameters":"unextracted","training_cutoff":"unextracted","training_data":"unextracted","version":"unextracted","weights_licence":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Kraken 2 assigns taxonomic labels to sequence reads by consulting a reference-derived minimizer database.","summary_source_ids":["evidence-official-40f46dc4b9fcc306773b","evidence-official-33fe0374c8ebbc2c7294","evidence-official-9ba2b6dc494c7b6df951","evidence-official-1f756fefa823dd402b0f"],"summary_source_locator":"docs/MANUAL.markdown: Introduction, Standard Kraken 2 Database, Classification and Confidence scoring","sections":[{"title":"How it works","body":"Kraken 2 breaks query sequences into k-mers and looks up selected minimizers in a compact hash table. Each stored minimizer is associated with a lowest-common-ancestor taxonomic label. The classifier combines that evidence to assign a taxon; the reference database and confidence settings are therefore part of the evaluated procedure.","source_ids":["evidence-official-40f46dc4b9fcc306773b","evidence-official-33fe0374c8ebbc2c7294","evidence-official-9ba2b6dc494c7b6df951","evidence-official-1f756fefa823dd402b0f"],"source_locator":"docs/MANUAL.markdown: Introduction, Standard Kraken 2 Database, Classification and Confidence scoring"},{"title":"Versions and reproducibility","body":"Kraken 2 is a rewrite of Kraken 1 and is not backwards compatible. Read/contig input, not a learned fixed token window.","source_ids":["evidence-official-40f46dc4b9fcc306773b","evidence-official-33fe0374c8ebbc2c7294","evidence-official-9ba2b6dc494c7b6df951","evidence-official-1f756fefa823dd402b0f"],"source_locator":"docs/MANUAL.markdown: Introduction, Standard Kraken 2 Database, Classification and Confidence scoring"}],"facts":[{"label":"Model type","value":"Minimizer-based taxonomic classifier","status":"source_checked","source_ids":["evidence-official-40f46dc4b9fcc306773b","evidence-official-33fe0374c8ebbc2c7294","evidence-official-9ba2b6dc494c7b6df951","evidence-official-1f756fefa823dd402b0f"],"source_locator":"docs/MANUAL.markdown: Introduction, Standard Kraken 2 Database, Classification and Confidence scoring"},{"label":"Architecture","value":"Minimizer-based sequence classification using a compact hash table and lowest-common-ancestor taxonomy assignments.","status":"source_checked","source_ids":["evidence-official-40f46dc4b9fcc306773b","evidence-official-33fe0374c8ebbc2c7294","evidence-official-9ba2b6dc494c7b6df951","evidence-official-1f756fefa823dd402b0f"],"source_locator":"docs/MANUAL.markdown: Introduction, Standard Kraken 2 Database, Classification and Confidence scoring"},{"label":"Inputs","value":"DNA reads or, in translated-search mode, sequences searched against an amino-acid database.","status":"source_checked","source_ids":["evidence-official-40f46dc4b9fcc306773b","evidence-official-33fe0374c8ebbc2c7294","evidence-official-9ba2b6dc494c7b6df951","evidence-official-1f756fefa823dd402b0f"],"source_locator":"docs/MANUAL.markdown: Introduction, Standard Kraken 2 Database, Classification and Confidence scoring"},{"label":"Outputs","value":"Per-read taxonomic assignments and aggregate classification reports.","status":"source_checked","source_ids":["evidence-official-40f46dc4b9fcc306773b","evidence-official-33fe0374c8ebbc2c7294","evidence-official-9ba2b6dc494c7b6df951","evidence-official-1f756fefa823dd402b0f"],"source_locator":"docs/MANUAL.markdown: Introduction, Standard Kraken 2 Database, Classification and Confidence scoring"},{"label":"Parameters","value":"Inapplicable as a neural parameter total; k-mer/minimizer length, confidence and database choices are algorithm settings.","status":"inapplicable","source_ids":["evidence-official-40f46dc4b9fcc306773b","evidence-official-33fe0374c8ebbc2c7294","evidence-official-9ba2b6dc494c7b6df951","evidence-official-1f756fefa823dd402b0f"],"source_locator":"docs/MANUAL.markdown: Introduction, Standard Kraken 2 Database, Classification and Confidence scoring"},{"label":"Known versions","value":"Kraken 2 is a rewrite of Kraken 1 and is not backwards compatible.","status":"source_checked","source_ids":["evidence-official-40f46dc4b9fcc306773b","evidence-official-33fe0374c8ebbc2c7294","evidence-official-9ba2b6dc494c7b6df951","evidence-official-1f756fefa823dd402b0f"],"source_locator":"docs/MANUAL.markdown: Introduction, Standard Kraken 2 Database, Classification and Confidence scoring"},{"label":"Training data","value":"Not neural pretraining: build a database from selected reference sequences and taxonomy.","status":"source_checked","source_ids":["evidence-official-40f46dc4b9fcc306773b","evidence-official-33fe0374c8ebbc2c7294","evidence-official-9ba2b6dc494c7b6df951","evidence-official-1f756fefa823dd402b0f"],"source_locator":"docs/MANUAL.markdown: Introduction, Standard Kraken 2 Database, Classification and Confidence scoring"},{"label":"Training cutoff","value":"Inapplicable to neural pretraining; reference-database and input-data dates must be recorded for each run.","status":"inapplicable","source_ids":["evidence-official-40f46dc4b9fcc306773b","evidence-official-33fe0374c8ebbc2c7294","evidence-official-9ba2b6dc494c7b6df951","evidence-official-1f756fefa823dd402b0f"],"source_locator":"docs/MANUAL.markdown: Introduction, Standard Kraken 2 Database, Classification and Confidence scoring"},{"label":"Context limits","value":"Read/contig input, not a learned fixed token window.","status":"source_checked","source_ids":["evidence-official-40f46dc4b9fcc306773b","evidence-official-33fe0374c8ebbc2c7294","evidence-official-9ba2b6dc494c7b6df951","evidence-official-1f756fefa823dd402b0f"],"source_locator":"docs/MANUAL.markdown: Introduction, Standard Kraken 2 Database, Classification and Confidence scoring"},{"label":"Weights licence","value":"Inapplicable to this classifier: database contents and their licences replace neural weights.","status":"inapplicable","source_ids":["evidence-official-40f46dc4b9fcc306773b","evidence-official-33fe0374c8ebbc2c7294","evidence-official-9ba2b6dc494c7b6df951","evidence-official-1f756fefa823dd402b0f"],"source_locator":"docs/MANUAL.markdown: Introduction, Standard Kraken 2 Database, Classification and Confidence scoring"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/DerrickWood/kraken2","status":"source_checked","source_ids":["evidence-official-40f46dc4b9fcc306773b","evidence-official-33fe0374c8ebbc2c7294","evidence-official-9ba2b6dc494c7b6df951","evidence-official-1f756fefa823dd402b0f"],"source_locator":"docs/MANUAL.markdown: Introduction, Standard Kraken 2 Database, Classification and Confidence scoring"},{"label":"Code licence","value":"MIT","status":"source_checked","source_ids":["evidence-official-bf1a5e03cd84873f4b04"],"source_locator":"LICENSE: licence text"}],"strengths":[{"text":"A reference-based procedural comparator with explicit database construction and confidence settings.","source_ids":["evidence-official-40f46dc4b9fcc306773b","evidence-official-33fe0374c8ebbc2c7294","evidence-official-9ba2b6dc494c7b6df951","evidence-official-1f756fefa823dd402b0f"],"source_locator":"docs/MANUAL.markdown: Introduction, Standard Kraken 2 Database, Classification and Confidence scoring"}],"limitations":[{"text":"Classification depends on the reference database, taxonomy version and minimizer configuration. Compact hashing can introduce false matches; software version alone does not identify a reproducible classifier.","source_ids":["evidence-official-40f46dc4b9fcc306773b","evidence-official-33fe0374c8ebbc2c7294","evidence-official-9ba2b6dc494c7b6df951","evidence-official-1f756fefa823dd402b0f"],"source_locator":"docs/MANUAL.markdown: Introduction, Standard Kraken 2 Database, Classification and Confidence scoring"}],"diagram":{"title":"Kraken 2 workflow","steps":["Reference genomes and taxonomy","Minimizer database","Read minimizer lookup","Taxonomic assignment"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-40f46dc4b9fcc306773b","evidence-official-33fe0374c8ebbc2c7294","evidence-official-9ba2b6dc494c7b6df951","evidence-official-1f756fefa823dd402b0f"],"source_locator":"docs/MANUAL.markdown: Introduction, Standard Kraken 2 Database, Classification and Confidence scoring"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}},"description":"Sequence-based metagenomic classification","facets":{"areas":["microbiome"]},"id":"discovery-model-kraken-2","kind":"model","links":[{"relation":"applicable_to","target_id":"discovery-benchmark-cami"}],"name":"Kraken 2","source_ids":["src-discovery-derrickwood-kraken2"],"status":"discovered"}
{"attributes":{"access":"official_source_linked","benchmark_applicability":"candidate; not evidence of a reported evaluation","candidate_benchmark_ids":[],"entity_level":"method","reported_name":"LipidBlast","version":null,"historical_missing_metadata":{"checkpoint":"unextracted","code_licence":"unextracted","parameters":"unextracted","training_cutoff":"unextracted","training_data":"unextracted","version":"unextracted","weights_licence":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"LipidBlast is an in-silico tandem mass-spectral library used to annotate lipids.","summary_source_ids":["evidence-official-fadad0bf451a225696d1"],"summary_source_locator":"Official LipidBlast page: Short Introduction, Download, Picture service and FAQ","sections":[{"title":"How it works","body":"LipidBlast is an in-silico tandem mass-spectral library used to annotate lipids. Rule/template-generated lipid fragmentation library queried by a compatible mass-spectral search engine. The documented inputs are LC-MS/MS spectra or, in the documented accurate-mass-only workflow, precursor masses. The output consists of candidate lipid annotations from spectral or mass matching.","source_ids":["evidence-official-fadad0bf451a225696d1"],"source_locator":"Official LipidBlast page: Short Introduction, Download, Picture service and FAQ"},{"title":"Versions and reproducibility","body":"Original LipidBlast library and later MS-DIAL integration are distinct releases. Mass-spectral peaks and library/search settings; no neural token window.","source_ids":["evidence-official-fadad0bf451a225696d1"],"source_locator":"Official LipidBlast page: Short Introduction, Download, Picture service and FAQ"}],"facts":[{"label":"Model type","value":"Rule-based in-silico lipid spectral library","status":"source_checked","source_ids":["evidence-official-fadad0bf451a225696d1"],"source_locator":"Official LipidBlast page: Short Introduction, Download, Picture service and FAQ"},{"label":"Architecture","value":"Rule/template-generated lipid fragmentation library queried by a compatible mass-spectral search engine.","status":"source_checked","source_ids":["evidence-official-fadad0bf451a225696d1"],"source_locator":"Official LipidBlast page: Short Introduction, Download, Picture service and FAQ"},{"label":"Inputs","value":"LC-MS/MS spectra or, in the documented accurate-mass-only workflow, precursor masses.","status":"source_checked","source_ids":["evidence-official-fadad0bf451a225696d1"],"source_locator":"Official LipidBlast page: Short Introduction, Download, Picture service and FAQ"},{"label":"Outputs","value":"Candidate lipid annotations from spectral or mass matching.","status":"source_checked","source_ids":["evidence-official-fadad0bf451a225696d1"],"source_locator":"Official LipidBlast page: Short Introduction, Download, Picture service and FAQ"},{"label":"Parameters","value":"Inapplicable as a neural parameter count.","status":"inapplicable","source_ids":["evidence-official-fadad0bf451a225696d1"],"source_locator":"Official LipidBlast page: Short Introduction, Download, Picture service and FAQ"},{"label":"Known versions","value":"Original LipidBlast library and later MS-DIAL integration are distinct releases.","status":"source_checked","source_ids":["evidence-official-fadad0bf451a225696d1"],"source_locator":"Official LipidBlast page: Short Introduction, Download, Picture service and FAQ"},{"label":"Training data","value":"Expert fragmentation templates and reference evidence, not neural pretraining.","status":"source_checked","source_ids":["evidence-official-fadad0bf451a225696d1"],"source_locator":"Official LipidBlast page: Short Introduction, Download, Picture service and FAQ"},{"label":"Training cutoff","value":"Inapplicable to neural pretraining; reference-database and input-data dates must be recorded for each run.","status":"inapplicable","source_ids":["evidence-official-fadad0bf451a225696d1"],"source_locator":"Official LipidBlast page: Short Introduction, Download, Picture service and FAQ"},{"label":"Context limits","value":"Mass-spectral peaks and library/search settings; no neural token window.","status":"source_checked","source_ids":["evidence-official-fadad0bf451a225696d1"],"source_locator":"Official LipidBlast page: Short Introduction, Download, Picture service and FAQ"},{"label":"Weights licence","value":"Inapplicable: library and software terms must be checked separately. The page only explicitly grants CC-BY for parts of the publication software supplement.","status":"inapplicable","source_ids":["evidence-official-fadad0bf451a225696d1"],"source_locator":"Official LipidBlast page: Short Introduction, Download, Picture service and FAQ"},{"label":"Access","value":"Official project documentation and implementation: https://fiehnlab.ucdavis.edu/projects/LipidBlast/","status":"source_checked","source_ids":["evidence-official-fadad0bf451a225696d1"],"source_locator":"Official LipidBlast page: Short Introduction, Download, Picture service and FAQ"},{"label":"Code licence","value":"The inspected LipidBlast project page does not establish separate software/library redistribution terms. The accompanying article licence is not substituted for those terms.","status":"unreported","source_ids":["evidence-official-fadad0bf451a225696d1"],"source_locator":"page.html: inspected official source"}],"strengths":[{"text":"Supports library search across multiple instrument types and provides inspectable reference spectra/templates.","source_ids":["evidence-official-fadad0bf451a225696d1"],"source_locator":"Official LipidBlast page: Short Introduction, Download, Picture service and FAQ"}],"limitations":[{"text":"The project notes missing lipid classes and ambiguous accurate-mass matches. Identification by mass alone cannot resolve isobars; it is not intended for GC-MS.","source_ids":["evidence-official-fadad0bf451a225696d1"],"source_locator":"Official LipidBlast page: Short Introduction, Download, Picture service and FAQ"}],"diagram":{"title":"LipidBlast workflow","steps":["Lipid fragmentation templates","In-silico library","Experimental-spectrum search","Candidate lipid annotation"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-fadad0bf451a225696d1"],"source_locator":"Official LipidBlast page: Short Introduction, Download, Picture service and FAQ"},"coverage":"limited","gaps":["Code licence: The inspected LipidBlast project page does not establish separate software/library redistribution terms. The accompanying article licence is not substituted for those terms."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}},"description":"Rule-based lipid fragmentation library matching","facets":{"areas":["lipidomics"]},"id":"discovery-model-lipidblast","kind":"model","links":[],"name":"LipidBlast","source_ids":["src-discovery-lipidblast"],"status":"discovered"}
{"attributes":{"access":"official_source_linked","benchmark_applicability":"candidate; not evidence of a reported evaluation","candidate_benchmark_ids":[],"entity_level":"method","reported_name":"LipidFinder","version":null,"historical_missing_metadata":{"checkpoint":"unextracted","code_licence":"unextracted","parameters":"unextracted","training_cutoff":"unextracted","training_data":"unextracted","version":"unextracted","weights_licence":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"LipidFinder filters aligned LC-MS features and assigns putative lipid identities using reference databases.","summary_source_ids":["evidence-official-95a620349435ae6651c5","evidence-official-3759a2f34cb49263eb43"],"summary_source_locator":"Official LipidFinder page: workflow description, applicability warning and source-code link","sections":[{"title":"How it works","body":"LipidFinder filters aligned LC-MS features and assigns putative lipid identities using reference databases. Configurable filtering and database-search workflow over high-resolution LC-MS features. The documented inputs are pre-aligned high-resolution LC-MS measurements, including chromatographic information. The output consists of filtered lipid-like features and putative lipid classifications.","source_ids":["evidence-official-95a620349435ae6651c5","evidence-official-3759a2f34cb49263eb43"],"source_locator":"Official LipidFinder page: workflow description, applicability warning and source-code link"},{"title":"Versions and reproducibility","body":"LipidFinder original workflow and LipidFinder 2.0 are distinguished in the official page citations. Chromatographic features and mass-resolution constraints, not sequence tokens.","source_ids":["evidence-official-95a620349435ae6651c5","evidence-official-3759a2f34cb49263eb43"],"source_locator":"Official LipidFinder page: workflow description, applicability warning and source-code link"}],"facts":[{"label":"Model type","value":"Lipidomics preprocessing and identification workflow","status":"source_checked","source_ids":["evidence-official-95a620349435ae6651c5","evidence-official-3759a2f34cb49263eb43"],"source_locator":"Official LipidFinder page: workflow description, applicability warning and source-code link"},{"label":"Architecture","value":"Configurable filtering and database-search workflow over high-resolution LC-MS features.","status":"source_checked","source_ids":["evidence-official-95a620349435ae6651c5","evidence-official-3759a2f34cb49263eb43"],"source_locator":"Official LipidFinder page: workflow description, applicability warning and source-code link"},{"label":"Inputs","value":"Pre-aligned high-resolution LC-MS measurements, including chromatographic information.","status":"source_checked","source_ids":["evidence-official-95a620349435ae6651c5","evidence-official-3759a2f34cb49263eb43"],"source_locator":"Official LipidFinder page: workflow description, applicability warning and source-code link"},{"label":"Outputs","value":"Filtered lipid-like features and putative lipid classifications.","status":"source_checked","source_ids":["evidence-official-95a620349435ae6651c5","evidence-official-3759a2f34cb49263eb43"],"source_locator":"Official LipidFinder page: workflow description, applicability warning and source-code link"},{"label":"Parameters","value":"Inapplicable as a neural parameter count; filtering tolerances and database choices are configuration.","status":"inapplicable","source_ids":["evidence-official-95a620349435ae6651c5","evidence-official-3759a2f34cb49263eb43"],"source_locator":"Official LipidFinder page: workflow description, applicability warning and source-code link"},{"label":"Known versions","value":"LipidFinder original workflow and LipidFinder 2.0 are distinguished in the official page citations.","status":"source_checked","source_ids":["evidence-official-95a620349435ae6651c5","evidence-official-3759a2f34cb49263eb43"],"source_locator":"Official LipidFinder page: workflow description, applicability warning and source-code link"},{"label":"Training data","value":"No neural pretraining; reference databases and user settings define the workflow.","status":"source_checked","source_ids":["evidence-official-95a620349435ae6651c5","evidence-official-3759a2f34cb49263eb43"],"source_locator":"Official LipidFinder page: workflow description, applicability warning and source-code link"},{"label":"Training cutoff","value":"Inapplicable to neural pretraining; reference-database and input-data dates must be recorded for each run.","status":"inapplicable","source_ids":["evidence-official-95a620349435ae6651c5","evidence-official-3759a2f34cb49263eb43"],"source_locator":"Official LipidFinder page: workflow description, applicability warning and source-code link"},{"label":"Context limits","value":"Chromatographic features and mass-resolution constraints, not sequence tokens.","status":"source_checked","source_ids":["evidence-official-95a620349435ae6651c5","evidence-official-3759a2f34cb49263eb43"],"source_locator":"Official LipidFinder page: workflow description, applicability warning and source-code link"},{"label":"Weights licence","value":"Inapplicable: no neural model checkpoint.","status":"inapplicable","source_ids":["evidence-official-95a620349435ae6651c5","evidence-official-3759a2f34cb49263eb43"],"source_locator":"Official LipidFinder page: workflow description, applicability warning and source-code link"},{"label":"Access","value":"Official project documentation and implementation: https://www.lipidmaps.org/resources/tools/lipidfinder/","status":"source_checked","source_ids":["evidence-official-95a620349435ae6651c5","evidence-official-3759a2f34cb49263eb43"],"source_locator":"Official LipidFinder page: workflow description, applicability warning and source-code link"},{"label":"Code licence","value":"MIT","status":"source_checked","source_ids":["evidence-official-5631fb75aa8e5a7a89e1"],"source_locator":"LICENSE: licence text"}],"strengths":[{"text":"Separates likely lipids from contaminants, adducts and noise before database annotation.","source_ids":["evidence-official-95a620349435ae6651c5","evidence-official-3759a2f34cb49263eb43"],"source_locator":"Official LipidFinder page: workflow description, applicability warning and source-code link"}],"limitations":[{"text":"The official page explicitly excludes shotgun lipidomics, MS/MS and low-resolution data from the described workflow. Putative annotations need separate identification evidence.","source_ids":["evidence-official-95a620349435ae6651c5","evidence-official-3759a2f34cb49263eb43"],"source_locator":"Official LipidFinder page: workflow description, applicability warning and source-code link"}],"diagram":{"title":"LipidFinder workflow","steps":["Aligned LC-MS features","Filtering and contaminant handling","Database search","Putative lipid classes"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-95a620349435ae6651c5","evidence-official-3759a2f34cb49263eb43"],"source_locator":"Official LipidFinder page: workflow description, applicability warning and source-code link"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}},"description":"LC-MS lipid feature filtering and annotation","facets":{"areas":["lipidomics"]},"id":"discovery-model-lipidfinder","kind":"model","links":[],"name":"LipidFinder","source_ids":["src-discovery-lipidfinder"],"status":"discovered"}
{"attributes":{"access":"official_source_linked","benchmark_applicability":"candidate; not evidence of a reported evaluation","candidate_benchmark_ids":["discovery-benchmark-massspecgym"],"entity_level":"method","reported_name":"matchms","version":null,"historical_missing_metadata":{"checkpoint":"unextracted","code_licence":"unextracted","parameters":"unextracted","training_cutoff":"unextracted","training_data":"unextracted","version":"unextracted","weights_licence":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"matchms provides reproducible processing and comparison of tandem mass spectra.","summary_source_ids":["evidence-official-e508b40498c3a3ba53d5"],"summary_source_locator":"README.rst: introductory description and License","sections":[{"title":"How it works","body":"matchms provides reproducible processing and comparison of tandem mass spectra. Modular import, metadata/peak processing and pairwise-similarity pipeline; classical cosine and external learned measures are distinct options. The documented inputs are MS/MS spectra in supported formats such as mzML, mzXML, MSP, MGF and JSON. The output consists of processed spectra and pairwise similarity scores.","source_ids":["evidence-official-e508b40498c3a3ba53d5"],"source_locator":"README.rst: introductory description and License"},{"title":"Versions and reproducibility","body":"matchms version plus the exact filter/similarity pipeline. Spectral peaks and workflow memory requirements; no universal token context.","source_ids":["evidence-official-e508b40498c3a3ba53d5"],"source_locator":"README.rst: introductory description and License"}],"facts":[{"label":"Model type","value":"Mass-spectral processing and similarity software","status":"source_checked","source_ids":["evidence-official-e508b40498c3a3ba53d5"],"source_locator":"README.rst: introductory description and License"},{"label":"Architecture","value":"Modular import, metadata/peak processing and pairwise-similarity pipeline; classical cosine and external learned measures are distinct options.","status":"source_checked","source_ids":["evidence-official-e508b40498c3a3ba53d5"],"source_locator":"README.rst: introductory description and License"},{"label":"Inputs","value":"MS/MS spectra in supported formats such as mzML, mzXML, MSP, MGF and JSON.","status":"source_checked","source_ids":["evidence-official-e508b40498c3a3ba53d5"],"source_locator":"README.rst: introductory description and License"},{"label":"Outputs","value":"Processed spectra and pairwise similarity scores.","status":"source_checked","source_ids":["evidence-official-e508b40498c3a3ba53d5"],"source_locator":"README.rst: introductory description and License"},{"label":"Parameters","value":"Inapplicable for the framework; external learned similarity plugins have separate parameters.","status":"inapplicable","source_ids":["evidence-official-e508b40498c3a3ba53d5"],"source_locator":"README.rst: introductory description and License"},{"label":"Known versions","value":"matchms version plus the exact filter/similarity pipeline.","status":"source_checked","source_ids":["evidence-official-e508b40498c3a3ba53d5"],"source_locator":"README.rst: introductory description and License"},{"label":"Training data","value":"No universal pretraining; selected learned plugins must provide their own training provenance.","status":"source_checked","source_ids":["evidence-official-e508b40498c3a3ba53d5"],"source_locator":"README.rst: introductory description and License"},{"label":"Training cutoff","value":"Inapplicable to neural pretraining; reference-database and input-data dates must be recorded for each run.","status":"inapplicable","source_ids":["evidence-official-e508b40498c3a3ba53d5"],"source_locator":"README.rst: introductory description and License"},{"label":"Context limits","value":"Spectral peaks and workflow memory requirements; no universal token context.","status":"source_checked","source_ids":["evidence-official-e508b40498c3a3ba53d5"],"source_locator":"README.rst: introductory description and License"},{"label":"Weights licence","value":"Inapplicable to core classical processing; external learned plugins have separate weight terms.","status":"inapplicable","source_ids":["evidence-official-e508b40498c3a3ba53d5"],"source_locator":"README.rst: introductory description and License"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/matchms/matchms","status":"source_checked","source_ids":["evidence-official-e508b40498c3a3ba53d5"],"source_locator":"README.rst: introductory description and License"},{"label":"Code licence","value":"Apache-2.0","status":"source_checked","source_ids":["evidence-official-71e8ebc066164f0ce44e"],"source_locator":"LICENSE: licence text"}],"strengths":[{"text":"Separates preprocessing from the similarity measure and supports custom metrics and sparse result storage.","source_ids":["evidence-official-e508b40498c3a3ba53d5"],"source_locator":"README.rst: introductory description and License"}],"limitations":[{"text":"matchms is a framework, not one universal scoring model. Filter order, thresholds and selected similarity function must be recorded.","source_ids":["evidence-official-e508b40498c3a3ba53d5"],"source_locator":"README.rst: introductory description and License"}],"diagram":{"title":"matchms workflow","steps":["Imported MS/MS spectra","Metadata and peak processing","Selected similarity function","Similarity matrix"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-e508b40498c3a3ba53d5"],"source_locator":"README.rst: introductory description and License"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}},"description":"Mass spectral processing and similarity matching","facets":{"areas":["metabolomics"]},"id":"discovery-model-matchms","kind":"model","links":[{"relation":"applicable_to","target_id":"discovery-benchmark-massspecgym"}],"name":"matchms","source_ids":["src-discovery-matchms-matchms"],"status":"discovered"}
{"attributes":{"access":"official_source_linked","benchmark_applicability":"candidate; not evidence of a reported evaluation","candidate_benchmark_ids":["discovery-benchmark-cami"],"entity_level":"method","reported_name":"MetaPhlAn","version":null,"historical_missing_metadata":{"checkpoint":"unextracted","code_licence":"unextracted","parameters":"unextracted","training_cutoff":"unextracted","training_data":"unextracted","version":"unextracted","weights_licence":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"MetaPhlAn profiles microbial community composition from shotgun metagenomic reads using clade-specific marker genes.","summary_source_ids":["evidence-official-72fdaed158e4ff851fbe","evidence-official-37f825d1226e59cbbfbf"],"summary_source_locator":"README.md: description; MetaPhlAn 4 paper: Building the expanded SGB catalog and Discussion","sections":[{"title":"How it works","body":"MetaPhlAn profiles microbial community composition from shotgun metagenomic reads using clade-specific marker genes. Reference marker-gene profiling; MetaPhlAn 4 organizes reference and metagenome-assembled genomes into species-level genome bins. The documented inputs are shotgun metagenomic reads and a selected MetaPhlAn marker database. The output consists of taxonomic relative-abundance profiles; StrainPhlAn is a separate strain-level analysis.","source_ids":["evidence-official-72fdaed158e4ff851fbe","evidence-official-37f825d1226e59cbbfbf"],"source_locator":"README.md: description; MetaPhlAn 4 paper: Building the expanded SGB catalog and Discussion"},{"title":"Versions and reproducibility","body":"MetaPhlAn 4 paper and 4.2-linked current documentation; database version is a separate reproducibility requirement. Shotgun reads; no fixed neural token context.","source_ids":["evidence-official-72fdaed158e4ff851fbe","evidence-official-37f825d1226e59cbbfbf"],"source_locator":"README.md: description; MetaPhlAn 4 paper: Building the expanded SGB catalog and Discussion"}],"facts":[{"label":"Model type","value":"Marker-based taxonomic profiling","status":"source_checked","source_ids":["evidence-official-72fdaed158e4ff851fbe","evidence-official-37f825d1226e59cbbfbf"],"source_locator":"README.md: description; MetaPhlAn 4 paper: Building the expanded SGB catalog and Discussion"},{"label":"Architecture","value":"Reference marker-gene profiling; MetaPhlAn 4 organizes reference and metagenome-assembled genomes into species-level genome bins.","status":"source_checked","source_ids":["evidence-official-72fdaed158e4ff851fbe","evidence-official-37f825d1226e59cbbfbf"],"source_locator":"README.md: description; MetaPhlAn 4 paper: Building the expanded SGB catalog and Discussion"},{"label":"Inputs","value":"Shotgun metagenomic reads and a selected MetaPhlAn marker database.","status":"source_checked","source_ids":["evidence-official-72fdaed158e4ff851fbe","evidence-official-37f825d1226e59cbbfbf"],"source_locator":"README.md: description; MetaPhlAn 4 paper: Building the expanded SGB catalog and Discussion"},{"label":"Outputs","value":"Taxonomic relative-abundance profiles; StrainPhlAn is a separate strain-level analysis.","status":"source_checked","source_ids":["evidence-official-72fdaed158e4ff851fbe","evidence-official-37f825d1226e59cbbfbf"],"source_locator":"README.md: description; MetaPhlAn 4 paper: Building the expanded SGB catalog and Discussion"},{"label":"Parameters","value":"Inapplicable as a neural parameter count.","status":"inapplicable","source_ids":["evidence-official-72fdaed158e4ff851fbe","evidence-official-37f825d1226e59cbbfbf"],"source_locator":"README.md: description; MetaPhlAn 4 paper: Building the expanded SGB catalog and Discussion"},{"label":"Known versions","value":"MetaPhlAn 4 paper and 4.2-linked current documentation; database version is a separate reproducibility requirement.","status":"source_checked","source_ids":["evidence-official-72fdaed158e4ff851fbe","evidence-official-37f825d1226e59cbbfbf"],"source_locator":"README.md: description; MetaPhlAn 4 paper: Building the expanded SGB catalog and Discussion"},{"label":"Training data","value":"Reference-derived marker database rather than neural pretraining; record the exact database release.","status":"source_checked","source_ids":["evidence-official-72fdaed158e4ff851fbe","evidence-official-37f825d1226e59cbbfbf"],"source_locator":"README.md: description; MetaPhlAn 4 paper: Building the expanded SGB catalog and Discussion"},{"label":"Training cutoff","value":"Inapplicable to neural pretraining; reference-database and input-data dates must be recorded for each run.","status":"inapplicable","source_ids":["evidence-official-72fdaed158e4ff851fbe","evidence-official-37f825d1226e59cbbfbf"],"source_locator":"README.md: description; MetaPhlAn 4 paper: Building the expanded SGB catalog and Discussion"},{"label":"Context limits","value":"Shotgun reads; no fixed neural token context.","status":"source_checked","source_ids":["evidence-official-72fdaed158e4ff851fbe","evidence-official-37f825d1226e59cbbfbf"],"source_locator":"README.md: description; MetaPhlAn 4 paper: Building the expanded SGB catalog and Discussion"},{"label":"Weights licence","value":"Inapplicable to this procedural method; marker databases have their own provenance and terms.","status":"inapplicable","source_ids":["evidence-official-72fdaed158e4ff851fbe","evidence-official-37f825d1226e59cbbfbf"],"source_locator":"README.md: description; MetaPhlAn 4 paper: Building the expanded SGB catalog and Discussion"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/biobakery/MetaPhlAn","status":"source_checked","source_ids":["evidence-official-72fdaed158e4ff851fbe","evidence-official-37f825d1226e59cbbfbf"],"source_locator":"README.md: description; MetaPhlAn 4 paper: Building the expanded SGB catalog and Discussion"},{"label":"Code licence","value":"MIT","status":"source_checked","source_ids":["evidence-official-1a0775cac85be75449ef"],"source_locator":"license.txt: licence text"}],"strengths":[{"text":"Marker-based profiling can incorporate characterized and previously uncharacterized species groups.","source_ids":["evidence-official-72fdaed158e4ff851fbe","evidence-official-37f825d1226e59cbbfbf"],"source_locator":"README.md: description; MetaPhlAn 4 paper: Building the expanded SGB catalog and Discussion"}],"limitations":[{"text":"Coverage depends on the marker database and habitat. The MetaPhlAn 4 paper identifies remaining gaps for under-studied environmental communities; newer software/databases may have different scope.","source_ids":["evidence-official-72fdaed158e4ff851fbe","evidence-official-37f825d1226e59cbbfbf"],"source_locator":"README.md: description; MetaPhlAn 4 paper: Building the expanded SGB catalog and Discussion"}],"diagram":{"title":"MetaPhlAn workflow","steps":["Metagenomic reads","Marker-gene mapping","Species-group quantification","Relative-abundance profile"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-72fdaed158e4ff851fbe","evidence-official-37f825d1226e59cbbfbf"],"source_locator":"README.md: description; MetaPhlAn 4 paper: Building the expanded SGB catalog and Discussion"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}},"description":"Marker-based microbial profiling","facets":{"areas":["microbiome"]},"id":"discovery-model-metaphlan","kind":"model","links":[{"relation":"applicable_to","target_id":"discovery-benchmark-cami"}],"name":"MetaPhlAn","source_ids":["src-discovery-biobakery-metaphlan"],"status":"discovered"}
{"attributes":{"access":"official_source_linked","benchmark_applicability":"candidate; not evidence of a reported evaluation","candidate_benchmark_ids":["discovery-benchmark-cafa"],"entity_level":"method","reported_name":"MMseqs2","version":null,"historical_missing_metadata":{"checkpoint":"unextracted","code_licence":"unextracted","parameters":"unextracted","training_cutoff":"unextracted","training_data":"unextracted","version":"unextracted","weights_licence":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"MMseqs2 searches and clusters large protein and nucleotide sequence collections.","summary_source_ids":["evidence-official-81e51077d7d3352a6de4"],"summary_source_locator":"README.md: opening, Publications, Installation and user-guide pointers","sections":[{"title":"How it works","body":"MMseqs2 searches and clusters large protein and nucleotide sequence collections. Sequence/profile search and clustering software with CPU and selected GPU execution paths. The documented inputs are protein or nucleotide query sequences and a reference database, or sequences to cluster. The output consists of sequence hits, alignments, clusters or configured taxonomic assignments.","source_ids":["evidence-official-81e51077d7d3352a6de4"],"source_locator":"README.md: opening, Publications, Installation and user-guide pointers"},{"title":"Versions and reproducibility","body":"MMseqs2 software release and database build must both be pinned; GPU capability differs by build. Database/query scale and implementation limits, not a neural token window.","source_ids":["evidence-official-81e51077d7d3352a6de4"],"source_locator":"README.md: opening, Publications, Installation and user-guide pointers"}],"facts":[{"label":"Model type","value":"Sequence search and clustering software","status":"source_checked","source_ids":["evidence-official-81e51077d7d3352a6de4"],"source_locator":"README.md: opening, Publications, Installation and user-guide pointers"},{"label":"Architecture","value":"Sequence/profile search and clustering software with CPU and selected GPU execution paths.","status":"source_checked","source_ids":["evidence-official-81e51077d7d3352a6de4"],"source_locator":"README.md: opening, Publications, Installation and user-guide pointers"},{"label":"Inputs","value":"Protein or nucleotide query sequences and a reference database, or sequences to cluster.","status":"source_checked","source_ids":["evidence-official-81e51077d7d3352a6de4"],"source_locator":"README.md: opening, Publications, Installation and user-guide pointers"},{"label":"Outputs","value":"Sequence hits, alignments, clusters or configured taxonomic assignments.","status":"source_checked","source_ids":["evidence-official-81e51077d7d3352a6de4"],"source_locator":"README.md: opening, Publications, Installation and user-guide pointers"},{"label":"Parameters","value":"Inapplicable as a neural parameter count; search and clustering settings apply.","status":"inapplicable","source_ids":["evidence-official-81e51077d7d3352a6de4"],"source_locator":"README.md: opening, Publications, Installation and user-guide pointers"},{"label":"Known versions","value":"MMseqs2 software release and database build must both be pinned; GPU capability differs by build.","status":"source_checked","source_ids":["evidence-official-81e51077d7d3352a6de4"],"source_locator":"README.md: opening, Publications, Installation and user-guide pointers"},{"label":"Training data","value":"Reference sequence/profile databases rather than neural pretraining.","status":"source_checked","source_ids":["evidence-official-81e51077d7d3352a6de4"],"source_locator":"README.md: opening, Publications, Installation and user-guide pointers"},{"label":"Training cutoff","value":"Inapplicable to neural pretraining; reference-database and input-data dates must be recorded for each run.","status":"inapplicable","source_ids":["evidence-official-81e51077d7d3352a6de4"],"source_locator":"README.md: opening, Publications, Installation and user-guide pointers"},{"label":"Context limits","value":"Database/query scale and implementation limits, not a neural token window.","status":"source_checked","source_ids":["evidence-official-81e51077d7d3352a6de4"],"source_locator":"README.md: opening, Publications, Installation and user-guide pointers"},{"label":"Weights licence","value":"Inapplicable: no neural checkpoint in the core search/clustering method.","status":"inapplicable","source_ids":["evidence-official-81e51077d7d3352a6de4"],"source_locator":"README.md: opening, Publications, Installation and user-guide pointers"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/soedinglab/MMseqs2","status":"source_checked","source_ids":["evidence-official-81e51077d7d3352a6de4"],"source_locator":"README.md: opening, Publications, Installation and user-guide pointers"},{"label":"Code licence","value":"MIT","status":"source_checked","source_ids":["evidence-official-950c72bc9d4bc037f6e2"],"source_locator":"LICENSE.md: licence text"}],"strengths":[{"text":"Provides reusable search and clustering procedures suitable as homology-based comparators.","source_ids":["evidence-official-81e51077d7d3352a6de4"],"source_locator":"README.md: opening, Publications, Installation and user-guide pointers"}],"limitations":[{"text":"Hardware paths, sensitivity settings and reference database versions affect an evaluation. A sequence-similarity hit is not itself a validated functional measurement.","source_ids":["evidence-official-81e51077d7d3352a6de4"],"source_locator":"README.md: opening, Publications, Installation and user-guide pointers"}],"diagram":{"title":"MMseqs2 workflow","steps":["Sequence collection","Configured search or clustering","Sequence relationships","Hits or clusters"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-81e51077d7d3352a6de4"],"source_locator":"README.md: opening, Publications, Installation and user-guide pointers"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}},"description":"Sequence search and clustering","facets":{"areas":["protein-function"]},"id":"discovery-model-mmseqs2","kind":"model","links":[{"relation":"applicable_to","target_id":"discovery-benchmark-cafa"}],"name":"MMseqs2","source_ids":["src-discovery-soedinglab-mmseqs2"],"status":"discovered"}
{"attributes":{"access":"official_source_linked","benchmark_applicability":"candidate; not evidence of a reported evaluation","candidate_benchmark_ids":["discovery-benchmark-massspecgym"],"entity_level":"family","reported_name":"MSAlign","version":null,"historical_missing_metadata":{"checkpoint":"unextracted","code_licence":"unextracted","parameters":"unextracted","training_cutoff":"unextracted","training_data":"unextracted","version":"unextracted","weights_licence":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"MSAlign retrieves candidate molecules from tandem mass spectra by aligning pretrained molecular and spectral representations.","summary_source_ids":["evidence-official-ba08b650243c9212de3f","evidence-official-760ab2fa8c396aeb796c"],"summary_source_locator":"MSAlign paper Section 3 Architecture and Training, Section 4 on splitting, and Section 5.1 Experimental Setting","sections":[{"title":"How it works","body":"MSAlign retrieves candidate molecules from tandem mass spectra by aligning pretrained molecular and spectral representations. Frozen DreaMS and ChemBERTa encoders connected by lightweight MLP projections trained with a candidate-based contrastive objective. The documented inputs are MS/MS spectrum and a set of candidate molecular structures. The output consists of candidate-molecule retrieval scores in a shared representation space.","source_ids":["evidence-official-ba08b650243c9212de3f","evidence-official-760ab2fa8c396aeb796c"],"source_locator":"MSAlign paper Section 3 Architecture and Training, Section 4 on splitting, and Section 5.1 Experimental Setting"},{"title":"Versions and reproducibility","body":"MSAlign arXiv:2605.19752v1, submitted 19 May 2026. The applicable input limits require configuration-specific checking.","source_ids":["evidence-official-ba08b650243c9212de3f","evidence-official-760ab2fa8c396aeb796c"],"source_locator":"MSAlign paper Section 3 Architecture and Training, Section 4 on splitting, and Section 5.1 Experimental Setting"}],"facts":[{"label":"Model type","value":"Frozen spectral/molecular encoders with learned alignment projections","status":"source_checked","source_ids":["evidence-official-ba08b650243c9212de3f","evidence-official-760ab2fa8c396aeb796c"],"source_locator":"MSAlign paper Section 3 Architecture and Training, Section 4 on splitting, and Section 5.1 Experimental Setting"},{"label":"Architecture","value":"Frozen DreaMS and ChemBERTa encoders connected by lightweight MLP projections trained with a candidate-based contrastive objective.","status":"source_checked","source_ids":["evidence-official-ba08b650243c9212de3f","evidence-official-760ab2fa8c396aeb796c"],"source_locator":"MSAlign paper Section 3 Architecture and Training, Section 4 on splitting, and Section 5.1 Experimental Setting"},{"label":"Inputs","value":"MS/MS spectrum and a set of candidate molecular structures.","status":"source_checked","source_ids":["evidence-official-ba08b650243c9212de3f","evidence-official-760ab2fa8c396aeb796c"],"source_locator":"MSAlign paper Section 3 Architecture and Training, Section 4 on splitting, and Section 5.1 Experimental Setting"},{"label":"Outputs","value":"Candidate-molecule retrieval scores in a shared representation space.","status":"source_checked","source_ids":["evidence-official-ba08b650243c9212de3f","evidence-official-760ab2fa8c396aeb796c"],"source_locator":"MSAlign paper Section 3 Architecture and Training, Section 4 on splitting, and Section 5.1 Experimental Setting"},{"label":"Parameters","value":"Approximately 4M trainable projection parameters; frozen DreaMS and ChemBERTa backbones are reported as 96M and 92M respectively.","status":"source_checked","source_ids":["evidence-official-ba08b650243c9212de3f","evidence-official-760ab2fa8c396aeb796c"],"source_locator":"MSAlign paper Section 3 Architecture and Training, Section 4 on splitting, and Section 5.1 Experimental Setting"},{"label":"Known versions","value":"MSAlign arXiv:2605.19752v1, submitted 19 May 2026.","status":"source_checked","source_ids":["evidence-official-ba08b650243c9212de3f","evidence-official-760ab2fa8c396aeb796c"],"source_locator":"MSAlign paper Section 3 Architecture and Training, Section 4 on splitting, and Section 5.1 Experimental Setting"},{"label":"Training data","value":"Projection layers are fitted separately on the NPLIB1, MassSpecGym or Spectraverse training splits. The paper distinguishes spectrum/molecule pair counts from unique molecules and controls candidate retrieval using mass matching.","status":"source_checked","source_ids":["evidence-official-ba08b650243c9212de3f","evidence-official-760ab2fa8c396aeb796c"],"source_locator":"MSAlign paper Section 3 Architecture and Training, Section 4 on splitting, and Section 5.1 Experimental Setting"},{"label":"Training cutoff","value":"NPLIB1, MassSpecGym and Spectraverse are separately split training resources. The inspected paper does not define one latest measurement date covering all three.","status":"unreported","source_ids":["evidence-official-ba08b650243c9212de3f","evidence-official-760ab2fa8c396aeb796c"],"source_locator":"MSAlign paper Section 3 Architecture and Training, Section 4 on splitting, and Section 5.1 Experimental Setting"},{"label":"Context limits","value":"The pipeline inherits spectrum and molecule preprocessing from its frozen DreaMS and ChemBERTa encoders. The inspected MSAlign architecture section does not state a single joint input limit.","status":"unreported","source_ids":["evidence-official-ba08b650243c9212de3f","evidence-official-760ab2fa8c396aeb796c"],"source_locator":"MSAlign paper Section 3 Architecture and Training, Section 4 on splitting, and Section 5.1 Experimental Setting"},{"label":"Weights licence","value":"The inspected preprint does not state distribution terms for the learned projection checkpoints. Frozen encoder licences remain separate from projection-weight rights.","status":"unreported","source_ids":["evidence-official-ba08b650243c9212de3f","evidence-official-760ab2fa8c396aeb796c"],"source_locator":"MSAlign paper Section 3 Architecture and Training, Section 4 on splitting, and Section 5.1 Experimental Setting; page.html: inspected official source"},{"label":"Access","value":"Official project documentation and implementation: https://arxiv.org/abs/2605.19752","status":"source_checked","source_ids":["evidence-official-ba08b650243c9212de3f","evidence-official-760ab2fa8c396aeb796c"],"source_locator":"MSAlign paper Section 3 Architecture and Training, Section 4 on splitting, and Section 5.1 Experimental Setting"},{"label":"Code licence","value":"The inspected preprint describes the algorithm but does not supply a separate code licence; no code-distribution permission is inferred from the paper licence.","status":"unreported","source_ids":["evidence-official-ba08b650243c9212de3f"],"source_locator":"page.html: inspected official source"}],"strengths":[{"text":"Keeps the large encoders frozen while learning alignment projections for the retrieval task.","source_ids":["evidence-official-ba08b650243c9212de3f","evidence-official-760ab2fa8c396aeb796c"],"source_locator":"MSAlign paper Section 3 Architecture and Training, Section 4 on splitting, and Section 5.1 Experimental Setting"}],"limitations":[{"text":"Candidate construction and splitting change the retrieval problem. The authors explicitly discuss the trade-off between leakage control and distribution shift.","source_ids":["evidence-official-ba08b650243c9212de3f","evidence-official-760ab2fa8c396aeb796c"],"source_locator":"MSAlign paper Section 3 Architecture and Training, Section 4 on splitting, and Section 5.1 Experimental Setting"}],"diagram":{"title":"MSAlign workflow","steps":["Spectrum and candidate molecules","Frozen DreaMS and ChemBERTa","Learned projection alignment","Candidate retrieval"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-ba08b650243c9212de3f","evidence-official-760ab2fa8c396aeb796c"],"source_locator":"MSAlign paper Section 3 Architecture and Training, Section 4 on splitting, and Section 5.1 Experimental Setting"},"coverage":"limited","gaps":["Training cutoff: NPLIB1, MassSpecGym and Spectraverse are separately split training resources. The inspected paper does not define one latest measurement date covering all three.","Context limits: The pipeline inherits spectrum and molecule preprocessing from its frozen DreaMS and ChemBERTa encoders. The inspected MSAlign architecture section does not state a single joint input limit.","Weights licence: The inspected preprint does not state distribution terms for the learned projection checkpoints. Frozen encoder licences remain separate from projection-weight rights.","Code licence: The inspected preprint describes the algorithm but does not supply a separate code licence; no code-distribution permission is inferred from the paper licence."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}},"description":"Spectrum-to-molecule representation alignment","facets":{"areas":["metabolomics"]},"id":"discovery-model-msalign","kind":"model","links":[{"relation":"applicable_to","target_id":"discovery-benchmark-massspecgym"}],"name":"MSAlign","source_ids":["src-discovery-msalign"],"status":"discovered"}
{"attributes":{"access":"official_source_linked","benchmark_applicability":"candidate; not evidence of a reported evaluation","candidate_benchmark_ids":["discovery-benchmark-geneb"],"entity_level":"family","reported_name":"Nucleotide Transformer","version":null,"historical_missing_metadata":{"checkpoint":"unextracted","code_licence":"unextracted","parameters":"unextracted","training_cutoff":"unextracted","training_data":"unextracted","version":"unextracted","weights_licence":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Nucleotide Transformer is a family of DNA encoders pretrained on human or multispecies sequence corpora.","summary_source_ids":["evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-aadbeb0f10ec55d34f5f"],"summary_source_locator":"docs/nucleotide_transformer.md: Model Variants and Sizes, Tokenization and How to use; Nucleotide Transformer paper Methods: Architecture, Pre-training datasets and Training","sections":[{"title":"How it works","body":"Nucleotide Transformer is a family of DNA encoders pretrained on human or multispecies sequence corpora. Encoder-only transformers with 6-mer tokens; v1 uses learned positional encodings and v2 uses rotary positions and SwiGLU. The documented inputs are DNA sequences tokenized into 6-mers, with single-base handling of N and remainder bases. The output consists of contextual representations used in specified downstream prediction workflows.","source_ids":["evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-aadbeb0f10ec55d34f5f"],"source_locator":"docs/nucleotide_transformer.md: Model Variants and Sizes, Tokenization and How to use; Nucleotide Transformer paper Methods: Architecture, Pre-training datasets and Training"},{"title":"Versions and reproducibility","body":"NT-v1 human-reference/1000G/multispecies variants and NT-v2 50M/100M/250M/500M. NT-v3 is a separate architecture described elsewhere in the repository. v1: approximately 6kb; v2: 2,048 tokens, approximately 12kb. Exact base count depends on special and ambiguous tokens.","source_ids":["evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-aadbeb0f10ec55d34f5f"],"source_locator":"docs/nucleotide_transformer.md: Model Variants and Sizes, Tokenization and How to use; Nucleotide Transformer paper Methods: Architecture, Pre-training datasets and Training"}],"facts":[{"label":"Model type","value":"DNA transformer encoder family","status":"source_checked","source_ids":["evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-aadbeb0f10ec55d34f5f"],"source_locator":"docs/nucleotide_transformer.md: Model Variants and Sizes, Tokenization and How to use; Nucleotide Transformer paper Methods: Architecture, Pre-training datasets and Training"},{"label":"Architecture","value":"Encoder-only transformers with 6-mer tokens; v1 uses learned positional encodings and v2 uses rotary positions and SwiGLU.","status":"source_checked","source_ids":["evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-aadbeb0f10ec55d34f5f"],"source_locator":"docs/nucleotide_transformer.md: Model Variants and Sizes, Tokenization and How to use; Nucleotide Transformer paper Methods: Architecture, Pre-training datasets and Training"},{"label":"Inputs","value":"DNA sequences tokenized into 6-mers, with single-base handling of N and remainder bases.","status":"source_checked","source_ids":["evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-aadbeb0f10ec55d34f5f"],"source_locator":"docs/nucleotide_transformer.md: Model Variants and Sizes, Tokenization and How to use; Nucleotide Transformer paper Methods: Architecture, Pre-training datasets and Training"},{"label":"Outputs","value":"Contextual representations used in specified downstream prediction workflows.","status":"source_checked","source_ids":["evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-aadbeb0f10ec55d34f5f"],"source_locator":"docs/nucleotide_transformer.md: Model Variants and Sizes, Tokenization and How to use; Nucleotide Transformer paper Methods: Architecture, Pre-training datasets and Training"},{"label":"Parameters","value":"50M to 2.5B across the documented v1/v2 family.","status":"source_checked","source_ids":["evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-aadbeb0f10ec55d34f5f"],"source_locator":"docs/nucleotide_transformer.md: Model Variants and Sizes, Tokenization and How to use; Nucleotide Transformer paper Methods: Architecture, Pre-training datasets and Training"},{"label":"Known versions","value":"NT-v1 human-reference/1000G/multispecies variants and NT-v2 50M/100M/250M/500M. NT-v3 is a separate architecture described elsewhere in the repository.","status":"source_checked","source_ids":["evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-aadbeb0f10ec55d34f5f"],"source_locator":"docs/nucleotide_transformer.md: Model Variants and Sizes, Tokenization and How to use; Nucleotide Transformer paper Methods: Architecture, Pre-training datasets and Training"},{"label":"Training data","value":"v1 variants use GRCh38, 3,202 human genomes or 850 multispecies genomes; v2 uses the multispecies corpus.","status":"source_checked","source_ids":["evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-aadbeb0f10ec55d34f5f"],"source_locator":"docs/nucleotide_transformer.md: Model Variants and Sizes, Tokenization and How to use; Nucleotide Transformer paper Methods: Architecture, Pre-training datasets and Training"},{"label":"Training cutoff","value":"The paper specifies human-reference, 1000 Genomes and multispecies training collections by variant. A single latest-deposition date for all sequences is not supplied in the inspected pretraining-data section.","status":"unreported","source_ids":["evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-aadbeb0f10ec55d34f5f"],"source_locator":"docs/nucleotide_transformer.md: Model Variants and Sizes, Tokenization and How to use; Nucleotide Transformer paper Methods: Architecture, Pre-training datasets and Training"},{"label":"Context limits","value":"v1: approximately 6kb; v2: 2,048 tokens, approximately 12kb. Exact base count depends on special and ambiguous tokens.","status":"source_checked","source_ids":["evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-aadbeb0f10ec55d34f5f"],"source_locator":"docs/nucleotide_transformer.md: Model Variants and Sizes, Tokenization and How to use; Nucleotide Transformer paper Methods: Architecture, Pre-training datasets and Training"},{"label":"Weights licence","value":"Separate checkpoint-distribution terms are not stated in the inspected release documentation and licence material. The source-code licence alone is not recorded as an explicit weight grant.","status":"unreported","source_ids":["evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-aadbeb0f10ec55d34f5f","evidence-official-7e4b193e47ba209860a1"],"source_locator":"docs/nucleotide_transformer.md: Model Variants and Sizes, Tokenization and How to use; Nucleotide Transformer paper Methods: Architecture, Pre-training datasets and Training; LICENSE.md: licence text"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/instadeepai/nucleotide-transformer","status":"source_checked","source_ids":["evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-aadbeb0f10ec55d34f5f"],"source_locator":"docs/nucleotide_transformer.md: Model Variants and Sizes, Tokenization and How to use; Nucleotide Transformer paper Methods: Architecture, Pre-training datasets and Training"},{"label":"Code licence","value":"CC-BY-NC-SA-4.0","status":"source_checked","source_ids":["evidence-official-7e4b193e47ba209860a1"],"source_locator":"LICENSE.md: licence text"}],"strengths":[{"text":"The family provides multiple sizes and training corpora, allowing those factors to be compared explicitly.","source_ids":["evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-aadbeb0f10ec55d34f5f"],"source_locator":"docs/nucleotide_transformer.md: Model Variants and Sizes, Tokenization and How to use; Nucleotide Transformer paper Methods: Architecture, Pre-training datasets and Training"}],"limitations":[{"text":"Training corpora and architectures differ across v1 and v2. A family-level name is insufficient to reproduce a score.","source_ids":["evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-aadbeb0f10ec55d34f5f"],"source_locator":"docs/nucleotide_transformer.md: Model Variants and Sizes, Tokenization and How to use; Nucleotide Transformer paper Methods: Architecture, Pre-training datasets and Training"}],"diagram":{"title":"Nucleotide Transformer workflow","steps":["DNA sequence","6-mer tokenization","Selected NT encoder","Representation or adapted predictor"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-aadbeb0f10ec55d34f5f"],"source_locator":"docs/nucleotide_transformer.md: Model Variants and Sizes, Tokenization and How to use; Nucleotide Transformer paper Methods: Architecture, Pre-training datasets and Training"},"coverage":"limited","gaps":["Training cutoff: The paper specifies human-reference, 1000 Genomes and multispecies training collections by variant. A single latest-deposition date for all sequences is not supplied in the inspected pretraining-data section.","Weights licence: Separate checkpoint-distribution terms are not stated in the inspected release documentation and licence material. The source-code licence alone is not recorded as an explicit weight grant."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}},"description":"Genomic representation model family","facets":{"areas":["genomics"]},"id":"discovery-model-nucleotide-transformer","kind":"model","links":[{"relation":"applicable_to","target_id":"discovery-benchmark-geneb"}],"name":"Nucleotide Transformer","source_ids":["src-discovery-instadeepai-nucleotide-transformer"],"status":"discovered"}
{"attributes":{"access":"official_source_linked","benchmark_applicability":"candidate; not evidence of a reported evaluation","candidate_benchmark_ids":["discovery-benchmark-casp"],"entity_level":"method","reported_name":"OpenFold","version":null,"historical_missing_metadata":{"checkpoint":"unextracted","code_licence":"unextracted","parameters":"unextracted","training_cutoff":"unextracted","training_data":"unextracted","version":"unextracted","weights_licence":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"OpenFold is a trainable PyTorch implementation of AlphaFold 2 and AlphaFold-Multimer workflows.","summary_source_ids":["evidence-official-c0ac4d10941b082e5f86","evidence-official-53934b67953ba33b2c95"],"summary_source_locator":"docs/source/original_readme.md: Features, Inference and Copyright Notice","sections":[{"title":"How it works","body":"OpenFold is a trainable PyTorch implementation of AlphaFold 2 and AlphaFold-Multimer workflows. AlphaFold-compatible folding architecture with configurable attention implementations and a training pipeline. The documented inputs are protein sequence, sequence alignments and optional structural templates, depending on the inference mode. The output consists of predicted protein structures from selected OpenFold or compatible AlphaFold parameters.","source_ids":["evidence-official-c0ac4d10941b082e5f86","evidence-official-53934b67953ba33b2c95"],"source_locator":"docs/source/original_readme.md: Features, Inference and Copyright Notice"},{"title":"Versions and reproducibility","body":"Documented monomer v2.0.1 and multimer v2.3.2 compatibility; distinguish OpenFold and imported AlphaFold parameters. Memory- and configuration-dependent; low-memory attention and CPU offloading are documented.","source_ids":["evidence-official-c0ac4d10941b082e5f86","evidence-official-53934b67953ba33b2c95"],"source_locator":"docs/source/original_readme.md: Features, Inference and Copyright Notice"}],"facts":[{"label":"Model type","value":"Trainable AlphaFold2-compatible structure predictor","status":"source_checked","source_ids":["evidence-official-c0ac4d10941b082e5f86","evidence-official-53934b67953ba33b2c95"],"source_locator":"docs/source/original_readme.md: Features, Inference and Copyright Notice"},{"label":"Architecture","value":"AlphaFold-compatible folding architecture with configurable attention implementations and a training pipeline.","status":"source_checked","source_ids":["evidence-official-c0ac4d10941b082e5f86","evidence-official-53934b67953ba33b2c95"],"source_locator":"docs/source/original_readme.md: Features, Inference and Copyright Notice"},{"label":"Inputs","value":"Protein sequence, sequence alignments and optional structural templates, depending on the inference mode.","status":"source_checked","source_ids":["evidence-official-c0ac4d10941b082e5f86","evidence-official-53934b67953ba33b2c95"],"source_locator":"docs/source/original_readme.md: Features, Inference and Copyright Notice"},{"label":"Outputs","value":"Predicted protein structures from selected OpenFold or compatible AlphaFold parameters.","status":"source_checked","source_ids":["evidence-official-c0ac4d10941b082e5f86","evidence-official-53934b67953ba33b2c95"],"source_locator":"docs/source/original_readme.md: Features, Inference and Copyright Notice"},{"label":"Parameters","value":"The official documentation supports different AlphaFold2-compatible configurations and heads; this family record does not select one checkpoint with one verified total.","status":"unreported","source_ids":["evidence-official-c0ac4d10941b082e5f86","evidence-official-53934b67953ba33b2c95"],"source_locator":"docs/source/original_readme.md: Features, Inference and Copyright Notice"},{"label":"Known versions","value":"Documented monomer v2.0.1 and multimer v2.3.2 compatibility; distinguish OpenFold and imported AlphaFold parameters.","status":"source_checked","source_ids":["evidence-official-c0ac4d10941b082e5f86","evidence-official-53934b67953ba33b2c95"],"source_locator":"docs/source/original_readme.md: Features, Inference and Copyright Notice"},{"label":"Training data","value":"The documentation releases approximately 400,000 MSAs and PDB70 template-hit files; the actual training schedule/checkpoint remains a separate identity.","status":"source_checked","source_ids":["evidence-official-c0ac4d10941b082e5f86","evidence-official-53934b67953ba33b2c95"],"source_locator":"docs/source/original_readme.md: Features, Inference and Copyright Notice"},{"label":"Training cutoff","value":"The inspected implementation documentation does not supply one common cutoff across its supported original, retrained and extended-context model configurations.","status":"unreported","source_ids":["evidence-official-c0ac4d10941b082e5f86","evidence-official-53934b67953ba33b2c95"],"source_locator":"docs/source/original_readme.md: Features, Inference and Copyright Notice"},{"label":"Context limits","value":"Memory- and configuration-dependent; low-memory attention and CPU offloading are documented.","status":"source_checked","source_ids":["evidence-official-c0ac4d10941b082e5f86","evidence-official-53934b67953ba33b2c95"],"source_locator":"docs/source/original_readme.md: Features, Inference and Copyright Notice"},{"label":"Weights licence","value":"The copyright notice specifies CC-BY-4.0 for DeepMind pretrained parameters; this does not automatically establish all OpenFold checkpoint terms.","status":"source_checked","source_ids":["evidence-official-c0ac4d10941b082e5f86","evidence-official-53934b67953ba33b2c95"],"source_locator":"docs/source/original_readme.md: Features, Inference and Copyright Notice"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/aqlaboratory/openfold","status":"source_checked","source_ids":["evidence-official-c0ac4d10941b082e5f86","evidence-official-53934b67953ba33b2c95"],"source_locator":"docs/source/original_readme.md: Features, Inference and Copyright Notice"},{"label":"Code licence","value":"Apache-2.0","status":"source_checked","source_ids":["evidence-official-c7d214a82cfd1afc827b"],"source_locator":"LICENSE: licence text"}],"strengths":[{"text":"Provides training code, alignment resources and conversion between compatible parameter formats.","source_ids":["evidence-official-c0ac4d10941b082e5f86","evidence-official-53934b67953ba33b2c95"],"source_locator":"docs/source/original_readme.md: Features, Inference and Copyright Notice"}],"limitations":[{"text":"OpenFold-trained weights and DeepMind weights are separate artifacts. The documentation notes deliberate differences, including omitted model ensembling.","source_ids":["evidence-official-c0ac4d10941b082e5f86","evidence-official-53934b67953ba33b2c95"],"source_locator":"docs/source/original_readme.md: Features, Inference and Copyright Notice"}],"diagram":{"title":"OpenFold workflow","steps":["Sequence and alignments","AlphaFold-compatible folding network","Structure module","Predicted protein structure"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-c0ac4d10941b082e5f86","evidence-official-53934b67953ba33b2c95"],"source_locator":"docs/source/original_readme.md: Features, Inference and Copyright Notice"},"coverage":"limited","gaps":["Parameters: The official documentation supports different AlphaFold2-compatible configurations and heads; this family record does not select one checkpoint with one verified total.","Training cutoff: The inspected implementation documentation does not supply one common cutoff across its supported original, retrained and extended-context model configurations."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}},"description":"Trainable protein structure prediction implementation","facets":{"areas":["protein-structure"]},"id":"discovery-model-openfold","kind":"model","links":[{"relation":"applicable_to","target_id":"discovery-benchmark-casp"}],"name":"OpenFold","source_ids":["src-discovery-aqlaboratory-openfold"],"status":"discovered"}
{"attributes":{"access":"official_source_linked","benchmark_applicability":"candidate; not evidence of a reported evaluation","candidate_benchmark_ids":[],"entity_level":"family","reported_name":"Pangolin","version":null,"historical_missing_metadata":{"checkpoint":"unextracted","code_licence":"unextracted","parameters":"unextracted","training_cutoff":"unextracted","training_data":"unextracted","version":"unextracted","weights_licence":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Pangolin predicts splice-site strength and changes caused by genetic variants.","summary_source_ids":["evidence-official-eaf3efab8850672f44e4","evidence-official-de1ee43bdd6f04de9850"],"summary_source_locator":"Paper: Deep neural network architecture and Generating training and test sets; README.md: Usage","sections":[{"title":"How it works","body":"Pangolin predicts splice-site strength and changes caused by genetic variants. Dilated convolutional network with 16 residual blocks and skip connections; separate probability and usage outputs for heart, liver, brain and testis. The documented inputs are VCF or CSV variants, reference FASTA and matching gene annotations; custom sequence inference is also available. The output consists of predicted increases/decreases in splice-site strength and their positions.","source_ids":["evidence-official-eaf3efab8850672f44e4","evidence-official-de1ee43bdd6f04de9850"],"source_locator":"Paper: Deep neural network architecture and Generating training and test sets; README.md: Usage"},{"title":"Versions and reproducibility","body":"Pangolin implementation; gene-annotation database and selected weights must be recorded with a run. 5,000 bases upstream and downstream each output position; minimum 10,001-base input for one prediction, with 15,000-base training blocks producing 5,000 central outputs.","source_ids":["evidence-official-eaf3efab8850672f44e4","evidence-official-de1ee43bdd6f04de9850"],"source_locator":"Paper: Deep neural network architecture and Generating training and test sets; README.md: Usage"}],"facts":[{"label":"Model type","value":"Dilated convolutional splicing predictor","status":"source_checked","source_ids":["evidence-official-eaf3efab8850672f44e4","evidence-official-de1ee43bdd6f04de9850"],"source_locator":"Paper: Deep neural network architecture and Generating training and test sets; README.md: Usage"},{"label":"Architecture","value":"Dilated convolutional network with 16 residual blocks and skip connections; separate probability and usage outputs for heart, liver, brain and testis.","status":"source_checked","source_ids":["evidence-official-eaf3efab8850672f44e4","evidence-official-de1ee43bdd6f04de9850"],"source_locator":"Paper: Deep neural network architecture and Generating training and test sets; README.md: Usage"},{"label":"Inputs","value":"VCF or CSV variants, reference FASTA and matching gene annotations; custom sequence inference is also available.","status":"source_checked","source_ids":["evidence-official-eaf3efab8850672f44e4","evidence-official-de1ee43bdd6f04de9850"],"source_locator":"Paper: Deep neural network architecture and Generating training and test sets; README.md: Usage"},{"label":"Outputs","value":"Predicted increases/decreases in splice-site strength and their positions.","status":"source_checked","source_ids":["evidence-official-eaf3efab8850672f44e4","evidence-official-de1ee43bdd6f04de9850"],"source_locator":"Paper: Deep neural network architecture and Generating training and test sets; README.md: Usage"},{"label":"Parameters","value":"The reviewed architecture section specifies the dilated residual network, but does not give a complete parameter total for the released ensemble.","status":"unreported","source_ids":["evidence-official-eaf3efab8850672f44e4","evidence-official-de1ee43bdd6f04de9850"],"source_locator":"Paper: Deep neural network architecture and Generating training and test sets; README.md: Usage"},{"label":"Known versions","value":"Pangolin implementation; gene-annotation database and selected weights must be recorded with a run.","status":"source_checked","source_ids":["evidence-official-eaf3efab8850672f44e4","evidence-official-de1ee43bdd6f04de9850"],"source_locator":"Paper: Deep neural network architecture and Generating training and test sets; README.md: Usage"},{"label":"Training data","value":"Human, rhesus macaque, mouse and rat sequence/splicing data. Human test chromosomes 1, 3, 5, 7 and 9 are held out, with homologous training genes filtered using Ensembl BioMart.","status":"source_checked","source_ids":["evidence-official-eaf3efab8850672f44e4","evidence-official-de1ee43bdd6f04de9850"],"source_locator":"Paper: Deep neural network architecture and Generating training and test sets; README.md: Usage"},{"label":"Training cutoff","value":"Training annotations are GENCODE 34 (human), Ensembl 100 (rhesus), GENCODE M25 (mouse) and Ensembl 101 (rat). These component releases do not establish one latest RNA-seq collection date.","status":"source_checked","source_ids":["evidence-official-eaf3efab8850672f44e4","evidence-official-de1ee43bdd6f04de9850"],"source_locator":"Paper: Deep neural network architecture and Generating training and test sets; README.md: Usage"},{"label":"Context limits","value":"5,000 bases upstream and downstream each output position; minimum 10,001-base input for one prediction, with 15,000-base training blocks producing 5,000 central outputs.","status":"source_checked","source_ids":["evidence-official-eaf3efab8850672f44e4","evidence-official-de1ee43bdd6f04de9850"],"source_locator":"Paper: Deep neural network architecture and Generating training and test sets; README.md: Usage"},{"label":"Weights licence","value":"Separate checkpoint-distribution terms are not stated in the inspected release documentation and licence material. The source-code licence alone is not recorded as an explicit weight grant.","status":"unreported","source_ids":["evidence-official-eaf3efab8850672f44e4","evidence-official-de1ee43bdd6f04de9850","evidence-official-dd29c6cbb629171059a6"],"source_locator":"Paper: Deep neural network architecture and Generating training and test sets; README.md: Usage; LICENSE: licence text"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/tkzeng/Pangolin","status":"source_checked","source_ids":["evidence-official-eaf3efab8850672f44e4","evidence-official-de1ee43bdd6f04de9850"],"source_locator":"Paper: Deep neural network architecture and Generating training and test sets; README.md: Usage"},{"label":"Code licence","value":"GPL-3.0; inspect the pinned licence and any file-specific terms.","status":"source_checked","source_ids":["evidence-official-dd29c6cbb629171059a6"],"source_locator":"LICENSE: licence text"}],"strengths":[{"text":"Supports custom sequences and annotation-aware variant scoring with configurable search distance.","source_ids":["evidence-official-eaf3efab8850672f44e4","evidence-official-de1ee43bdd6f04de9850"],"source_locator":"Paper: Deep neural network architecture and Generating training and test sets; README.md: Usage"}],"limitations":[{"text":"Only substitutions and simple insertions/deletions are supported. The documented tool skips variants outside annotated genes, near chromosome ends, inconsistent with the reference or beyond supported deletion lengths.","source_ids":["evidence-official-eaf3efab8850672f44e4","evidence-official-de1ee43bdd6f04de9850"],"source_locator":"Paper: Deep neural network architecture and Generating training and test sets; README.md: Usage"}],"diagram":{"title":"Pangolin workflow","steps":["Variant and reference genome","Construct sequence inputs","Splice-strength prediction","Reference/alternate comparison"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-eaf3efab8850672f44e4","evidence-official-de1ee43bdd6f04de9850"],"source_locator":"Paper: Deep neural network architecture and Generating training and test sets; README.md: Usage"},"coverage":"limited","gaps":["Parameters: The reviewed architecture section specifies the dilated residual network, but does not give a complete parameter total for the released ensemble.","Weights licence: Separate checkpoint-distribution terms are not stated in the inspected release documentation and licence material. The source-code licence alone is not recorded as an explicit weight grant."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}},"description":"Splice site strength prediction","facets":{"areas":["genomics"]},"id":"discovery-model-pangolin","kind":"model","links":[],"name":"Pangolin","source_ids":["src-discovery-tkzeng-pangolin"],"status":"discovered"}
{"attributes":{"access":"official_source_linked","benchmark_applicability":"candidate; not evidence of a reported evaluation","candidate_benchmark_ids":["discovery-benchmark-proteinbench"],"entity_level":"family","reported_name":"ProteinMPNN","version":null,"historical_missing_metadata":{"checkpoint":"unextracted","code_licence":"unextracted","parameters":"unextracted","training_cutoff":"unextracted","training_data":"unextracted","version":"unextracted","weights_licence":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"ProteinMPNN designs amino-acid sequences for a supplied protein backbone.","summary_source_ids":["evidence-official-ea568a52ea3ab476db6a","evidence-official-7b6fdf915d9c6950ad0d","evidence-official-019aff235ae2c3cf29d6","evidence-official-92b0ae57a020cbc40fe7","evidence-official-0b8003691827eb09b0a4"],"summary_source_locator":"ProteinMPNN paper: main-text model development; training/README.md: multi-chain training set, list.csv and cluster manifests; protein_mpnn_utils.py: ProteinMPNN; Supplementary Materials: Methods for training multi-chain models / Training data","sections":[{"title":"How it works","body":"ProteinMPNN converts a supplied backbone into a graph whose edges encode interatomic distances. Message-passing layers update node and edge features, and an autoregressive decoder samples amino acids while conditioning on the backbone and previously assigned residues. Fixed residues, tied positions and chain choices change the design task and must accompany its result.","source_ids":["evidence-official-ea568a52ea3ab476db6a","evidence-official-7b6fdf915d9c6950ad0d","evidence-official-019aff235ae2c3cf29d6","evidence-official-92b0ae57a020cbc40fe7","evidence-official-0b8003691827eb09b0a4"],"source_locator":"ProteinMPNN paper: main-text model development; training/README.md: multi-chain training set, list.csv and cluster manifests; protein_mpnn_utils.py: ProteinMPNN; Supplementary Materials: Methods for training multi-chain models / Training data"},{"title":"Versions and reproducibility","body":"v_48_002, v_48_010, v_48_020 and v_48_030; distinct soluble and C-alpha-only weights. Structure-size and memory dependent. README --max_length is an implementation guard, not a validated scientific context limit.","source_ids":["evidence-official-ea568a52ea3ab476db6a","evidence-official-7b6fdf915d9c6950ad0d","evidence-official-019aff235ae2c3cf29d6","evidence-official-92b0ae57a020cbc40fe7","evidence-official-0b8003691827eb09b0a4"],"source_locator":"ProteinMPNN paper: main-text model development; training/README.md: multi-chain training set, list.csv and cluster manifests; protein_mpnn_utils.py: ProteinMPNN; Supplementary Materials: Methods for training multi-chain models / Training data"}],"facts":[{"label":"Model type","value":"Structure-conditioned message-passing sequence design model","status":"source_checked","source_ids":["evidence-official-ea568a52ea3ab476db6a","evidence-official-7b6fdf915d9c6950ad0d","evidence-official-019aff235ae2c3cf29d6","evidence-official-92b0ae57a020cbc40fe7","evidence-official-0b8003691827eb09b0a4"],"source_locator":"ProteinMPNN paper: main-text model development; training/README.md: multi-chain training set, list.csv and cluster manifests; protein_mpnn_utils.py: ProteinMPNN; Supplementary Materials: Methods for training multi-chain models / Training data"},{"label":"Architecture","value":"Message-passing encoder-decoder with structural interatomic-distance features and edge updates; sequences are sampled with the configured autoregressive decoding procedure.","status":"source_checked","source_ids":["evidence-official-ea568a52ea3ab476db6a","evidence-official-7b6fdf915d9c6950ad0d","evidence-official-019aff235ae2c3cf29d6","evidence-official-92b0ae57a020cbc40fe7","evidence-official-0b8003691827eb09b0a4"],"source_locator":"ProteinMPNN paper: main-text model development; training/README.md: multi-chain training set, list.csv and cluster manifests; protein_mpnn_utils.py: ProteinMPNN; Supplementary Materials: Methods for training multi-chain models / Training data"},{"label":"Inputs","value":"Protein backbone coordinates, with optional fixed residues, chain choices, tied positions and amino-acid constraints.","status":"source_checked","source_ids":["evidence-official-ea568a52ea3ab476db6a","evidence-official-7b6fdf915d9c6950ad0d","evidence-official-019aff235ae2c3cf29d6","evidence-official-92b0ae57a020cbc40fe7","evidence-official-0b8003691827eb09b0a4"],"source_locator":"ProteinMPNN paper: main-text model development; training/README.md: multi-chain training set, list.csv and cluster manifests; protein_mpnn_utils.py: ProteinMPNN; Supplementary Materials: Methods for training multi-chain models / Training data"},{"label":"Outputs","value":"Designed sequences, sequence scores and conditional amino-acid probabilities.","status":"source_checked","source_ids":["evidence-official-ea568a52ea3ab476db6a","evidence-official-7b6fdf915d9c6950ad0d","evidence-official-019aff235ae2c3cf29d6","evidence-official-92b0ae57a020cbc40fe7","evidence-official-0b8003691827eb09b0a4"],"source_locator":"ProteinMPNN paper: main-text model development; training/README.md: multi-chain training set, list.csv and cluster manifests; protein_mpnn_utils.py: ProteinMPNN; Supplementary Materials: Methods for training multi-chain models / Training data"},{"label":"Parameters","value":"The inspected model implementation is configured through encoder/decoder depth and feature width. The paper and training README do not state an exact total for every released checkpoint.","status":"unreported","source_ids":["evidence-official-ea568a52ea3ab476db6a","evidence-official-7b6fdf915d9c6950ad0d","evidence-official-019aff235ae2c3cf29d6","evidence-official-92b0ae57a020cbc40fe7","evidence-official-0b8003691827eb09b0a4"],"source_locator":"ProteinMPNN paper: main-text model development; training/README.md: multi-chain training set, list.csv and cluster manifests; protein_mpnn_utils.py: ProteinMPNN; Supplementary Materials: Methods for training multi-chain models / Training data"},{"label":"Known versions","value":"v_48_002, v_48_010, v_48_020 and v_48_030; distinct soluble and C-alpha-only weights.","status":"source_checked","source_ids":["evidence-official-ea568a52ea3ab476db6a","evidence-official-7b6fdf915d9c6950ad0d","evidence-official-019aff235ae2c3cf29d6","evidence-official-92b0ae57a020cbc40fe7","evidence-official-0b8003691827eb09b0a4"],"source_locator":"ProteinMPNN paper: main-text model development; training/README.md: multi-chain training set, list.csv and cluster manifests; protein_mpnn_utils.py: ProteinMPNN; Supplementary Materials: Methods for training multi-chain models / Training data"},{"label":"Training data","value":"Released multi-chain training set of PDB biological units, with chain metadata and validation/test cluster manifests. The documented set is dated 2 August 2021.","status":"source_checked","source_ids":["evidence-official-ea568a52ea3ab476db6a","evidence-official-7b6fdf915d9c6950ad0d","evidence-official-019aff235ae2c3cf29d6","evidence-official-92b0ae57a020cbc40fe7","evidence-official-0b8003691827eb09b0a4"],"source_locator":"ProteinMPNN paper: main-text model development; training/README.md: multi-chain training set, list.csv and cluster manifests; protein_mpnn_utils.py: ProteinMPNN; Supplementary Materials: Methods for training multi-chain models / Training data"},{"label":"Training cutoff","value":"The released PDB training-set snapshot is dated 2021-08-02; preserve its chain-level deposition metadata and cluster split for a run.","status":"source_checked","source_ids":["evidence-official-ea568a52ea3ab476db6a","evidence-official-7b6fdf915d9c6950ad0d","evidence-official-019aff235ae2c3cf29d6","evidence-official-92b0ae57a020cbc40fe7","evidence-official-0b8003691827eb09b0a4"],"source_locator":"ProteinMPNN paper: main-text model development; training/README.md: multi-chain training set, list.csv and cluster manifests; protein_mpnn_utils.py: ProteinMPNN; Supplementary Materials: Methods for training multi-chain models / Training data"},{"label":"Context limits","value":"Structure-size and memory dependent. README --max_length is an implementation guard, not a validated scientific context limit.","status":"source_checked","source_ids":["evidence-official-ea568a52ea3ab476db6a","evidence-official-7b6fdf915d9c6950ad0d","evidence-official-019aff235ae2c3cf29d6","evidence-official-92b0ae57a020cbc40fe7","evidence-official-0b8003691827eb09b0a4"],"source_locator":"ProteinMPNN paper: main-text model development; training/README.md: multi-chain training set, list.csv and cluster manifests; protein_mpnn_utils.py: ProteinMPNN; Supplementary Materials: Methods for training multi-chain models / Training data"},{"label":"Weights licence","value":"Separate checkpoint-distribution terms are not stated in the inspected release documentation and licence material. The source-code licence alone is not recorded as an explicit weight grant.","status":"unreported","source_ids":["evidence-official-ea568a52ea3ab476db6a","evidence-official-7b6fdf915d9c6950ad0d","evidence-official-019aff235ae2c3cf29d6","evidence-official-92b0ae57a020cbc40fe7","evidence-official-0b8003691827eb09b0a4","evidence-official-eebe7c91156963e6ddc0"],"source_locator":"ProteinMPNN paper: main-text model development; training/README.md: multi-chain training set, list.csv and cluster manifests; protein_mpnn_utils.py: ProteinMPNN; Supplementary Materials: Methods for training multi-chain models / Training data; LICENSE: licence text"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/dauparas/ProteinMPNN","status":"source_checked","source_ids":["evidence-official-ea568a52ea3ab476db6a","evidence-official-7b6fdf915d9c6950ad0d","evidence-official-019aff235ae2c3cf29d6","evidence-official-92b0ae57a020cbc40fe7","evidence-official-0b8003691827eb09b0a4"],"source_locator":"ProteinMPNN paper: main-text model development; training/README.md: multi-chain training set, list.csv and cluster manifests; protein_mpnn_utils.py: ProteinMPNN; Supplementary Materials: Methods for training multi-chain models / Training data"},{"label":"Code licence","value":"MIT","status":"source_checked","source_ids":["evidence-official-eebe7c91156963e6ddc0"],"source_locator":"LICENSE: licence text"}],"strengths":[{"text":"The interface makes design constraints explicit and includes full-backbone, C-alpha-only and soluble-protein weight sets.","source_ids":["evidence-official-ea568a52ea3ab476db6a","evidence-official-7b6fdf915d9c6950ad0d","evidence-official-019aff235ae2c3cf29d6","evidence-official-92b0ae57a020cbc40fe7","evidence-official-0b8003691827eb09b0a4"],"source_locator":"ProteinMPNN paper: main-text model development; training/README.md: multi-chain training set, list.csv and cluster manifests; protein_mpnn_utils.py: ProteinMPNN; Supplementary Materials: Methods for training multi-chain models / Training data"}],"limitations":[{"text":"The requested backbone and constraints are part of the evaluated problem. A command-line maximum-length guard is not evidence that designs at that size have been validated.","source_ids":["evidence-official-ea568a52ea3ab476db6a","evidence-official-7b6fdf915d9c6950ad0d","evidence-official-019aff235ae2c3cf29d6","evidence-official-92b0ae57a020cbc40fe7","evidence-official-0b8003691827eb09b0a4"],"source_locator":"ProteinMPNN paper: main-text model development; training/README.md: multi-chain training set, list.csv and cluster manifests; protein_mpnn_utils.py: ProteinMPNN; Supplementary Materials: Methods for training multi-chain models / Training data"}],"diagram":{"title":"ProteinMPNN workflow","steps":["Protein backbone","Structural graph features","Message-passing model","Constrained sequence sampling"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-ea568a52ea3ab476db6a","evidence-official-7b6fdf915d9c6950ad0d","evidence-official-019aff235ae2c3cf29d6","evidence-official-92b0ae57a020cbc40fe7","evidence-official-0b8003691827eb09b0a4"],"source_locator":"ProteinMPNN paper: main-text model development; training/README.md: multi-chain training set, list.csv and cluster manifests; protein_mpnn_utils.py: ProteinMPNN; Supplementary Materials: Methods for training multi-chain models / Training data"},"coverage":"limited","gaps":["Parameters: The inspected model implementation is configured through encoder/decoder depth and feature width. The paper and training README do not state an exact total for every released checkpoint.","Weights licence: Separate checkpoint-distribution terms are not stated in the inspected release documentation and licence material. The source-code licence alone is not recorded as an explicit weight grant."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}},"description":"Structure-conditioned protein sequence design","facets":{"areas":["protein-structure"]},"id":"discovery-model-proteinmpnn","kind":"model","links":[{"relation":"applicable_to","target_id":"discovery-benchmark-proteinbench"}],"name":"ProteinMPNN","source_ids":["src-discovery-dauparas-proteinmpnn"],"status":"discovered"}
{"attributes":{"access":"official_source_linked","benchmark_applicability":"candidate; not evidence of a reported evaluation","candidate_benchmark_ids":["discovery-benchmark-proteinbench"],"entity_level":"family","reported_name":"RFdiffusion","version":null,"historical_missing_metadata":{"checkpoint":"unextracted","code_licence":"unextracted","parameters":"unextracted","training_cutoff":"unextracted","training_data":"unextracted","version":"unextracted","weights_licence":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"RFdiffusion generates protein structures, optionally conditioned on a motif, target or symmetry constraint.","summary_source_ids":["evidence-official-c7befbb0fbd8620d5c3c","evidence-official-c0351b621280a9c081da","evidence-official-28c1720afecf4240f3f7"],"summary_source_locator":"RFdiffusion paper Main text and Extended Data Figure 1 (pretrained representations); README.md: model variants, contigs and licence; Supplementary Methods Sections 1.4 and 4.1, Table 6","sections":[{"title":"How it works","body":"RFdiffusion generates protein structures, optionally conditioned on a motif, target or symmetry constraint. Diffusion-based protein structure generation with checkpoint-specific conditioning and denoising configuration. The documented inputs are unconditional length specification or structural constraints such as a motif, target and contig map. The output consists of generated protein backbone designs for downstream sequence design and assessment.","source_ids":["evidence-official-c7befbb0fbd8620d5c3c","evidence-official-c0351b621280a9c081da","evidence-official-28c1720afecf4240f3f7"],"source_locator":"RFdiffusion paper Main text and Extended Data Figure 1 (pretrained representations); README.md: model variants, contigs and licence; Supplementary Methods Sections 1.4 and 4.1, Table 6"},{"title":"Versions and reproducibility","body":"Base, active-site, sequence-inpainting and other conditioning-specific checkpoints; preserve the selected weight identity. The RFdiffusion training crop is 384 residues (supplementary Table 6). This crop size is not an inference maximum; contig lengths and conditional task settings remain explicit.","source_ids":["evidence-official-c7befbb0fbd8620d5c3c","evidence-official-c0351b621280a9c081da","evidence-official-28c1720afecf4240f3f7"],"source_locator":"RFdiffusion paper Main text and Extended Data Figure 1 (pretrained representations); README.md: model variants, contigs and licence; Supplementary Methods Sections 1.4 and 4.1, Table 6"}],"facts":[{"label":"Model type","value":"Diffusion-based protein backbone generator","status":"source_checked","source_ids":["evidence-official-c7befbb0fbd8620d5c3c","evidence-official-c0351b621280a9c081da","evidence-official-28c1720afecf4240f3f7"],"source_locator":"RFdiffusion paper Main text and Extended Data Figure 1 (pretrained representations); README.md: model variants, contigs and licence; Supplementary Methods Sections 1.4 and 4.1, Table 6"},{"label":"Architecture","value":"Diffusion-based protein structure generation with checkpoint-specific conditioning and denoising configuration.","status":"source_checked","source_ids":["evidence-official-c7befbb0fbd8620d5c3c","evidence-official-c0351b621280a9c081da","evidence-official-28c1720afecf4240f3f7"],"source_locator":"RFdiffusion paper Main text and Extended Data Figure 1 (pretrained representations); README.md: model variants, contigs and licence; Supplementary Methods Sections 1.4 and 4.1, Table 6"},{"label":"Inputs","value":"Unconditional length specification or structural constraints such as a motif, target and contig map.","status":"source_checked","source_ids":["evidence-official-c7befbb0fbd8620d5c3c","evidence-official-c0351b621280a9c081da","evidence-official-28c1720afecf4240f3f7"],"source_locator":"RFdiffusion paper Main text and Extended Data Figure 1 (pretrained representations); README.md: model variants, contigs and licence; Supplementary Methods Sections 1.4 and 4.1, Table 6"},{"label":"Outputs","value":"Generated protein backbone designs for downstream sequence design and assessment.","status":"source_checked","source_ids":["evidence-official-c7befbb0fbd8620d5c3c","evidence-official-c0351b621280a9c081da","evidence-official-28c1720afecf4240f3f7"],"source_locator":"RFdiffusion paper Main text and Extended Data Figure 1 (pretrained representations); README.md: model variants, contigs and licence; Supplementary Methods Sections 1.4 and 4.1, Table 6"},{"label":"Parameters","value":"The complete supplementary architecture and training sections specify RoseTTAFold-derived modules and training settings but do not state a total for each released conditional checkpoint.","status":"unreported","source_ids":["evidence-official-c7befbb0fbd8620d5c3c","evidence-official-c0351b621280a9c081da","evidence-official-28c1720afecf4240f3f7"],"source_locator":"RFdiffusion paper Main text and Extended Data Figure 1 (pretrained representations); README.md: model variants, contigs and licence; Supplementary Methods Sections 1.4 and 4.1, Table 6"},{"label":"Known versions","value":"Base, active-site, sequence-inpainting and other conditioning-specific checkpoints; preserve the selected weight identity.","status":"source_checked","source_ids":["evidence-official-c7befbb0fbd8620d5c3c","evidence-official-c0351b621280a9c081da","evidence-official-28c1720afecf4240f3f7"],"source_locator":"RFdiffusion paper Main text and Extended Data Figure 1 (pretrained representations); README.md: model variants, contigs and licence; Supplementary Methods Sections 1.4 and 4.1, Table 6"},{"label":"Training data","value":"Fine-tunes pretrained RoseTTAFold to denoise protein backbone structures from the PDB; unconditional and task-conditioned variants are distinct configurations.","status":"source_checked","source_ids":["evidence-official-c7befbb0fbd8620d5c3c","evidence-official-c0351b621280a9c081da","evidence-official-28c1720afecf4240f3f7"],"source_locator":"RFdiffusion paper Main text and Extended Data Figure 1 (pretrained representations); README.md: model variants, contigs and licence; Supplementary Methods Sections 1.4 and 4.1, Table 6"},{"label":"Training cutoff","value":"The supplementary RoseTTAFold pretraining description specifies a 2 August 2021 PDB cutoff and additional AlphaFold2 models. This is pretraining provenance, not a date for every conditional design fine-tune.","status":"source_checked","source_ids":["evidence-official-c7befbb0fbd8620d5c3c","evidence-official-c0351b621280a9c081da","evidence-official-28c1720afecf4240f3f7"],"source_locator":"RFdiffusion paper Main text and Extended Data Figure 1 (pretrained representations); README.md: model variants, contigs and licence; Supplementary Methods Sections 1.4 and 4.1, Table 6"},{"label":"Context limits","value":"The RFdiffusion training crop is 384 residues (supplementary Table 6). This crop size is not an inference maximum; contig lengths and conditional task settings remain explicit.","status":"source_checked","source_ids":["evidence-official-c7befbb0fbd8620d5c3c","evidence-official-c0351b621280a9c081da","evidence-official-28c1720afecf4240f3f7"],"source_locator":"RFdiffusion paper Main text and Extended Data Figure 1 (pretrained representations); README.md: model variants, contigs and licence; Supplementary Methods Sections 1.4 and 4.1, Table 6"},{"label":"Weights licence","value":"BSD licence in the inspected LICENSE explicitly covers both source code and linked downloadable model weights.","status":"source_checked","source_ids":["evidence-official-c7befbb0fbd8620d5c3c","evidence-official-c0351b621280a9c081da","evidence-official-28c1720afecf4240f3f7"],"source_locator":"RFdiffusion paper Main text and Extended Data Figure 1 (pretrained representations); README.md: model variants, contigs and licence; Supplementary Methods Sections 1.4 and 4.1, Table 6"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/RosettaCommons/RFdiffusion","status":"source_checked","source_ids":["evidence-official-c7befbb0fbd8620d5c3c","evidence-official-c0351b621280a9c081da","evidence-official-28c1720afecf4240f3f7"],"source_locator":"RFdiffusion paper Main text and Extended Data Figure 1 (pretrained representations); README.md: model variants, contigs and licence; Supplementary Methods Sections 1.4 and 4.1, Table 6"},{"label":"Code licence","value":"BSD-3-Clause","status":"source_checked","source_ids":["evidence-official-fa91833592f376f6fb51"],"source_locator":"LICENSE: licence text"}],"strengths":[{"text":"Supports motif scaffolding, symmetry, binder design and partial diffusion through explicit conditioning options.","source_ids":["evidence-official-c7befbb0fbd8620d5c3c","evidence-official-c0351b621280a9c081da","evidence-official-28c1720afecf4240f3f7"],"source_locator":"RFdiffusion paper Main text and Extended Data Figure 1 (pretrained representations); README.md: model variants, contigs and licence; Supplementary Methods Sections 1.4 and 4.1, Table 6"}],"limitations":[{"text":"Different conditioning modes use different trained weights. Backbone generation alone does not establish a functional experimentally validated protein.","source_ids":["evidence-official-c7befbb0fbd8620d5c3c","evidence-official-c0351b621280a9c081da","evidence-official-28c1720afecf4240f3f7"],"source_locator":"RFdiffusion paper Main text and Extended Data Figure 1 (pretrained representations); README.md: model variants, contigs and licence; Supplementary Methods Sections 1.4 and 4.1, Table 6"}],"diagram":{"title":"RFdiffusion workflow","steps":["Length or structural constraints","Diffusion sampling","Generated backbone","Separate sequence design and assessment"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-c7befbb0fbd8620d5c3c","evidence-official-c0351b621280a9c081da","evidence-official-28c1720afecf4240f3f7"],"source_locator":"RFdiffusion paper Main text and Extended Data Figure 1 (pretrained representations); README.md: model variants, contigs and licence; Supplementary Methods Sections 1.4 and 4.1, Table 6"},"coverage":"limited","gaps":["Parameters: The complete supplementary architecture and training sections specify RoseTTAFold-derived modules and training settings but do not state a total for each released conditional checkpoint."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}},"description":"Protein structure generation","facets":{"areas":["protein-structure"]},"id":"discovery-model-rfdiffusion","kind":"model","links":[{"relation":"applicable_to","target_id":"discovery-benchmark-proteinbench"}],"name":"RFdiffusion","source_ids":["src-discovery-rosettacommons-rfdiffusion"],"status":"discovered"}
{"attributes":{"access":"official_source_linked","benchmark_applicability":"candidate; not evidence of a reported evaluation","candidate_benchmark_ids":["discovery-benchmark-beacon"],"entity_level":"family","reported_name":"RNA-FM","version":null,"historical_missing_metadata":{"checkpoint":"unextracted","code_licence":"unextracted","parameters":"unextracted","training_cutoff":"unextracted","training_data":"unextracted","version":"unextracted","weights_licence":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"RNA-FM learns contextual representations of RNA nucleotides for downstream RNA analyses.","summary_source_ids":["evidence-official-fd8e332abdf04a75195b","evidence-official-e17e3864c464d981afc7"],"summary_source_locator":"README.md: Quick Start, embedding examples and RNA foundation model comparison table","sections":[{"title":"How it works","body":"RNA-FM converts an RNA sequence into one token per nucleotide. Twelve transformer encoder blocks use self-attention to produce a 640-dimensional representation at each position. During pretraining, the model learns to recover masked nucleotides from their surrounding sequence. The resulting representations can be supplied to a separately specified downstream model; they are not, by themselves, a structure or functional prediction.","source_ids":["evidence-final-model-rna-fm-paper"],"source_locator":"arXiv:2204.00300v5, Methods: ncRNA data collection and preprocessing and RNA foundation model training details (p.22)"},{"title":"Versions and reproducibility","body":"The nucleotide-based rna_fm_t12 and codon-based mrna_fm_t12 interfaces are distinct. The original RNA-FM paper sets a training input-length limit of 1,024 and describes a usable limit of 1,022 nucleotides. That limit should not be assigned to mRNA-FM without checking its separate configuration.","source_ids":["evidence-official-fd8e332abdf04a75195b","evidence-official-e17e3864c464d981afc7","evidence-final-model-rna-fm-paper"],"source_locator":"Official README: Quick Start and RNA-FM/mRNA-FM examples; arXiv:2204.00300v5, Methods: training input length and SARS-CoV-2 genome embedding extraction (pp.22–23)"}],"facts":[{"label":"Model type","value":"Masked-token RNA transformer encoder","status":"source_checked","source_ids":["evidence-official-fd8e332abdf04a75195b","evidence-official-e17e3864c464d981afc7"],"source_locator":"README.md: Quick Start, embedding examples and RNA foundation model comparison table"},{"label":"Architecture","value":"12-layer masked-token transformer encoder with hidden width 640 and 20 attention heads; nucleotide tokens produce contextual representations.","status":"source_checked","source_ids":["evidence-final-model-rna-fm-paper"],"source_locator":"arXiv:2204.00300v5, Methods: ncRNA data collection and preprocessing and RNA foundation model training details (p.22)"},{"label":"Inputs","value":"RNA sequences tokenized at nucleotide resolution.","status":"source_checked","source_ids":["evidence-official-fd8e332abdf04a75195b","evidence-official-e17e3864c464d981afc7"],"source_locator":"README.md: Quick Start, embedding examples and RNA foundation model comparison table"},{"label":"Outputs","value":"Contextual token embeddings for a specified downstream RNA task.","status":"source_checked","source_ids":["evidence-official-fd8e332abdf04a75195b","evidence-official-e17e3864c464d981afc7"],"source_locator":"README.md: Quick Start, embedding examples and RNA foundation model comparison table"},{"label":"Parameters","value":"99M, as printed in the official Foundation Models table.","status":"source_checked","source_ids":["evidence-official-fd8e332abdf04a75195b","evidence-official-e17e3864c464d981afc7"],"source_locator":"README.md: Quick Start, embedding examples and RNA foundation model comparison table"},{"label":"Known versions","value":"rna_fm_t12 and mrna_fm_t12 are separate pretrained interfaces.","status":"source_checked","source_ids":["evidence-official-fd8e332abdf04a75195b","evidence-official-e17e3864c464d981afc7"],"source_locator":"README.md: Quick Start, embedding examples and RNA foundation model comparison table"},{"label":"Training data","value":"23.7 million non-coding RNA sequences collected from RNAcentral. The authors replace T with U and remove identical sequences using CD-HIT-EST at 100% identity, naming the resulting corpus RNAcentral100.","status":"source_checked","source_ids":["evidence-final-model-rna-fm-paper"],"source_locator":"arXiv:2204.00300v5, Methods: ncRNA data collection and preprocessing and RNA foundation model training details (p.22)"},{"label":"Training cutoff","value":"The inspected Methods and official README do not establish an exact dated RNAcentral release. RNAcentral100 is the authors’ processed-corpus label, not a verified release number.","status":"unreported","source_ids":["evidence-final-model-rna-fm-paper","evidence-official-fd8e332abdf04a75195b"],"source_locator":"arXiv:2204.00300v5, Methods: ncRNA data collection and preprocessing (p.22); official README: Foundation Models table"},{"label":"Context limits","value":"The original paper sets a training input-length limit of 1,024 and describes a usable input limit of 1,022 nucleotides. These are the original RNA-FM settings, not a validated limit for later codon-based mRNA-FM checkpoints.","status":"source_checked","source_ids":["evidence-final-model-rna-fm-paper"],"source_locator":"arXiv:2204.00300v5, Methods: RNA foundation model training details (pp.22–23) and RNA-FM application input-limit statement (p.23)"},{"label":"Weights licence","value":"Separate checkpoint-distribution terms are not stated in the inspected release documentation and licence material. The source-code licence alone is not recorded as an explicit weight grant.","status":"unreported","source_ids":["evidence-official-fd8e332abdf04a75195b","evidence-official-e17e3864c464d981afc7","evidence-official-3fde3df73e79e455bd86"],"source_locator":"README.md: Quick Start, embedding examples and RNA foundation model comparison table; LICENSE: licence text"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/ml4bio/RNA-FM","status":"source_checked","source_ids":["evidence-official-fd8e332abdf04a75195b","evidence-official-e17e3864c464d981afc7"],"source_locator":"README.md: Quick Start, embedding examples and RNA foundation model comparison table"},{"label":"Code licence","value":"MIT","status":"source_checked","source_ids":["evidence-official-3fde3df73e79e455bd86"],"source_locator":"LICENSE: licence text"}],"strengths":[{"text":"The repository exposes embedding extraction and examples for downstream RNA analyses.","source_ids":["evidence-official-fd8e332abdf04a75195b","evidence-official-e17e3864c464d981afc7"],"source_locator":"README.md: Quick Start, embedding examples and RNA foundation model comparison table"}],"limitations":[{"text":"Base-level RNA-FM and codon-level mRNA-FM are not interchangeable. The task head and tokenization must be specified in each evaluation.","source_ids":["evidence-official-fd8e332abdf04a75195b","evidence-official-e17e3864c464d981afc7"],"source_locator":"README.md: Quick Start, embedding examples and RNA foundation model comparison table"}],"diagram":{"title":"RNA-FM workflow","steps":["RNA sequence","Nucleotide tokenizer","12-layer transformer encoder","Contextual nucleotide representations"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-fd8e332abdf04a75195b","evidence-official-e17e3864c464d981afc7"],"source_locator":"README.md: Quick Start, embedding examples and RNA foundation model comparison table"},"coverage":"limited","gaps":["Weights licence: Separate checkpoint-distribution terms are not stated in the inspected release documentation and licence material. The source-code licence alone is not recorded as an explicit weight grant.","Training cutoff: An exact dated RNAcentral release is not established by the inspected original Methods or official README."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied. Follow-up retrieved the complete original RNA-FM PDF and inspected the full pinned Geneformer repository inventory; unavailable labels were updated only where new evidence resolved the earlier retrieval gap. Follow-up audit reconciles the narrative with the verified RNA-FM input limit and clarifies the original paper’s RNAcentral100 preprocessing definition; mRNA-FM remains separate."}}},"description":"RNA sequence representation family","facets":{"areas":["rna"]},"id":"discovery-model-rna-fm","kind":"model","links":[{"relation":"applicable_to","target_id":"discovery-benchmark-beacon"}],"name":"RNA-FM","source_ids":["src-discovery-ml4bio-rna-fm"],"status":"discovered"}
{"attributes":{"access":"official_source_linked","benchmark_applicability":"candidate; not evidence of a reported evaluation","candidate_benchmark_ids":["discovery-benchmark-perturbench"],"entity_level":"family","reported_name":"scGPT","version":null,"historical_missing_metadata":{"checkpoint":"unextracted","code_licence":"unextracted","parameters":"unextracted","training_cutoff":"unextracted","training_data":"unextracted","version":"unextracted","weights_licence":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"scGPT learns representations of single-cell molecular measurements and supports task-specific adaptation.","summary_source_ids":["evidence-official-0e4dac85aa0e0ff1d4b2","evidence-official-d8e2d4f2e752370b6565","evidence-official-efa6ec9d886a21534a20"],"summary_source_locator":"May 2023 scGPT preprint Sections 4.1–4.3 and 4.8; current scgpt/model/model.py and README pretrained model table; README.md: Pretrained scGPT Model Zoo, whole-human row","sections":[{"title":"How it works","body":"scGPT combines each gene identity with its expression-value encoding before transformer attention. The implementation supports several expression encoders and cell-pooling choices. Task heads then predict expression or cell labels; optional masking and batch objectives depend on the training configuration.","source_ids":["evidence-official-0e4dac85aa0e0ff1d4b2","evidence-official-d8e2d4f2e752370b6565","evidence-official-efa6ec9d886a21534a20"],"source_locator":"May 2023 scGPT preprint Sections 4.1–4.3 and 4.8; current scgpt/model/model.py and README pretrained model table; README.md: Pretrained scGPT Model Zoo, whole-human row"},{"title":"Versions and reproducibility","body":"The May 2023 preprint reports an early 10M-cell model. The current whole-human checkpoint table reports 33M normal human cells; these sources describe different releases. The applicable input limits require configuration-specific checking.","source_ids":["evidence-official-0e4dac85aa0e0ff1d4b2","evidence-official-d8e2d4f2e752370b6565","evidence-official-efa6ec9d886a21534a20"],"source_locator":"May 2023 scGPT preprint Sections 4.1–4.3 and 4.8; current scgpt/model/model.py and README pretrained model table; README.md: Pretrained scGPT Model Zoo, whole-human row"}],"facts":[{"label":"Model type","value":"Generative single-cell transformer","status":"source_checked","source_ids":["evidence-official-0e4dac85aa0e0ff1d4b2","evidence-official-d8e2d4f2e752370b6565","evidence-official-efa6ec9d886a21534a20"],"source_locator":"May 2023 scGPT preprint Sections 4.1–4.3 and 4.8; current scgpt/model/model.py and README pretrained model table; README.md: Pretrained scGPT Model Zoo, whole-human row"},{"label":"Architecture","value":"Transformer backbone combining learned gene-token embeddings with expression-value encodings and optional batch encodings. Separate expression, cell-classification and optional masked-value or batch-discriminator heads support configured tasks.","status":"source_checked","source_ids":["evidence-official-0e4dac85aa0e0ff1d4b2","evidence-official-d8e2d4f2e752370b6565","evidence-official-efa6ec9d886a21534a20"],"source_locator":"May 2023 scGPT preprint Sections 4.1–4.3 and 4.8; current scgpt/model/model.py and README pretrained model table; README.md: Pretrained scGPT Model Zoo, whole-human row"},{"label":"Inputs","value":"Gene-expression measurements with the checkpoint-matched gene vocabulary.","status":"source_checked","source_ids":["evidence-official-0e4dac85aa0e0ff1d4b2","evidence-official-d8e2d4f2e752370b6565","evidence-official-efa6ec9d886a21534a20"],"source_locator":"May 2023 scGPT preprint Sections 4.1–4.3 and 4.8; current scgpt/model/model.py and README pretrained model table; README.md: Pretrained scGPT Model Zoo, whole-human row"},{"label":"Outputs","value":"Cell/gene representations and task-specific predictions after the relevant workflow.","status":"source_checked","source_ids":["evidence-official-0e4dac85aa0e0ff1d4b2","evidence-official-d8e2d4f2e752370b6565","evidence-official-efa6ec9d886a21534a20"],"source_locator":"May 2023 scGPT preprint Sections 4.1–4.3 and 4.8; current scgpt/model/model.py and README pretrained model table; README.md: Pretrained scGPT Model Zoo, whole-human row"},{"label":"Parameters","value":"The May 2023 model has 12 transformer blocks, width 512 and eight heads. The inspected current model-zoo table does not state the exact parameter total of its separate 33M-cell checkpoint.","status":"unreported","source_ids":["evidence-official-0e4dac85aa0e0ff1d4b2","evidence-official-d8e2d4f2e752370b6565","evidence-official-efa6ec9d886a21534a20"],"source_locator":"May 2023 scGPT preprint Sections 4.1–4.3 and 4.8; current scgpt/model/model.py and README pretrained model table; README.md: Pretrained scGPT Model Zoo, whole-human row"},{"label":"Known versions","value":"The May 2023 preprint reports an early 10M-cell model. The current whole-human checkpoint table reports 33M normal human cells; these sources describe different releases.","status":"source_checked","source_ids":["evidence-official-0e4dac85aa0e0ff1d4b2","evidence-official-d8e2d4f2e752370b6565","evidence-official-efa6ec9d886a21534a20"],"source_locator":"May 2023 scGPT preprint Sections 4.1–4.3 and 4.8; current scgpt/model/model.py and README pretrained model table; README.md: Pretrained scGPT Model Zoo, whole-human row"},{"label":"Training data","value":"The current whole-human model-zoo checkpoint uses 33M normal human cells, alongside separately released organ-specific and pan-cancer models. The earlier May 2023 preprint describes 10M training cells; its corpus is not the current checkpoint corpus.","status":"source_checked","source_ids":["evidence-official-0e4dac85aa0e0ff1d4b2","evidence-official-d8e2d4f2e752370b6565","evidence-official-efa6ec9d886a21534a20"],"source_locator":"May 2023 scGPT preprint Sections 4.1–4.3 and 4.8; current scgpt/model/model.py and README pretrained model table; README.md: Pretrained scGPT Model Zoo, whole-human row"},{"label":"Training cutoff","value":"The reviewed early manuscript and current whole-human model-zoo entry describe different corpora; neither supplies a shared latest-study date for the current checkpoint.","status":"unreported","source_ids":["evidence-official-0e4dac85aa0e0ff1d4b2","evidence-official-d8e2d4f2e752370b6565","evidence-official-efa6ec9d886a21534a20"],"source_locator":"May 2023 scGPT preprint Sections 4.1–4.3 and 4.8; current scgpt/model/model.py and README pretrained model table; README.md: Pretrained scGPT Model Zoo, whole-human row"},{"label":"Context limits","value":"The implementation accepts variable gene sets matched to its vocabulary. The reviewed model-zoo entry does not specify one validated maximum gene sequence for the current whole-human checkpoint.","status":"unreported","source_ids":["evidence-official-0e4dac85aa0e0ff1d4b2","evidence-official-d8e2d4f2e752370b6565","evidence-official-efa6ec9d886a21534a20"],"source_locator":"May 2023 scGPT preprint Sections 4.1–4.3 and 4.8; current scgpt/model/model.py and README pretrained model table; README.md: Pretrained scGPT Model Zoo, whole-human row"},{"label":"Weights licence","value":"Separate checkpoint-distribution terms are not stated in the inspected release documentation and licence material. The source-code licence alone is not recorded as an explicit weight grant.","status":"unreported","source_ids":["evidence-official-0e4dac85aa0e0ff1d4b2","evidence-official-d8e2d4f2e752370b6565","evidence-official-efa6ec9d886a21534a20","evidence-official-43484d4de29aacd65ed7"],"source_locator":"May 2023 scGPT preprint Sections 4.1–4.3 and 4.8; current scgpt/model/model.py and README pretrained model table; README.md: Pretrained scGPT Model Zoo, whole-human row; LICENSE: licence text"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/bowang-lab/scGPT","status":"source_checked","source_ids":["evidence-official-0e4dac85aa0e0ff1d4b2","evidence-official-d8e2d4f2e752370b6565","evidence-official-efa6ec9d886a21534a20"],"source_locator":"May 2023 scGPT preprint Sections 4.1–4.3 and 4.8; current scgpt/model/model.py and README pretrained model table; README.md: Pretrained scGPT Model Zoo, whole-human row"},{"label":"Code licence","value":"MIT","status":"source_checked","source_ids":["evidence-official-43484d4de29aacd65ed7"],"source_locator":"LICENSE: licence text"}],"strengths":[{"text":"Releases include whole-human and organ-specific checkpoints, with tutorials for reference mapping and other downstream tasks.","source_ids":["evidence-official-0e4dac85aa0e0ff1d4b2","evidence-official-d8e2d4f2e752370b6565","evidence-official-efa6ec9d886a21534a20"],"source_locator":"May 2023 scGPT preprint Sections 4.1–4.3 and 4.8; current scgpt/model/model.py and README pretrained model table; README.md: Pretrained scGPT Model Zoo, whole-human row"}],"limitations":[{"text":"Checkpoint choice and vocabulary must match the biological context. A whole-human pretrained encoder and a fine-tuned annotation or perturbation model are distinct evaluated configurations.","source_ids":["evidence-official-0e4dac85aa0e0ff1d4b2","evidence-official-d8e2d4f2e752370b6565","evidence-official-efa6ec9d886a21534a20"],"source_locator":"May 2023 scGPT preprint Sections 4.1–4.3 and 4.8; current scgpt/model/model.py and README pretrained model table; README.md: Pretrained scGPT Model Zoo, whole-human row"}],"diagram":{"title":"scGPT workflow","steps":["Gene expression and vocabulary","scGPT encoder","Cell and gene representations","Task-specific adaptation"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-0e4dac85aa0e0ff1d4b2","evidence-official-d8e2d4f2e752370b6565","evidence-official-efa6ec9d886a21534a20"],"source_locator":"May 2023 scGPT preprint Sections 4.1–4.3 and 4.8; current scgpt/model/model.py and README pretrained model table; README.md: Pretrained scGPT Model Zoo, whole-human row"},"coverage":"limited","gaps":["Parameters: The May 2023 model has 12 transformer blocks, width 512 and eight heads. The inspected current model-zoo table does not state the exact parameter total of its separate 33M-cell checkpoint.","Training cutoff: The reviewed early manuscript and current whole-human model-zoo entry describe different corpora; neither supplies a shared latest-study date for the current checkpoint.","Context limits: The implementation accepts variable gene sets matched to its vocabulary. The reviewed model-zoo entry does not specify one validated maximum gene sequence for the current whole-human checkpoint.","Weights licence: Separate checkpoint-distribution terms are not stated in the inspected release documentation and licence material. The source-code licence alone is not recorded as an explicit weight grant."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}},"description":"Single-cell multi-omics model","facets":{"areas":["single-cell"]},"id":"discovery-model-scgpt","kind":"model","links":[{"relation":"applicable_to","target_id":"discovery-benchmark-perturbench"}],"name":"scGPT","source_ids":["src-discovery-bowang-lab-scgpt"],"status":"discovered"}
{"attributes":{"access":"official_source_linked","benchmark_applicability":"candidate; not evidence of a reported evaluation","candidate_benchmark_ids":["discovery-benchmark-scib"],"entity_level":"family","reported_name":"scVI","version":null,"historical_missing_metadata":{"checkpoint":"unextracted","code_licence":"unextracted","parameters":"unextracted","training_cutoff":"unextracted","training_data":"unextracted","version":"unextracted","weights_licence":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"scVI models single-cell RNA counts with a probabilistic latent-variable model that accounts for observed covariates.","summary_source_ids":["evidence-official-550f6719c2f22f4d57b3","evidence-official-0cbab80c2d1769287a1f"],"summary_source_locator":"docs/user_guide/models/scvi.md: Preliminaries, Generative process, Inference and Limitations","sections":[{"title":"How it works","body":"scVI models single-cell RNA counts with a probabilistic latent-variable model that accounts for observed covariates. Variational autoencoder with a count likelihood and neural encoder/decoder; likelihood and batch/dispersion settings are configurable. The documented inputs are cell-by-gene count matrix, optionally with batch, donor or other covariates. The output consists of low-dimensional cell representations, normalized expression and probabilistic downstream quantities.","source_ids":["evidence-official-550f6719c2f22f4d57b3","evidence-official-0cbab80c2d1769287a1f"],"source_locator":"docs/user_guide/models/scvi.md: Preliminaries, Generative process, Inference and Limitations"},{"title":"Versions and reproducibility","body":"scVI model within scvi-tools; package version, likelihood, covariates and checkpoint are evaluation-specific. Gene-feature matrix rather than a fixed sequence-token window.","source_ids":["evidence-official-550f6719c2f22f4d57b3","evidence-official-0cbab80c2d1769287a1f"],"source_locator":"docs/user_guide/models/scvi.md: Preliminaries, Generative process, Inference and Limitations"}],"facts":[{"label":"Model type","value":"Variational autoencoder for count data","status":"source_checked","source_ids":["evidence-official-550f6719c2f22f4d57b3","evidence-official-0cbab80c2d1769287a1f"],"source_locator":"docs/user_guide/models/scvi.md: Preliminaries, Generative process, Inference and Limitations"},{"label":"Architecture","value":"Variational autoencoder with a count likelihood and neural encoder/decoder; likelihood and batch/dispersion settings are configurable.","status":"source_checked","source_ids":["evidence-official-550f6719c2f22f4d57b3","evidence-official-0cbab80c2d1769287a1f"],"source_locator":"docs/user_guide/models/scvi.md: Preliminaries, Generative process, Inference and Limitations"},{"label":"Inputs","value":"Cell-by-gene count matrix, optionally with batch, donor or other covariates.","status":"source_checked","source_ids":["evidence-official-550f6719c2f22f4d57b3","evidence-official-0cbab80c2d1769287a1f"],"source_locator":"docs/user_guide/models/scvi.md: Preliminaries, Generative process, Inference and Limitations"},{"label":"Outputs","value":"Low-dimensional cell representations, normalized expression and probabilistic downstream quantities.","status":"source_checked","source_ids":["evidence-official-550f6719c2f22f4d57b3","evidence-official-0cbab80c2d1769287a1f"],"source_locator":"docs/user_guide/models/scvi.md: Preliminaries, Generative process, Inference and Limitations"},{"label":"Parameters","value":"Configuration-dependent, including gene count and encoder/decoder dimensions.","status":"source_checked","source_ids":["evidence-official-550f6719c2f22f4d57b3","evidence-official-0cbab80c2d1769287a1f"],"source_locator":"docs/user_guide/models/scvi.md: Preliminaries, Generative process, Inference and Limitations"},{"label":"Known versions","value":"scVI model within scvi-tools; package version, likelihood, covariates and checkpoint are evaluation-specific.","status":"source_checked","source_ids":["evidence-official-550f6719c2f22f4d57b3","evidence-official-0cbab80c2d1769287a1f"],"source_locator":"docs/user_guide/models/scvi.md: Preliminaries, Generative process, Inference and Limitations"},{"label":"Training data","value":"Fitted to the user-selected count matrix or a specified pretrained reference; scVI is not one universal checkpoint.","status":"source_checked","source_ids":["evidence-official-550f6719c2f22f4d57b3","evidence-official-0cbab80c2d1769287a1f"],"source_locator":"docs/user_guide/models/scvi.md: Preliminaries, Generative process, Inference and Limitations"},{"label":"Training cutoff","value":"Inapplicable as one universal pretraining date: scVI is fitted to the supplied dataset, whose collection date and train/test split belong to the evaluation.","status":"inapplicable","source_ids":["evidence-official-550f6719c2f22f4d57b3","evidence-official-0cbab80c2d1769287a1f"],"source_locator":"docs/user_guide/models/scvi.md: Preliminaries, Generative process, Inference and Limitations"},{"label":"Context limits","value":"Gene-feature matrix rather than a fixed sequence-token window.","status":"source_checked","source_ids":["evidence-official-550f6719c2f22f4d57b3","evidence-official-0cbab80c2d1769287a1f"],"source_locator":"docs/user_guide/models/scvi.md: Preliminaries, Generative process, Inference and Limitations"},{"label":"Weights licence","value":"No universal weights release applies to a model fitted on each dataset; any reused checkpoint requires its own licence.","status":"source_checked","source_ids":["evidence-official-550f6719c2f22f4d57b3","evidence-official-0cbab80c2d1769287a1f"],"source_locator":"docs/user_guide/models/scvi.md: Preliminaries, Generative process, Inference and Limitations"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/scverse/scvi-tools","status":"source_checked","source_ids":["evidence-official-550f6719c2f22f4d57b3","evidence-official-0cbab80c2d1769287a1f"],"source_locator":"docs/user_guide/models/scvi.md: Preliminaries, Generative process, Inference and Limitations"},{"label":"Code licence","value":"BSD-3-Clause","status":"source_checked","source_ids":["evidence-official-6851724e3bcb7e9d2781"],"source_locator":"LICENSE: licence text"}],"strengths":[{"text":"Models count observations directly and supports batch-conditioned expression estimates and reference-to-query transfer.","source_ids":["evidence-official-550f6719c2f22f4d57b3","evidence-official-0cbab80c2d1769287a1f"],"source_locator":"docs/user_guide/models/scvi.md: Preliminaries, Generative process, Inference and Limitations"}],"limitations":[{"text":"The documentation notes that the latent space is less interpretable than a linear method and efficient inference generally benefits from a GPU. Covariates and likelihood must be reported.","source_ids":["evidence-official-550f6719c2f22f4d57b3","evidence-official-0cbab80c2d1769287a1f"],"source_locator":"docs/user_guide/models/scvi.md: Preliminaries, Generative process, Inference and Limitations"}],"diagram":{"title":"scVI workflow","steps":["RNA counts and covariates","Variational encoder","Latent cell state","Count decoder and estimates"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-550f6719c2f22f4d57b3","evidence-official-0cbab80c2d1769287a1f"],"source_locator":"docs/user_guide/models/scvi.md: Preliminaries, Generative process, Inference and Limitations"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}},"description":"Probabilistic single-cell expression model","facets":{"areas":["single-cell"]},"id":"discovery-model-scvi","kind":"model","links":[{"relation":"applicable_to","target_id":"discovery-benchmark-scib"}],"name":"scVI","source_ids":["src-discovery-scverse-scvi-tools"],"status":"discovered"}
{"attributes":{"access":"official_source_linked","benchmark_applicability":"candidate; not evidence of a reported evaluation","candidate_benchmark_ids":[],"entity_level":"family","reported_name":"SegmentNT","version":null,"historical_missing_metadata":{"checkpoint":"unextracted","code_licence":"unextracted","parameters":"unextracted","training_cutoff":"unextracted","training_data":"unextracted","version":"unextracted","weights_licence":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"SegmentNT labels genomic elements at individual nucleotide positions using a pretrained DNA backbone.","summary_source_ids":["evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-7e59c59bfea79722fc33","evidence-official-97071500fc3422c426d4","evidence-official-da4566a88ed9fa42fdcb"],"summary_source_locator":"SegmentNT paper Methods: SegmentNT architecture, Model ablations, Human genomic elements and Multispecies training; official model-card Training data and licence metadata","sections":[{"title":"How it works","body":"SegmentNT labels genomic elements at individual nucleotide positions using a pretrained DNA backbone. Nucleotide Transformer backbone with a one-dimensional U-Net segmentation head; YaRN rescales positions for longer inputs. The documented inputs are DNA sequences without N bases, tokenized into 6-mers under the documented length constraints. The output consists of per-base probabilities for 14 genomic-element classes.","source_ids":["evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-7e59c59bfea79722fc33","evidence-official-97071500fc3422c426d4","evidence-official-da4566a88ed9fa42fdcb"],"source_locator":"SegmentNT paper Methods: SegmentNT architecture, Model ablations, Human genomic elements and Multispecies training; official model-card Training data and licence metadata"},{"title":"Versions and reproducibility","body":"segment_nt and segment_nt_multi_species; SegmentEnformer and SegmentBorzoi are distinct pipelines. Trained on 30kb; inference up to 50kb requires the documented rescaling. Input token count also has divisibility constraints.","source_ids":["evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-7e59c59bfea79722fc33","evidence-official-97071500fc3422c426d4","evidence-official-da4566a88ed9fa42fdcb"],"source_locator":"SegmentNT paper Methods: SegmentNT architecture, Model ablations, Human genomic elements and Multispecies training; official model-card Training data and licence metadata"}],"facts":[{"label":"Model type","value":"DNA encoder with nucleotide-level segmentation head","status":"source_checked","source_ids":["evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-7e59c59bfea79722fc33","evidence-official-97071500fc3422c426d4","evidence-official-da4566a88ed9fa42fdcb"],"source_locator":"SegmentNT paper Methods: SegmentNT architecture, Model ablations, Human genomic elements and Multispecies training; official model-card Training data and licence metadata"},{"label":"Architecture","value":"Nucleotide Transformer backbone with a one-dimensional U-Net segmentation head; YaRN rescales positions for longer inputs.","status":"source_checked","source_ids":["evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-7e59c59bfea79722fc33","evidence-official-97071500fc3422c426d4","evidence-official-da4566a88ed9fa42fdcb"],"source_locator":"SegmentNT paper Methods: SegmentNT architecture, Model ablations, Human genomic elements and Multispecies training; official model-card Training data and licence metadata"},{"label":"Inputs","value":"DNA sequences without N bases, tokenized into 6-mers under the documented length constraints.","status":"source_checked","source_ids":["evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-7e59c59bfea79722fc33","evidence-official-97071500fc3422c426d4","evidence-official-da4566a88ed9fa42fdcb"],"source_locator":"SegmentNT paper Methods: SegmentNT architecture, Model ablations, Human genomic elements and Multispecies training; official model-card Training data and licence metadata"},{"label":"Outputs","value":"Per-base probabilities for 14 genomic-element classes.","status":"source_checked","source_ids":["evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-7e59c59bfea79722fc33","evidence-official-97071500fc3422c426d4","evidence-official-da4566a88ed9fa42fdcb"],"source_locator":"SegmentNT paper Methods: SegmentNT architecture, Model ablations, Human genomic elements and Multispecies training; official model-card Training data and licence metadata"},{"label":"Parameters","value":"563M for NT-v2 500M plus the 63M segmentation head; alternative encoder ablations are different complete pipelines.","status":"source_checked","source_ids":["evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-7e59c59bfea79722fc33","evidence-official-97071500fc3422c426d4","evidence-official-da4566a88ed9fa42fdcb"],"source_locator":"SegmentNT paper Methods: SegmentNT architecture, Model ablations, Human genomic elements and Multispecies training; official model-card Training data and licence metadata"},{"label":"Known versions","value":"segment_nt and segment_nt_multi_species; SegmentEnformer and SegmentBorzoi are distinct pipelines.","status":"source_checked","source_ids":["evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-7e59c59bfea79722fc33","evidence-official-97071500fc3422c426d4","evidence-official-da4566a88ed9fa42fdcb"],"source_locator":"SegmentNT paper Methods: SegmentNT architecture, Model ablations, Human genomic elements and Multispecies training; official model-card Training data and licence metadata"},{"label":"Training data","value":"Human 14-element segmentation labels from GENCODE V44 and ENCODE SCREEN/DHS annotations. Human chromosomes 20 and 21 are held out for testing and 22 for validation. A separate multispecies model adds mouse, chicken, fly, zebrafish and worm.","status":"source_checked","source_ids":["evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-7e59c59bfea79722fc33","evidence-official-97071500fc3422c426d4","evidence-official-da4566a88ed9fa42fdcb"],"source_locator":"SegmentNT paper Methods: SegmentNT architecture, Model ablations, Human genomic elements and Multispecies training; official model-card Training data and licence metadata"},{"label":"Training cutoff","value":"GENCODE V44 and the named ENCODE SCREEN/DHS resources define label provenance. The inspected paper does not give one latest-experiment date shared by every resource.","status":"source_checked","source_ids":["evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-7e59c59bfea79722fc33","evidence-official-97071500fc3422c426d4","evidence-official-da4566a88ed9fa42fdcb"],"source_locator":"SegmentNT paper Methods: SegmentNT architecture, Model ablations, Human genomic elements and Multispecies training; official model-card Training data and licence metadata"},{"label":"Context limits","value":"Trained on 30kb; inference up to 50kb requires the documented rescaling. Input token count also has divisibility constraints.","status":"source_checked","source_ids":["evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-7e59c59bfea79722fc33","evidence-official-97071500fc3422c426d4","evidence-official-da4566a88ed9fa42fdcb"],"source_locator":"SegmentNT paper Methods: SegmentNT architecture, Model ablations, Human genomic elements and Multispecies training; official model-card Training data and licence metadata"},{"label":"Weights licence","value":"CC-BY-NC-SA-4.0 declared by the official InstaDeepAI/segment_nt model card.","status":"source_checked","source_ids":["evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-7e59c59bfea79722fc33","evidence-official-97071500fc3422c426d4","evidence-official-da4566a88ed9fa42fdcb"],"source_locator":"SegmentNT paper Methods: SegmentNT architecture, Model ablations, Human genomic elements and Multispecies training; official model-card Training data and licence metadata"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/instadeepai/nucleotide-transformer","status":"source_checked","source_ids":["evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-7e59c59bfea79722fc33","evidence-official-97071500fc3422c426d4","evidence-official-da4566a88ed9fa42fdcb"],"source_locator":"SegmentNT paper Methods: SegmentNT architecture, Model ablations, Human genomic elements and Multispecies training; official model-card Training data and licence metadata"},{"label":"Code licence","value":"CC-BY-NC-SA-4.0","status":"source_checked","source_ids":["evidence-official-7e4b193e47ba209860a1"],"source_locator":"LICENSE.md: licence text"}],"strengths":[{"text":"Links sequence representation learning to explicitly localized annotations such as exons, splice sites and regulatory elements.","source_ids":["evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-7e59c59bfea79722fc33","evidence-official-97071500fc3422c426d4","evidence-official-da4566a88ed9fa42fdcb"],"source_locator":"SegmentNT paper Methods: SegmentNT architecture, Model ablations, Human genomic elements and Multispecies training; official model-card Training data and licence metadata"}],"limitations":[{"text":"The documented implementation does not accept N bases. Its 30kb training context and reported 50kb generalization should not be treated as unlimited context.","source_ids":["evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-7e59c59bfea79722fc33","evidence-official-97071500fc3422c426d4","evidence-official-da4566a88ed9fa42fdcb"],"source_locator":"SegmentNT paper Methods: SegmentNT architecture, Model ablations, Human genomic elements and Multispecies training; official model-card Training data and licence metadata"}],"diagram":{"title":"SegmentNT workflow","steps":["DNA sequence","NT backbone with position rescaling","1D U-Net head","Per-base element predictions"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-18d8f4d5fc3f6616922d","evidence-official-657e83427ab59f3aec83","evidence-official-56f02d45976d011d80aa","evidence-official-4920952f9b3c4b8909a0","evidence-official-7e59c59bfea79722fc33","evidence-official-97071500fc3422c426d4","evidence-official-da4566a88ed9fa42fdcb"],"source_locator":"SegmentNT paper Methods: SegmentNT architecture, Model ablations, Human genomic elements and Multispecies training; official model-card Training data and licence metadata"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}},"description":"Genomic sequence segmentation model","facets":{"areas":["genomics"]},"id":"discovery-model-segmentnt","kind":"model","links":[],"name":"SegmentNT","source_ids":["src-discovery-instadeepai-nucleotide-transformer"],"status":"discovered"}
{"attributes":{"access":"official_source_linked","benchmark_applicability":"candidate; not evidence of a reported evaluation","candidate_benchmark_ids":[],"entity_level":"family","reported_name":"SpliceAI","version":null,"historical_missing_metadata":{"checkpoint":"unextracted","code_licence":"unextracted","parameters":"unextracted","training_cutoff":"unextracted","training_data":"unextracted","version":"unextracted","weights_licence":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"SpliceAI annotates sequence variants with predicted splice acceptor and donor changes.","summary_source_ids":["evidence-official-9b820532ba8e3965f64e","evidence-official-ded281404bd1a0f3fdb7"],"summary_source_locator":"Jaganathan et al., Cell 2019, STAR Methods: SpliceAI architecture and Model training and testing (pp. e2–e3); README.md: usage and License","sections":[{"title":"How it works","body":"SpliceAI reads one-hot-encoded DNA through dilated convolutional residual blocks. Skip connections combine features at different depths, and a softmax layer assigns acceptor, donor or neither probabilities to the central positions. The 10kb version requires 5kb of sequence on each side of a scored position; variant scoring compares the reference and alternate predictions.","source_ids":["evidence-official-9b820532ba8e3965f64e","evidence-official-ded281404bd1a0f3fdb7"],"source_locator":"Jaganathan et al., Cell 2019, STAR Methods: SpliceAI architecture and Model training and testing (pp. e2–e3); README.md: usage and License"},{"title":"Versions and reproducibility","body":"The paper studies 80nt, 400nt, 2kb and 10kb receptive spans. Variant scoring averages five independently trained models; these are not five different assay results. SpliceAI-10k uses 5,000 flanking bases on each side. An input of length l + 10,000 produces predictions for l central positions; receptive span is distinct from maximum input length.","source_ids":["evidence-official-9b820532ba8e3965f64e","evidence-official-ded281404bd1a0f3fdb7"],"source_locator":"Jaganathan et al., Cell 2019, STAR Methods: SpliceAI architecture and Model training and testing (pp. e2–e3); README.md: usage and License"}],"facts":[{"label":"Model type","value":"Dilated convolutional splicing predictor","status":"source_checked","source_ids":["evidence-official-9b820532ba8e3965f64e","evidence-official-ded281404bd1a0f3fdb7"],"source_locator":"Jaganathan et al., Cell 2019, STAR Methods: SpliceAI architecture and Model training and testing (pp. e2–e3); README.md: usage and License"},{"label":"Architecture","value":"Residual one-dimensional convolutional network with dilated kernels and skip connections; a softmax head predicts acceptor, donor and neither at each central position.","status":"source_checked","source_ids":["evidence-official-9b820532ba8e3965f64e","evidence-official-ded281404bd1a0f3fdb7"],"source_locator":"Jaganathan et al., Cell 2019, STAR Methods: SpliceAI architecture and Model training and testing (pp. e2–e3); README.md: usage and License"},{"label":"Inputs","value":"VCF variants, reference FASTA and matching gene annotation, or custom one-hot-encoded sequence.","status":"source_checked","source_ids":["evidence-official-9b820532ba8e3965f64e","evidence-official-ded281404bd1a0f3fdb7"],"source_locator":"Jaganathan et al., Cell 2019, STAR Methods: SpliceAI architecture and Model training and testing (pp. e2–e3); README.md: usage and License"},{"label":"Outputs","value":"Acceptor/donor gain/loss scores and positions in VCF INFO annotations.","status":"source_checked","source_ids":["evidence-official-9b820532ba8e3965f64e","evidence-official-ded281404bd1a0f3fdb7"],"source_locator":"Jaganathan et al., Cell 2019, STAR Methods: SpliceAI architecture and Model training and testing (pp. e2–e3); README.md: usage and License"},{"label":"Parameters","value":"The original STAR Methods specifies residual blocks, dilation and receptive spans, but does not state a complete parameter count for each released five-model scoring ensemble.","status":"unreported","source_ids":["evidence-official-9b820532ba8e3965f64e","evidence-official-ded281404bd1a0f3fdb7"],"source_locator":"Jaganathan et al., Cell 2019, STAR Methods: SpliceAI architecture and Model training and testing (pp. e2–e3); README.md: usage and License"},{"label":"Known versions","value":"The paper studies 80nt, 400nt, 2kb and 10kb receptive spans. Variant scoring averages five independently trained models; these are not five different assay results.","status":"source_checked","source_ids":["evidence-official-9b820532ba8e3965f64e","evidence-official-ded281404bd1a0f3fdb7"],"source_locator":"Jaganathan et al., Cell 2019, STAR Methods: SpliceAI architecture and Model training and testing (pp. e2–e3); README.md: usage and License"},{"label":"Training data","value":"Human GRCh37 sequence and GENCODE V24lift37 principal protein-coding transcripts, split by chromosome with non-paralogous held-out test genes. The paper distinguishes GENCODE-only training from GTEx-junction-augmented models used for variant analyses.","status":"source_checked","source_ids":["evidence-official-9b820532ba8e3965f64e","evidence-official-ded281404bd1a0f3fdb7"],"source_locator":"Jaganathan et al., Cell 2019, STAR Methods: SpliceAI architecture and Model training and testing (pp. e2–e3); README.md: usage and License"},{"label":"Training cutoff","value":"GENCODE V24lift37 on GRCh37 defines the documented transcript annotations. GTEx-augmented training is separately described; the paper does not give one common latest-data date for both variants.","status":"source_checked","source_ids":["evidence-official-9b820532ba8e3965f64e","evidence-official-ded281404bd1a0f3fdb7"],"source_locator":"Jaganathan et al., Cell 2019, STAR Methods: SpliceAI architecture and Model training and testing (pp. e2–e3); README.md: usage and License"},{"label":"Context limits","value":"SpliceAI-10k uses 5,000 flanking bases on each side. An input of length l + 10,000 produces predictions for l central positions; receptive span is distinct from maximum input length.","status":"source_checked","source_ids":["evidence-official-9b820532ba8e3965f64e","evidence-official-ded281404bd1a0f3fdb7"],"source_locator":"Jaganathan et al., Cell 2019, STAR Methods: SpliceAI architecture and Model training and testing (pp. e2–e3); README.md: usage and License"},{"label":"Weights licence","value":"CC-BY-NC-4.0 for trained models; commercial use requires a separate licence. Code is PolyForm Strict 1.0.0.","status":"source_checked","source_ids":["evidence-official-9b820532ba8e3965f64e","evidence-official-ded281404bd1a0f3fdb7"],"source_locator":"Jaganathan et al., Cell 2019, STAR Methods: SpliceAI architecture and Model training and testing (pp. e2–e3); README.md: usage and License"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/illumina/SpliceAI","status":"source_checked","source_ids":["evidence-official-9b820532ba8e3965f64e","evidence-official-ded281404bd1a0f3fdb7"],"source_locator":"Jaganathan et al., Cell 2019, STAR Methods: SpliceAI architecture and Model training and testing (pp. e2–e3); README.md: usage and License"},{"label":"Code licence","value":"PolyForm Strict 1.0.0 for code; trained weights have separate terms.","status":"source_checked","source_ids":["evidence-official-9a47f0326b33870b3113"],"source_locator":"LICENSE: licence text"}],"strengths":[{"text":"Provides direct sequence inference and an annotation workflow with explicit genome and distance settings.","source_ids":["evidence-official-9b820532ba8e3965f64e","evidence-official-ded281404bd1a0f3fdb7"],"source_locator":"Jaganathan et al., Cell 2019, STAR Methods: SpliceAI architecture and Model training and testing (pp. e2–e3); README.md: usage and License"}],"limitations":[{"text":"The command-line pipeline skips unsupported variants and variants outside its gene annotations. Code, model weights and downloadable precomputed scores have distinct licensing provisions.","source_ids":["evidence-official-9b820532ba8e3965f64e","evidence-official-ded281404bd1a0f3fdb7"],"source_locator":"Jaganathan et al., Cell 2019, STAR Methods: SpliceAI architecture and Model training and testing (pp. e2–e3); README.md: usage and License"}],"diagram":{"title":"SpliceAI workflow","steps":["Variant plus sequence context","Reference and alternate predictions","Splice-site differences","Gain/loss annotations"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-9b820532ba8e3965f64e","evidence-official-ded281404bd1a0f3fdb7"],"source_locator":"Jaganathan et al., Cell 2019, STAR Methods: SpliceAI architecture and Model training and testing (pp. e2–e3); README.md: usage and License"},"coverage":"limited","gaps":["Parameters: The original STAR Methods specifies residual blocks, dilation and receptive spans, but does not state a complete parameter count for each released five-model scoring ensemble."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}},"description":"Splicing effect prediction","facets":{"areas":["genomics"]},"id":"discovery-model-spliceai","kind":"model","links":[],"name":"SpliceAI","source_ids":["src-discovery-illumina-spliceai"],"status":"discovered"}
{"attributes":{"access":"official_source_linked","benchmark_applicability":"candidate; not evidence of a reported evaluation","candidate_benchmark_ids":["discovery-benchmark-virtual-cell-challenge-2026"],"entity_level":"family","reported_name":"STATE","version":null,"historical_missing_metadata":{"checkpoint":"unextracted","code_licence":"unextracted","parameters":"unextracted","training_cutoff":"unextracted","training_data":"unextracted","version":"unextracted","weights_licence":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"State separates cellular representation learning from prediction of responses to perturbation.","summary_source_ids":["evidence-official-1df5b1865861177c0c75"],"summary_source_locator":"README.md: Getting started, State Transition Model, ST TOML configuration and Licenses","sections":[{"title":"How it works","body":"State separates cellular representation learning from prediction of responses to perturbation. State Embedding and State Transition are separate components; the documented transition workflow trains on specified expression features and perturbation metadata. The documented inputs are annData expression measurements, gene/cell-type labels and an explicit dataset/split configuration. The output consists of cell embeddings or predicted perturbed-expression matrices, depending on the selected component.","source_ids":["evidence-official-1df5b1865861177c0c75"],"source_locator":"README.md: Getting started, State Transition Model, ST TOML configuration and Licenses"},{"title":"Versions and reproducibility","body":"State Embedding (SE) and State Transition (ST), with separately trained checkpoints. The applicable input limits require configuration-specific checking.","source_ids":["evidence-official-1df5b1865861177c0c75"],"source_locator":"README.md: Getting started, State Transition Model, ST TOML configuration and Licenses"}],"facts":[{"label":"Model type","value":"Cell representation and perturbation-transition model family","status":"source_checked","source_ids":["evidence-official-1df5b1865861177c0c75"],"source_locator":"README.md: Getting started, State Transition Model, ST TOML configuration and Licenses"},{"label":"Architecture","value":"State Embedding and State Transition are separate components; the documented transition workflow trains on specified expression features and perturbation metadata.","status":"source_checked","source_ids":["evidence-official-1df5b1865861177c0c75"],"source_locator":"README.md: Getting started, State Transition Model, ST TOML configuration and Licenses"},{"label":"Inputs","value":"AnnData expression measurements, gene/cell-type labels and an explicit dataset/split configuration.","status":"source_checked","source_ids":["evidence-official-1df5b1865861177c0c75"],"source_locator":"README.md: Getting started, State Transition Model, ST TOML configuration and Licenses"},{"label":"Outputs","value":"Cell embeddings or predicted perturbed-expression matrices, depending on the selected component.","status":"source_checked","source_ids":["evidence-official-1df5b1865861177c0c75"],"source_locator":"README.md: Getting started, State Transition Model, ST TOML configuration and Licenses"},{"label":"Parameters","value":"The repository contains separate State Embedding and State Transition configurations; the selected complete pipeline is required before a parameter total can be assigned.","status":"unreported","source_ids":["evidence-official-1df5b1865861177c0c75"],"source_locator":"README.md: Getting started, State Transition Model, ST TOML configuration and Licenses"},{"label":"Known versions","value":"State Embedding (SE) and State Transition (ST), with separately trained checkpoints.","status":"source_checked","source_ids":["evidence-official-1df5b1865861177c0c75"],"source_locator":"README.md: Getting started, State Transition Model, ST TOML configuration and Licenses"},{"label":"Training data","value":"Examples include Replogle–Nadig genetic perturbations and Tahoe-100M; the actual trained model depends on its configuration.","status":"source_checked","source_ids":["evidence-official-1df5b1865861177c0c75"],"source_locator":"README.md: Getting started, State Transition Model, ST TOML configuration and Licenses"},{"label":"Training cutoff","value":"The reviewed README and training configuration do not provide one latest-data date shared by all State Embedding and task-fitted State Transition releases.","status":"unreported","source_ids":["evidence-official-1df5b1865861177c0c75"],"source_locator":"README.md: Getting started, State Transition Model, ST TOML configuration and Licenses"},{"label":"Context limits","value":"The transition workflow operates on the chosen expression features and perturbation metadata; the inspected family documentation does not define one universal gene/cell token budget.","status":"unreported","source_ids":["evidence-official-1df5b1865861177c0c75"],"source_locator":"README.md: Getting started, State Transition Model, ST TOML configuration and Licenses"},{"label":"Weights licence","value":"Arc Research Institute State Model Non-Commercial License, with Acceptable Use Policy; separate from code CC-BY-NC-SA-4.0.","status":"source_checked","source_ids":["evidence-official-1df5b1865861177c0c75"],"source_locator":"README.md: Getting started, State Transition Model, ST TOML configuration and Licenses"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/ArcInstitute/state","status":"source_checked","source_ids":["evidence-official-1df5b1865861177c0c75"],"source_locator":"README.md: Getting started, State Transition Model, ST TOML configuration and Licenses"},{"label":"Code licence","value":"CC-BY-NC-SA-4.0","status":"source_checked","source_ids":["evidence-official-ed0382dc026ea7080abd"],"source_locator":"LICENSE: licence text"}],"strengths":[{"text":"The command-line workflow makes training, inference and evaluation configurations explicit.","source_ids":["evidence-official-1df5b1865861177c0c75"],"source_locator":"README.md: Getting started, State Transition Model, ST TOML configuration and Licenses"}],"limitations":[{"text":"State Embedding and State Transition are not interchangeable model identities. Data omitted from zero-shot/few-shot split declarations default to training in the documented configuration format.","source_ids":["evidence-official-1df5b1865861177c0c75"],"source_locator":"README.md: Getting started, State Transition Model, ST TOML configuration and Licenses"}],"diagram":{"title":"STATE workflow","steps":["Expression and perturbation metadata","Selected embedding or transition component","Configured inference","Embeddings or predicted response"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-1df5b1865861177c0c75"],"source_locator":"README.md: Getting started, State Transition Model, ST TOML configuration and Licenses"},"coverage":"limited","gaps":["Parameters: The repository contains separate State Embedding and State Transition configurations; the selected complete pipeline is required before a parameter total can be assigned.","Training cutoff: The reviewed README and training configuration do not provide one latest-data date shared by all State Embedding and task-fitted State Transition releases.","Context limits: The transition workflow operates on the chosen expression features and perturbation metadata; the inspected family documentation does not define one universal gene/cell token budget."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}},"description":"Cell state and perturbation modelling","facets":{"areas":["single-cell"]},"id":"discovery-model-state","kind":"model","links":[{"relation":"applicable_to","target_id":"discovery-benchmark-virtual-cell-challenge-2026"}],"name":"STATE","source_ids":["src-discovery-arcinstitute-state"],"status":"discovered"}
{"attributes":{"access":"official_source_linked","benchmark_applicability":"candidate; not evidence of a reported evaluation","candidate_benchmark_ids":["discovery-benchmark-glycanml"],"entity_level":"family","reported_name":"SweetNet","version":null,"historical_missing_metadata":{"checkpoint":"unextracted","code_licence":"unextracted","parameters":"unextracted","training_cutoff":"unextracted","training_data":"unextracted","version":"unextracted","weights_licence":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"SweetNet predicts glycan properties and produces learned representations from glycan graphs.","summary_source_ids":["evidence-official-7c43b700aa87c22843a6","evidence-official-2bf115f7d072744af00a"],"summary_source_locator":"glycowork/ml/models.py: SweetNet.__init__ and forward; README.md: pretrained SweetNet description","sections":[{"title":"How it works","body":"SweetNet predicts glycan properties and produces learned representations from glycan graphs. Graph convolutional network; the inspected implementation has three graph-convolution layers, global mean pooling and fully connected prediction layers. The documented inputs are tokenized glycan graph nodes and glycosidic connectivity. The output consists of property predictions and optional intermediate glycan representations.","source_ids":["evidence-official-7c43b700aa87c22843a6","evidence-official-2bf115f7d072744af00a"],"source_locator":"glycowork/ml/models.py: SweetNet.__init__ and forward; README.md: pretrained SweetNet description"},{"title":"Versions and reproducibility","body":"SweetNet class in the pinned glycowork revision; checkpoint/task identity remains separate. The applicable input limits require configuration-specific checking.","source_ids":["evidence-official-7c43b700aa87c22843a6","evidence-official-2bf115f7d072744af00a"],"source_locator":"glycowork/ml/models.py: SweetNet.__init__ and forward; README.md: pretrained SweetNet description"}],"facts":[{"label":"Model type","value":"Glycan graph convolutional network","status":"source_checked","source_ids":["evidence-official-7c43b700aa87c22843a6","evidence-official-2bf115f7d072744af00a"],"source_locator":"glycowork/ml/models.py: SweetNet.__init__ and forward; README.md: pretrained SweetNet description"},{"label":"Architecture","value":"Graph convolutional network; the inspected implementation has three graph-convolution layers, global mean pooling and fully connected prediction layers.","status":"source_checked","source_ids":["evidence-official-7c43b700aa87c22843a6","evidence-official-2bf115f7d072744af00a"],"source_locator":"glycowork/ml/models.py: SweetNet.__init__ and forward; README.md: pretrained SweetNet description"},{"label":"Inputs","value":"Tokenized glycan graph nodes and glycosidic connectivity.","status":"source_checked","source_ids":["evidence-official-7c43b700aa87c22843a6","evidence-official-2bf115f7d072744af00a"],"source_locator":"glycowork/ml/models.py: SweetNet.__init__ and forward; README.md: pretrained SweetNet description"},{"label":"Outputs","value":"Property predictions and optional intermediate glycan representations.","status":"source_checked","source_ids":["evidence-official-7c43b700aa87c22843a6","evidence-official-2bf115f7d072744af00a"],"source_locator":"glycowork/ml/models.py: SweetNet.__init__ and forward; README.md: pretrained SweetNet description"},{"label":"Parameters","value":"Depends on vocabulary size, hidden dimension and output classes; default hidden dimension in the inspected class is 128.","status":"source_checked","source_ids":["evidence-official-7c43b700aa87c22843a6","evidence-official-2bf115f7d072744af00a"],"source_locator":"glycowork/ml/models.py: SweetNet.__init__ and forward; README.md: pretrained SweetNet description"},{"label":"Known versions","value":"SweetNet class in the pinned glycowork revision; checkpoint/task identity remains separate.","status":"source_checked","source_ids":["evidence-official-7c43b700aa87c22843a6","evidence-official-2bf115f7d072744af00a"],"source_locator":"glycowork/ml/models.py: SweetNet.__init__ and forward; README.md: pretrained SweetNet description"},{"label":"Training data","value":"The inspected repository describes pretrained glycan-to-species prediction, but does not identify the exact training snapshot for that downloadable model.","status":"unreported","source_ids":["evidence-official-7c43b700aa87c22843a6","evidence-official-2bf115f7d072744af00a"],"source_locator":"glycowork/ml/models.py: SweetNet.__init__ and forward; README.md: pretrained SweetNet description"},{"label":"Training cutoff","value":"The reviewed pretrained-model documentation does not state the last-included glycan or species annotation date for the checkpoint.","status":"unreported","source_ids":["evidence-official-7c43b700aa87c22843a6","evidence-official-2bf115f7d072744af00a"],"source_locator":"glycowork/ml/models.py: SweetNet.__init__ and forward; README.md: pretrained SweetNet description"},{"label":"Context limits","value":"SweetNet pools variable-size glycan graphs. The inspected model implementation does not declare one validated maximum graph size for the pretrained checkpoint.","status":"unreported","source_ids":["evidence-official-7c43b700aa87c22843a6","evidence-official-2bf115f7d072744af00a"],"source_locator":"glycowork/ml/models.py: SweetNet.__init__ and forward; README.md: pretrained SweetNet description"},{"label":"Weights licence","value":"Separate checkpoint-distribution terms are not stated in the inspected release documentation and licence material. The source-code licence alone is not recorded as an explicit weight grant.","status":"unreported","source_ids":["evidence-official-7c43b700aa87c22843a6","evidence-official-2bf115f7d072744af00a","evidence-official-6bcab1b3e31b52c38ea5"],"source_locator":"glycowork/ml/models.py: SweetNet.__init__ and forward; README.md: pretrained SweetNet description; LICENSE: licence text"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/BojarLab/glycowork","status":"source_checked","source_ids":["evidence-official-7c43b700aa87c22843a6","evidence-official-2bf115f7d072744af00a"],"source_locator":"glycowork/ml/models.py: SweetNet.__init__ and forward; README.md: pretrained SweetNet description"},{"label":"Code licence","value":"MIT","status":"source_checked","source_ids":["evidence-official-6bcab1b3e31b52c38ea5"],"source_locator":"LICENSE: licence text"}],"strengths":[{"text":"Models branched glycan connectivity directly instead of treating each glycan only as a linear string.","source_ids":["evidence-official-7c43b700aa87c22843a6","evidence-official-2bf115f7d072744af00a"],"source_locator":"glycowork/ml/models.py: SweetNet.__init__ and forward; README.md: pretrained SweetNet description"}],"limitations":[{"text":"The package contains several other models, including LectinOracle, whose protein inputs must not be assigned to SweetNet. The class configuration and pretrained task identify the actual model.","source_ids":["evidence-official-7c43b700aa87c22843a6","evidence-official-2bf115f7d072744af00a"],"source_locator":"glycowork/ml/models.py: SweetNet.__init__ and forward; README.md: pretrained SweetNet description"}],"diagram":{"title":"SweetNet workflow","steps":["Glycan graph","Three graph convolutions","Graph pooling","Property prediction"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-7c43b700aa87c22843a6","evidence-official-2bf115f7d072744af00a"],"source_locator":"glycowork/ml/models.py: SweetNet.__init__ and forward; README.md: pretrained SweetNet description"},"coverage":"limited","gaps":["Training data: The inspected repository describes pretrained glycan-to-species prediction, but does not identify the exact training snapshot for that downloadable model.","Training cutoff: The reviewed pretrained-model documentation does not state the last-included glycan or species annotation date for the checkpoint.","Context limits: SweetNet pools variable-size glycan graphs. The inspected model implementation does not declare one validated maximum graph size for the pretrained checkpoint.","Weights licence: Separate checkpoint-distribution terms are not stated in the inspected release documentation and licence material. The source-code licence alone is not recorded as an explicit weight grant."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}},"description":"Glycan graph learning model","facets":{"areas":["glycomics"]},"id":"discovery-model-sweetnet","kind":"model","links":[{"relation":"applicable_to","target_id":"discovery-benchmark-glycanml"}],"name":"SweetNet","source_ids":["src-discovery-bojarlab-glycowork"],"status":"discovered"}
{"attributes":{"entity_level":"method","reported_name":"Bepler","version":null,"historical_missing_metadata":{"checkpoint":"unreported","version":"unreported"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"The TAPE Bepler comparison uses a protein representation that combines bidirectional language modelling with supervised structural pretraining.","summary_source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-116dfd8f61e04b51c148","evidence-official-362b9cef071111dfe343"],"summary_source_locator":"README.md: opening reproducibility warning and Leaderboard; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3","sections":[{"title":"How it works","body":"The TAPE Bepler comparison uses a protein representation that combines bidirectional language modelling with supervised structural pretraining. The June 2019 TAPE preprint describes a two-layer bidirectional language model followed by three 512-unit bidirectional LSTMs, with contact and remote-homology supervision. The documented inputs are protein sequences for the task-specific TAPE evaluation. The output consists of protein representations followed by the applicable TAPE prediction head.","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-116dfd8f61e04b51c148","evidence-official-362b9cef071111dfe343"],"source_locator":"README.md: opening reproducibility warning and Leaderboard; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"title":"Versions and reproducibility","body":"June 2019 TAPE preprint configurations and the later PyTorch package are distinct; the current README warns it is not an exact reproduction. The TAPE preprint uses sequence-length-dependent batching; the exact task/checkpoint determines padding, truncation and resource constraints.","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-116dfd8f61e04b51c148","evidence-official-362b9cef071111dfe343"],"source_locator":"README.md: opening reproducibility warning and Leaderboard; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"}],"facts":[{"label":"Model type","value":"Bidirectional recurrent protein representation with structural pretraining","status":"source_checked","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-116dfd8f61e04b51c148","evidence-official-362b9cef071111dfe343"],"source_locator":"README.md: opening reproducibility warning and Leaderboard; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Architecture","value":"The June 2019 TAPE preprint describes a two-layer bidirectional language model followed by three 512-unit bidirectional LSTMs, with contact and remote-homology supervision.","status":"source_checked","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-116dfd8f61e04b51c148","evidence-official-362b9cef071111dfe343"],"source_locator":"README.md: opening reproducibility warning and Leaderboard; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Inputs","value":"Protein sequences for the task-specific TAPE evaluation.","status":"source_checked","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-116dfd8f61e04b51c148","evidence-official-362b9cef071111dfe343"],"source_locator":"README.md: opening reproducibility warning and Leaderboard; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Outputs","value":"Protein representations followed by the applicable TAPE prediction head.","status":"source_checked","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-116dfd8f61e04b51c148","evidence-official-362b9cef071111dfe343"],"source_locator":"README.md: opening reproducibility warning and Leaderboard; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Parameters","value":"The paper describes the language model and structural LSTMs but the exact historical checkpoint remains unresolved; no single verified total is assigned.","status":"unreported","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-116dfd8f61e04b51c148","evidence-official-362b9cef071111dfe343"],"source_locator":"README.md: opening reproducibility warning and Leaderboard; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Known versions","value":"June 2019 TAPE preprint configurations and the later PyTorch package are distinct; the current README warns it is not an exact reproduction.","status":"source_checked","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-116dfd8f61e04b51c148","evidence-official-362b9cef071111dfe343"],"source_locator":"README.md: opening reproducibility warning and Leaderboard; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Training data","value":"The June 2019 TAPE preprint uses 31M Pfam domains, with held-out families and a separate random split; downstream task heads are fitted on their own task data.","status":"source_checked","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-116dfd8f61e04b51c148","evidence-official-362b9cef071111dfe343"],"source_locator":"README.md: opening reproducibility warning and Leaderboard; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Training cutoff","value":"The June 2019 TAPE preprint identifies the Pfam-domain corpus and split procedure but does not state one latest-sequence deposition date. Later package defaults are a different implementation.","status":"unreported","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-116dfd8f61e04b51c148","evidence-official-362b9cef071111dfe343"],"source_locator":"README.md: opening reproducibility warning and Leaderboard; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Context limits","value":"The TAPE preprint uses sequence-length-dependent batching; the exact task/checkpoint determines padding, truncation and resource constraints.","status":"source_checked","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-116dfd8f61e04b51c148","evidence-official-362b9cef071111dfe343"],"source_locator":"README.md: opening reproducibility warning and Leaderboard; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Weights licence","value":"Separate checkpoint-distribution terms are not stated in the inspected release documentation and licence material. The source-code licence alone is not recorded as an explicit weight grant.","status":"unreported","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-116dfd8f61e04b51c148","evidence-official-362b9cef071111dfe343","evidence-official-e847949b10d83e3b0efe"],"source_locator":"README.md: opening reproducibility warning and Leaderboard; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3; LICENSE: licence text"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/songlab-cal/tape","status":"source_checked","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-116dfd8f61e04b51c148","evidence-official-362b9cef071111dfe343"],"source_locator":"README.md: opening reproducibility warning and Leaderboard; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Code licence","value":"BSD-3-Clause","status":"source_checked","source_ids":["evidence-official-e847949b10d83e3b0efe"],"source_locator":"LICENSE: licence text"}],"strengths":[{"text":"Provides a reference for testing whether structural supervision during pretraining transfers to downstream protein tasks.","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-116dfd8f61e04b51c148","evidence-official-362b9cef071111dfe343"],"source_locator":"README.md: opening reproducibility warning and Leaderboard; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"}],"limitations":[{"text":"The preprint identifies the architecture and training objectives, but the exact historical checkpoint is still unresolved. Later PyTorch package defaults are not an exact reproduction.","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-116dfd8f61e04b51c148","evidence-official-362b9cef071111dfe343"],"source_locator":"README.md: opening reproducibility warning and Leaderboard; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"}],"diagram":{"title":"TAPE Bepler workflow","steps":["Protein sequence","Bidirectional language model","Three bidirectional LSTMs","Specified downstream task head"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-116dfd8f61e04b51c148","evidence-official-362b9cef071111dfe343"],"source_locator":"README.md: opening reproducibility warning and Leaderboard; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},"coverage":"limited","gaps":["Parameters: The paper describes the language model and structural LSTMs but the exact historical checkpoint remains unresolved; no single verified total is assigned.","Training cutoff: The June 2019 TAPE preprint identifies the Pfam-domain corpus and split procedure but does not state one latest-sequence deposition date. Later package defaults are a different implementation.","Weights licence: Separate checkpoint-distribution terms are not stated in the inspected release documentation and licence material. The source-code licence alone is not recorded as an explicit weight grant.","The historical Bepler checkpoint/variant is unresolved; no architecture equivalence or result aggregation is asserted."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}},"description":"Method identity as printed in the historical TAPE leaderboard.","facets":{"areas":["protein-function"]},"id":"discovery-model-tape-bepler","kind":"model","links":[],"name":"TAPE Bepler","source_ids":["src-discovery-songlab-cal-tape"],"status":"discovered"}
{"attributes":{"entity_level":"method","reported_name":"LSTM","version":null,"historical_missing_metadata":{"checkpoint":"unreported","version":"unreported"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"The TAPE LSTM baseline represents protein sequences using recurrent networks that read residues in both directions.","summary_source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"summary_source_locator":"tape/models/modeling_lstm.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3","sections":[{"title":"How it works","body":"The TAPE LSTM baseline represents protein sequences using recurrent networks that read residues in both directions. Bidirectional recurrent protein encoder with three forward and three reverse LSTM layers in the default implementation. The documented inputs are amino-acid sequence in the tokenizer expected by the selected implementation. The output consists of residue/sequence representations and predictions from the chosen task head.","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_lstm.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"title":"Versions and reproducibility","body":"June 2019 TAPE preprint configurations and the later PyTorch package are distinct; the current README warns it is not an exact reproduction. The TAPE preprint uses sequence-length-dependent batching; the exact task/checkpoint determines padding, truncation and resource constraints.","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_lstm.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"}],"facts":[{"label":"Model type","value":"Bidirectional recurrent protein encoder","status":"source_checked","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_lstm.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Architecture","value":"Bidirectional recurrent protein encoder with three forward and three reverse LSTM layers in the default implementation.","status":"source_checked","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_lstm.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Inputs","value":"Amino-acid sequence in the tokenizer expected by the selected implementation.","status":"source_checked","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_lstm.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Outputs","value":"Residue/sequence representations and predictions from the chosen task head.","status":"source_checked","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_lstm.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Parameters","value":"Default input embedding size 128 and recurrent hidden size 1,024; these are dimensions, not parameter totals.","status":"source_checked","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_lstm.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Known versions","value":"June 2019 TAPE preprint configurations and the later PyTorch package are distinct; the current README warns it is not an exact reproduction.","status":"source_checked","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_lstm.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Training data","value":"The June 2019 TAPE preprint uses 31M Pfam domains, with held-out families and a separate random split; downstream task heads are fitted on their own task data.","status":"source_checked","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_lstm.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Training cutoff","value":"The June 2019 TAPE preprint identifies the Pfam-domain corpus and split procedure but does not state one latest-sequence deposition date. Later package defaults are a different implementation.","status":"unreported","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_lstm.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Context limits","value":"The TAPE preprint uses sequence-length-dependent batching; the exact task/checkpoint determines padding, truncation and resource constraints.","status":"source_checked","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_lstm.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Weights licence","value":"Separate checkpoint-distribution terms are not stated in the inspected release documentation and licence material. The source-code licence alone is not recorded as an explicit weight grant.","status":"unreported","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343","evidence-official-e847949b10d83e3b0efe"],"source_locator":"tape/models/modeling_lstm.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3; LICENSE: licence text"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/songlab-cal/tape","status":"source_checked","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_lstm.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Code licence","value":"BSD-3-Clause","status":"source_checked","source_ids":["evidence-official-e847949b10d83e3b0efe"],"source_locator":"LICENSE: licence text"}],"strengths":[{"text":"Provides a defined reference architecture that can be paired with the same downstream tasks as other protein representations.","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_lstm.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"}],"limitations":[{"text":"The maintainers explicitly state that the current PyTorch repository is not an exact reproduction of the original TensorFlow paper code; training maintenance was discontinued. Default input embedding size 128 and recurrent hidden size 1,024; these are dimensions, not parameter totals.","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_lstm.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"}],"diagram":{"title":"TAPE LSTM workflow","steps":["Protein sequence","Forward and reverse LSTMs","Residue/sequence representation","Task head"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_lstm.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},"coverage":"limited","gaps":["Training cutoff: The June 2019 TAPE preprint identifies the Pfam-domain corpus and split procedure but does not state one latest-sequence deposition date. Later package defaults are a different implementation.","Weights licence: Separate checkpoint-distribution terms are not stated in the inspected release documentation and licence material. The source-code licence alone is not recorded as an explicit weight grant."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}},"description":"Method identity as printed in the historical TAPE leaderboard.","facets":{"areas":["protein-function"]},"id":"discovery-model-tape-lstm","kind":"model","links":[],"name":"TAPE LSTM","source_ids":["src-discovery-songlab-cal-tape"],"status":"discovered"}
{"attributes":{"entity_level":"method","reported_name":"One Hot","version":null,"historical_missing_metadata":{"checkpoint":"unreported","version":"unreported"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"The TAPE one-hot baseline encodes each amino acid directly, allowing task performance to be assessed without a pretrained sequence representation.","summary_source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"summary_source_locator":"tape/models/modeling_onehot.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3","sections":[{"title":"How it works","body":"The TAPE one-hot baseline encodes each amino acid directly, allowing task performance to be assessed without a pretrained sequence representation. Direct one-hot residue representation followed by a separately fitted task head. The documented inputs are amino-acid sequence in the tokenizer expected by the selected implementation. The output consists of residue/sequence representations and predictions from the chosen task head.","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_onehot.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"title":"Versions and reproducibility","body":"June 2019 TAPE preprint configurations and the later PyTorch package are distinct; the current README warns it is not an exact reproduction. Variable-length residue input in the implementation; padding/masking and the downstream head determine the evaluated length handling.","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_onehot.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"}],"facts":[{"label":"Model type","value":"One-hot sequence representation and separate task head","status":"source_checked","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_onehot.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Architecture","value":"Direct one-hot residue representation followed by a separately fitted task head.","status":"source_checked","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_onehot.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Inputs","value":"Amino-acid sequence in the tokenizer expected by the selected implementation.","status":"source_checked","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_onehot.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Outputs","value":"Residue/sequence representations and predictions from the chosen task head.","status":"source_checked","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_onehot.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Parameters","value":"The encoding is parameter-free; the task-specific prediction head can still have learned parameters.","status":"source_checked","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_onehot.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Known versions","value":"June 2019 TAPE preprint configurations and the later PyTorch package are distinct; the current README warns it is not an exact reproduction.","status":"source_checked","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_onehot.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Training data","value":"The repository distributes a Pfam pretraining corpus and separate supervised task data. A specific checkpoint and adaptation must identify which training was actually used.","status":"source_checked","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_onehot.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Training cutoff","value":"Inapplicable to the fixed one-hot encoding; supervised task-head training dates belong to each evaluation.","status":"inapplicable","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_onehot.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Context limits","value":"Variable-length residue input in the implementation; padding/masking and the downstream head determine the evaluated length handling.","status":"source_checked","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_onehot.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Weights licence","value":"Inapplicable to the parameter-free residue encoding; a fitted downstream head requires its own checkpoint provenance and terms.","status":"inapplicable","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_onehot.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/songlab-cal/tape","status":"source_checked","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_onehot.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Code licence","value":"BSD-3-Clause","status":"source_checked","source_ids":["evidence-official-e847949b10d83e3b0efe"],"source_locator":"LICENSE: licence text"}],"strengths":[{"text":"Provides a defined reference architecture that can be paired with the same downstream tasks as other protein representations.","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_onehot.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"}],"limitations":[{"text":"The maintainers explicitly state that the current PyTorch repository is not an exact reproduction of the original TensorFlow paper code; training maintenance was discontinued. The encoding is parameter-free; the task-specific prediction head can still have learned parameters.","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_onehot.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"}],"diagram":{"title":"TAPE One Hot workflow","steps":["Protein sequence","One-hot residue vectors","Specified task head","Task predictions"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_onehot.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}},"description":"Method identity as printed in the historical TAPE leaderboard.","facets":{"areas":["protein-function"]},"id":"discovery-model-tape-one-hot","kind":"model","links":[],"name":"TAPE One Hot","source_ids":["src-discovery-songlab-cal-tape"],"status":"discovered"}
{"attributes":{"entity_level":"method","reported_name":"ResNet","version":null,"historical_missing_metadata":{"checkpoint":"unreported","version":"unreported"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"The TAPE ResNet baseline represents protein sequences using residual convolutional blocks before a task-specific prediction head.","summary_source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"summary_source_locator":"tape/models/modeling_resnet.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3","sections":[{"title":"How it works","body":"The TAPE ResNet baseline represents protein sequences using residual convolutional blocks before a task-specific prediction head. The June 2019 preprint uses 35 residual blocks with two dilated convolutions each, 256 filters and kernel width 9. The later PyTorch defaults use30 layers and width 512. The documented inputs are amino-acid sequence in the tokenizer expected by the selected implementation. The output consists of residue/sequence representations and predictions from the chosen task head.","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_resnet.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"title":"Versions and reproducibility","body":"June 2019 TAPE preprint configurations and the later PyTorch package are distinct; the current README warns it is not an exact reproduction. The TAPE preprint uses sequence-length-dependent batching; the exact task/checkpoint determines padding, truncation and resource constraints.","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_resnet.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"}],"facts":[{"label":"Model type","value":"Residual convolutional protein encoder","status":"source_checked","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_resnet.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Architecture","value":"The June 2019 preprint uses 35 residual blocks with two dilated convolutions each, 256 filters and kernel width 9. The later PyTorch defaults use30 layers and width 512.","status":"source_checked","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_resnet.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Inputs","value":"Amino-acid sequence in the tokenizer expected by the selected implementation.","status":"source_checked","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_resnet.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Outputs","value":"Residue/sequence representations and predictions from the chosen task head.","status":"source_checked","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_resnet.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Parameters","value":"The model also exposes separate downstream heads; count parameters for the exact selected head/configuration.","status":"source_checked","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_resnet.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Known versions","value":"June 2019 TAPE preprint configurations and the later PyTorch package are distinct; the current README warns it is not an exact reproduction.","status":"source_checked","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_resnet.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Training data","value":"The June 2019 TAPE preprint uses 31M Pfam domains, with held-out families and a separate random split; downstream task heads are fitted on their own task data.","status":"source_checked","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_resnet.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Training cutoff","value":"The June 2019 TAPE preprint identifies the Pfam-domain corpus and split procedure but does not state one latest-sequence deposition date. Later package defaults are a different implementation.","status":"unreported","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_resnet.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Context limits","value":"The TAPE preprint uses sequence-length-dependent batching; the exact task/checkpoint determines padding, truncation and resource constraints.","status":"source_checked","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_resnet.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Weights licence","value":"Separate checkpoint-distribution terms are not stated in the inspected release documentation and licence material. The source-code licence alone is not recorded as an explicit weight grant.","status":"unreported","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343","evidence-official-e847949b10d83e3b0efe"],"source_locator":"tape/models/modeling_resnet.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3; LICENSE: licence text"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/songlab-cal/tape","status":"source_checked","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_resnet.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Code licence","value":"BSD-3-Clause","status":"source_checked","source_ids":["evidence-official-e847949b10d83e3b0efe"],"source_locator":"LICENSE: licence text"}],"strengths":[{"text":"Provides a defined reference architecture that can be paired with the same downstream tasks as other protein representations.","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_resnet.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"}],"limitations":[{"text":"The maintainers explicitly state that the current PyTorch repository is not an exact reproduction of the original TensorFlow paper code; training maintenance was discontinued. The model also exposes separate downstream heads; count parameters for the exact selected head/configuration.","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_resnet.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"}],"diagram":{"title":"TAPE ResNet workflow","steps":["Protein sequence","Residue embeddings","Residual convolutions","Task head"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_resnet.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},"coverage":"limited","gaps":["Training cutoff: The June 2019 TAPE preprint identifies the Pfam-domain corpus and split procedure but does not state one latest-sequence deposition date. Later package defaults are a different implementation.","Weights licence: Separate checkpoint-distribution terms are not stated in the inspected release documentation and licence material. The source-code licence alone is not recorded as an explicit weight grant."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}},"description":"Method identity as printed in the historical TAPE leaderboard.","facets":{"areas":["protein-function"]},"id":"discovery-model-tape-resnet","kind":"model","links":[],"name":"TAPE ResNet","source_ids":["src-discovery-songlab-cal-tape"],"status":"discovered"}
{"attributes":{"entity_level":"method","reported_name":"Transformer","version":null,"historical_missing_metadata":{"checkpoint":"unreported","version":"unreported"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"The TAPE Transformer learns contextual protein representations using masked-residue pretraining and a task-specific prediction head.","summary_source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"summary_source_locator":"tape/models/modeling_bert.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3","sections":[{"title":"How it works","body":"The TAPE Transformer learns contextual protein representations using masked-residue pretraining and a task-specific prediction head. The June 2019 preprint uses 12 transformer layers, width 512 and eight heads (38M parameters). The later PyTorch BertConfig defaults to width 768 and 12 heads; these are different configurations. The documented inputs are amino-acid sequence in the tokenizer expected by the selected implementation. The output consists of residue/sequence representations and predictions from the chosen task head.","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_bert.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"title":"Versions and reproducibility","body":"June 2019 TAPE preprint configurations and the later PyTorch package are distinct; the current README warns it is not an exact reproduction. The inspected default BertConfig sets max_position_embeddings to 8,096; this is an implementation default, not evidence that a historical TAPE checkpoint was trained at that length.","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_bert.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"}],"facts":[{"label":"Model type","value":"BERT-style protein transformer encoder","status":"source_checked","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_bert.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Architecture","value":"The June 2019 preprint uses 12 transformer layers, width 512 and eight heads (38M parameters). The later PyTorch BertConfig defaults to width 768 and 12 heads; these are different configurations.","status":"source_checked","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_bert.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Inputs","value":"Amino-acid sequence in the tokenizer expected by the selected implementation.","status":"source_checked","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_bert.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Outputs","value":"Residue/sequence representations and predictions from the chosen task head.","status":"source_checked","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_bert.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Parameters","value":"38M for the June 2019 preprint Transformer; do not apply that total to the later PyTorch defaults or every task head.","status":"source_checked","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_bert.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Known versions","value":"June 2019 TAPE preprint configurations and the later PyTorch package are distinct; the current README warns it is not an exact reproduction.","status":"source_checked","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_bert.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Training data","value":"The June 2019 TAPE preprint uses 31M Pfam domains, with held-out families and a separate random split; downstream task heads are fitted on their own task data.","status":"source_checked","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_bert.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Training cutoff","value":"The June 2019 TAPE preprint identifies the Pfam-domain corpus and split procedure but does not state one latest-sequence deposition date. Later package defaults are a different implementation.","status":"unreported","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_bert.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Context limits","value":"The inspected default BertConfig sets max_position_embeddings to 8,096; this is an implementation default, not evidence that a historical TAPE checkpoint was trained at that length.","status":"source_checked","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_bert.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Weights licence","value":"Separate checkpoint-distribution terms are not stated in the inspected release documentation and licence material. The source-code licence alone is not recorded as an explicit weight grant.","status":"unreported","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343","evidence-official-e847949b10d83e3b0efe"],"source_locator":"tape/models/modeling_bert.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3; LICENSE: licence text"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/songlab-cal/tape","status":"source_checked","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_bert.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Code licence","value":"BSD-3-Clause","status":"source_checked","source_ids":["evidence-official-e847949b10d83e3b0efe"],"source_locator":"LICENSE: licence text"}],"strengths":[{"text":"Provides a defined reference architecture that can be paired with the same downstream tasks as other protein representations.","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_bert.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"}],"limitations":[{"text":"The maintainers explicitly state that the current PyTorch repository is not an exact reproduction of the original TensorFlow paper code; training maintenance was discontinued. The default implementation reserves 8,096 position embeddings; this is an implementation setting, not proof of training or evaluation at that length.","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_bert.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"}],"diagram":{"title":"TAPE Transformer workflow","steps":["Protein sequence","Residue and position embeddings","Transformer encoder","Task head"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_bert.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},"coverage":"limited","gaps":["Training cutoff: The June 2019 TAPE preprint identifies the Pfam-domain corpus and split procedure but does not state one latest-sequence deposition date. Later package defaults are a different implementation.","Weights licence: Separate checkpoint-distribution terms are not stated in the inspected release documentation and licence material. The source-code licence alone is not recorded as an explicit weight grant."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}},"description":"Method identity as printed in the historical TAPE leaderboard.","facets":{"areas":["protein-function"]},"id":"discovery-model-tape-transformer","kind":"model","links":[],"name":"TAPE Transformer","source_ids":["src-discovery-songlab-cal-tape"],"status":"discovered"}
{"attributes":{"entity_level":"method","reported_name":"Unirep","version":null,"historical_missing_metadata":{"checkpoint":"unreported","version":"unreported"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"The TAPE UniRep baseline uses a multiplicative recurrent network to represent protein sequences for downstream tasks.","summary_source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"summary_source_locator":"tape/models/modeling_unirep.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3","sections":[{"title":"How it works","body":"The TAPE UniRep baseline uses a multiplicative recurrent network to represent protein sequences for downstream tasks. Multiplicative LSTM protein encoder; the babbler-1900 configuration has recurrent hidden size 1,900. The documented inputs are amino-acid sequence in the tokenizer expected by the selected implementation. The output consists of residue/sequence representations and predictions from the chosen task head.","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_unirep.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"title":"Versions and reproducibility","body":"June 2019 TAPE preprint configurations and the later PyTorch package are distinct; the current README warns it is not an exact reproduction. The TAPE preprint uses sequence-length-dependent batching; the exact task/checkpoint determines padding, truncation and resource constraints.","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_unirep.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"}],"facts":[{"label":"Model type","value":"Multiplicative-LSTM protein encoder","status":"source_checked","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_unirep.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Architecture","value":"Multiplicative LSTM protein encoder; the babbler-1900 configuration has recurrent hidden size 1,900.","status":"source_checked","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_unirep.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Inputs","value":"Amino-acid sequence in the tokenizer expected by the selected implementation.","status":"source_checked","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_unirep.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Outputs","value":"Residue/sequence representations and predictions from the chosen task head.","status":"source_checked","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_unirep.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Parameters","value":"UniRep uses a different vocabulary from the other TAPE models; use its matching tokenizer.","status":"source_checked","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_unirep.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Known versions","value":"June 2019 TAPE preprint configurations and the later PyTorch package are distinct; the current README warns it is not an exact reproduction.","status":"source_checked","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_unirep.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Training data","value":"The June 2019 TAPE preprint uses 31M Pfam domains, with held-out families and a separate random split; downstream task heads are fitted on their own task data.","status":"source_checked","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_unirep.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Training cutoff","value":"The June 2019 TAPE preprint identifies the Pfam-domain corpus and split procedure but does not state one latest-sequence deposition date. Later package defaults are a different implementation.","status":"unreported","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_unirep.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Context limits","value":"The TAPE preprint uses sequence-length-dependent batching; the exact task/checkpoint determines padding, truncation and resource constraints.","status":"source_checked","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_unirep.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Weights licence","value":"Separate checkpoint-distribution terms are not stated in the inspected release documentation and licence material. The source-code licence alone is not recorded as an explicit weight grant.","status":"unreported","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343","evidence-official-e847949b10d83e3b0efe"],"source_locator":"tape/models/modeling_unirep.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3; LICENSE: licence text"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/songlab-cal/tape","status":"source_checked","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_unirep.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},{"label":"Code licence","value":"BSD-3-Clause","status":"source_checked","source_ids":["evidence-official-e847949b10d83e3b0efe"],"source_locator":"LICENSE: licence text"}],"strengths":[{"text":"Provides a defined reference architecture that can be paired with the same downstream tasks as other protein representations.","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_unirep.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"}],"limitations":[{"text":"The maintainers explicitly state that the current PyTorch repository is not an exact reproduction of the original TensorFlow paper code; training maintenance was discontinued. UniRep uses a different vocabulary from the other TAPE models; use its matching tokenizer.","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_unirep.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"}],"diagram":{"title":"TAPE Unirep workflow","steps":["Protein sequence","UniRep-specific tokens","Multiplicative LSTM","Representation or task head"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-028db5eeb224def23456","evidence-official-1aae5efd1b27ae9c761a","evidence-official-3cec54ed6ded9cb0a092","evidence-official-044ad0df4e4fe48a2125","evidence-official-12d43bf3076a93fc4ee7","evidence-official-c303652f3d79132a00c2","evidence-official-362b9cef071111dfe343"],"source_locator":"tape/models/modeling_unirep.py: configuration and model classes; README.md: opening maintenance warning, Examples and Data; June 2019 TAPE preprint Sections 4.1,5 and AppendixA.3"},"coverage":"limited","gaps":["Training cutoff: The June 2019 TAPE preprint identifies the Pfam-domain corpus and split procedure but does not state one latest-sequence deposition date. Later package defaults are a different implementation.","Weights licence: Separate checkpoint-distribution terms are not stated in the inspected release documentation and licence material. The source-code licence alone is not recorded as an explicit weight grant."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}},"description":"Method identity as printed in the historical TAPE leaderboard.","facets":{"areas":["protein-function"]},"id":"discovery-model-tape-unirep","kind":"model","links":[],"name":"TAPE Unirep","source_ids":["src-discovery-songlab-cal-tape"],"status":"discovered"}
{"attributes":{"access":"official_source_linked","benchmark_applicability":"candidate; not evidence of a reported evaluation","candidate_benchmark_ids":["discovery-benchmark-beacon"],"entity_level":"method","reported_name":"ViennaRNA RNAfold","version":null,"historical_missing_metadata":{"checkpoint":"unextracted","code_licence":"unextracted","parameters":"unextracted","training_cutoff":"unextracted","training_data":"unextracted","version":"unextracted","weights_licence":"unextracted"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"RNAfold predicts RNA secondary structure and thermodynamic ensemble quantities within the ViennaRNA package.","summary_source_ids":["evidence-official-1b2aa7207e3a05f297ab"],"summary_source_locator":"README.md: package capabilities, executable programs, Energy Parameters and License","sections":[{"title":"How it works","body":"RNAfold predicts RNA secondary structure and thermodynamic ensemble quantities within the ViennaRNA package. Energy-based RNA secondary-structure calculation; minimum-free-energy and partition-function modes are distinct outputs. The documented inputs are RNA nucleotide sequence with selected thermodynamic settings and supported constraints. The output consists of predicted secondary structure and, in ensemble mode, partition-function-derived probabilities.","source_ids":["evidence-official-1b2aa7207e3a05f297ab"],"source_locator":"README.md: package capabilities, executable programs, Energy Parameters and License"},{"title":"Versions and reproducibility","body":"RNAfold within ViennaRNA; software and energy-parameter set must be pinned. Implementation and memory dependent; the package README gives an upper representational length with an explicit large-memory caveat.","source_ids":["evidence-official-1b2aa7207e3a05f297ab"],"source_locator":"README.md: package capabilities, executable programs, Energy Parameters and License"}],"facts":[{"label":"Model type","value":"Thermodynamic RNA secondary-structure procedure","status":"source_checked","source_ids":["evidence-official-1b2aa7207e3a05f297ab"],"source_locator":"README.md: package capabilities, executable programs, Energy Parameters and License"},{"label":"Architecture","value":"Energy-based RNA secondary-structure calculation; minimum-free-energy and partition-function modes are distinct outputs.","status":"source_checked","source_ids":["evidence-official-1b2aa7207e3a05f297ab"],"source_locator":"README.md: package capabilities, executable programs, Energy Parameters and License"},{"label":"Inputs","value":"RNA nucleotide sequence with selected thermodynamic settings and supported constraints.","status":"source_checked","source_ids":["evidence-official-1b2aa7207e3a05f297ab"],"source_locator":"README.md: package capabilities, executable programs, Energy Parameters and License"},{"label":"Outputs","value":"Predicted secondary structure and, in ensemble mode, partition-function-derived probabilities.","status":"source_checked","source_ids":["evidence-official-1b2aa7207e3a05f297ab"],"source_locator":"README.md: package capabilities, executable programs, Energy Parameters and License"},{"label":"Parameters","value":"Thermodynamic parameter set and algorithm settings; not neural model parameters.","status":"source_checked","source_ids":["evidence-official-1b2aa7207e3a05f297ab"],"source_locator":"README.md: package capabilities, executable programs, Energy Parameters and License"},{"label":"Known versions","value":"RNAfold within ViennaRNA; software and energy-parameter set must be pinned.","status":"source_checked","source_ids":["evidence-official-1b2aa7207e3a05f297ab"],"source_locator":"README.md: package capabilities, executable programs, Energy Parameters and License"},{"label":"Training data","value":"Energy parameters are empirical model inputs, not a pretrained neural checkpoint.","status":"source_checked","source_ids":["evidence-official-1b2aa7207e3a05f297ab"],"source_locator":"README.md: package capabilities, executable programs, Energy Parameters and License"},{"label":"Training cutoff","value":"Inapplicable to neural pretraining; reference-database and input-data dates must be recorded for each run.","status":"inapplicable","source_ids":["evidence-official-1b2aa7207e3a05f297ab"],"source_locator":"README.md: package capabilities, executable programs, Energy Parameters and License"},{"label":"Context limits","value":"Implementation and memory dependent; the package README gives an upper representational length with an explicit large-memory caveat.","status":"source_checked","source_ids":["evidence-official-1b2aa7207e3a05f297ab"],"source_locator":"README.md: package capabilities, executable programs, Energy Parameters and License"},{"label":"Weights licence","value":"Inapplicable to neural weights; preserve the thermodynamic parameter-file version and licence.","status":"inapplicable","source_ids":["evidence-official-1b2aa7207e3a05f297ab"],"source_locator":"README.md: package capabilities, executable programs, Energy Parameters and License"},{"label":"Access","value":"Official project documentation and implementation: https://github.com/ViennaRNA/ViennaRNA","status":"source_checked","source_ids":["evidence-official-1b2aa7207e3a05f297ab"],"source_locator":"README.md: package capabilities, executable programs, Energy Parameters and License"},{"label":"Code licence","value":"ViennaRNA custom permissive licence; inspect the pinned COPYING terms.","status":"source_checked","source_ids":["evidence-official-4269946a73db94a15289"],"source_locator":"COPYING: licence text"}],"strengths":[{"text":"Provides an interpretable thermodynamic reference for comparison with learned RNA structure methods.","source_ids":["evidence-official-1b2aa7207e3a05f297ab"],"source_locator":"README.md: package capabilities, executable programs, Energy Parameters and License"}],"limitations":[{"text":"The package includes multiple programs with different tasks. RNAfold secondary structure must not be presented as three-dimensional RNA folding. Sequence length is constrained by memory and algorithmic cost.","source_ids":["evidence-official-1b2aa7207e3a05f297ab"],"source_locator":"README.md: package capabilities, executable programs, Energy Parameters and License"}],"diagram":{"title":"ViennaRNA RNAfold workflow","steps":["RNA sequence and energy settings","Secondary-structure energy calculation","Minimum-energy or ensemble mode","Structure and probabilities"],"caption":"Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.","source_ids":["evidence-official-1b2aa7207e3a05f297ab"],"source_locator":"README.md: package capabilities, executable programs, Energy Parameters and License"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied."}}},"description":"Thermodynamic RNA secondary structure prediction","facets":{"areas":["rna"]},"id":"discovery-model-viennarna-rnafold","kind":"model","links":[{"relation":"applicable_to","target_id":"discovery-benchmark-beacon"}],"name":"ViennaRNA RNAfold","source_ids":["src-discovery-viennarna-viennarna"],"status":"discovered"}
{"attributes":{"metric":"Spearman's rho","metric_direction":"higher","missing_metadata":{"denominator":"unextracted","seeds":"unreported","uncertainty":"unreported"},"numeric_value":"0.33","printed_value":"0.33","review":{"method":"automated table parse plus separate source read and manually enumerated row/value cross-check","notes":"All six rows checked for this task. Source checked, not experiment reproduced. Historical README does not resolve exact checkpoint or data release.","reviewed_at":"2026-09-16T10:35:25.806835+00:00","reviewer":"Codex research agent; no human review claimed"},"source_locator":"README.md > Leaderboard > Fluorescence; row Bepler; column Spearman's rho","uncertainty":null,"unit":"correlation"},"description":"Reported score transcribed from the complete historical task table.","facets":{"areas":["protein-function"]},"id":"discovery-result-tape-fluorescence-bepler-spearman-rho","kind":"result","links":[{"relation":"evaluation","target_id":"discovery-evaluation-tape-fluorescence-bepler-leaderboard-evaluation"}],"name":"TAPE Fluorescence Bepler Spearman rho","source_ids":["src-discovery-songlab-cal-tape"],"status":"source_checked"}
{"attributes":{"metric":"Spearman's rho","metric_direction":"higher","missing_metadata":{"denominator":"unextracted","seeds":"unreported","uncertainty":"unreported"},"numeric_value":"0.67","printed_value":"0.67","review":{"method":"automated table parse plus separate source read and manually enumerated row/value cross-check","notes":"All six rows checked for this task. Source checked, not experiment reproduced. Historical README does not resolve exact checkpoint or data release.","reviewed_at":"2026-09-16T10:35:25.806835+00:00","reviewer":"Codex research agent; no human review claimed"},"source_locator":"README.md > Leaderboard > Fluorescence; row LSTM; column Spearman's rho","uncertainty":null,"unit":"correlation"},"description":"Reported score transcribed from the complete historical task table.","facets":{"areas":["protein-function"]},"id":"discovery-result-tape-fluorescence-lstm-spearman-rho","kind":"result","links":[{"relation":"evaluation","target_id":"discovery-evaluation-tape-fluorescence-lstm-leaderboard-evaluation"}],"name":"TAPE Fluorescence LSTM Spearman rho","source_ids":["src-discovery-songlab-cal-tape"],"status":"source_checked"}
{"attributes":{"metric":"Spearman's rho","metric_direction":"higher","missing_metadata":{"denominator":"unextracted","seeds":"unreported","uncertainty":"unreported"},"numeric_value":"0.14","printed_value":"0.14","review":{"method":"automated table parse plus separate source read and manually enumerated row/value cross-check","notes":"All six rows checked for this task. Source checked, not experiment reproduced. Historical README does not resolve exact checkpoint or data release.","reviewed_at":"2026-09-16T10:35:25.806835+00:00","reviewer":"Codex research agent; no human review claimed"},"source_locator":"README.md > Leaderboard > Fluorescence; row One Hot; column Spearman's rho","uncertainty":null,"unit":"correlation"},"description":"Reported score transcribed from the complete historical task table.","facets":{"areas":["protein-function"]},"id":"discovery-result-tape-fluorescence-one-hot-spearman-rho","kind":"result","links":[{"relation":"evaluation","target_id":"discovery-evaluation-tape-fluorescence-one-hot-leaderboard-evaluation"}],"name":"TAPE Fluorescence One Hot Spearman rho","source_ids":["src-discovery-songlab-cal-tape"],"status":"source_checked"}
{"attributes":{"metric":"Spearman's rho","metric_direction":"higher","missing_metadata":{"denominator":"unextracted","seeds":"unreported","uncertainty":"unreported"},"numeric_value":"0.21","printed_value":"0.21","review":{"method":"automated table parse plus separate source read and manually enumerated row/value cross-check","notes":"All six rows checked for this task. Source checked, not experiment reproduced. Historical README does not resolve exact checkpoint or data release.","reviewed_at":"2026-09-16T10:35:25.806835+00:00","reviewer":"Codex research agent; no human review claimed"},"source_locator":"README.md > Leaderboard > Fluorescence; row ResNet; column Spearman's rho","uncertainty":null,"unit":"correlation"},"description":"Reported score transcribed from the complete historical task table.","facets":{"areas":["protein-function"]},"id":"discovery-result-tape-fluorescence-resnet-spearman-rho","kind":"result","links":[{"relation":"evaluation","target_id":"discovery-evaluation-tape-fluorescence-resnet-leaderboard-evaluation"}],"name":"TAPE Fluorescence ResNet Spearman rho","source_ids":["src-discovery-songlab-cal-tape"],"status":"source_checked"}
{"attributes":{"metric":"Spearman's rho","metric_direction":"higher","missing_metadata":{"denominator":"unextracted","seeds":"unreported","uncertainty":"unreported"},"numeric_value":"0.68","printed_value":"0.68","review":{"method":"automated table parse plus separate source read and manually enumerated row/value cross-check","notes":"All six rows checked for this task. Source checked, not experiment reproduced. Historical README does not resolve exact checkpoint or data release.","reviewed_at":"2026-09-16T10:35:25.806835+00:00","reviewer":"Codex research agent; no human review claimed"},"source_locator":"README.md > Leaderboard > Fluorescence; row Transformer; column Spearman's rho","uncertainty":null,"unit":"correlation"},"description":"Reported score transcribed from the complete historical task table.","facets":{"areas":["protein-function"]},"id":"discovery-result-tape-fluorescence-transformer-spearman-rho","kind":"result","links":[{"relation":"evaluation","target_id":"discovery-evaluation-tape-fluorescence-transformer-leaderboard-evaluation"}],"name":"TAPE Fluorescence Transformer Spearman rho","source_ids":["src-discovery-songlab-cal-tape"],"status":"source_checked"}
{"attributes":{"metric":"Spearman's rho","metric_direction":"higher","missing_metadata":{"denominator":"unextracted","seeds":"unreported","uncertainty":"unreported"},"numeric_value":"0.67","printed_value":"0.67","review":{"method":"automated table parse plus separate source read and manually enumerated row/value cross-check","notes":"All six rows checked for this task. Source checked, not experiment reproduced. Historical README does not resolve exact checkpoint or data release.","reviewed_at":"2026-09-16T10:35:25.806835+00:00","reviewer":"Codex research agent; no human review claimed"},"source_locator":"README.md > Leaderboard > Fluorescence; row Unirep; column Spearman's rho","uncertainty":null,"unit":"correlation"},"description":"Reported score transcribed from the complete historical task table.","facets":{"areas":["protein-function"]},"id":"discovery-result-tape-fluorescence-unirep-spearman-rho","kind":"result","links":[{"relation":"evaluation","target_id":"discovery-evaluation-tape-fluorescence-unirep-leaderboard-evaluation"}],"name":"TAPE Fluorescence Unirep Spearman rho","source_ids":["src-discovery-songlab-cal-tape"],"status":"source_checked"}
{"attributes":{"metric":"Spearman's rho","metric_direction":"higher","missing_metadata":{"denominator":"unextracted","seeds":"unreported","uncertainty":"unreported"},"numeric_value":"0.64","printed_value":"0.64","review":{"method":"automated table parse plus separate source read and manually enumerated row/value cross-check","notes":"All six rows checked for this task. Source checked, not experiment reproduced. Historical README does not resolve exact checkpoint or data release.","reviewed_at":"2026-09-16T10:35:25.806835+00:00","reviewer":"Codex research agent; no human review claimed"},"source_locator":"README.md > Leaderboard > Stability; row Bepler; column Spearman's rho","uncertainty":null,"unit":"correlation"},"description":"Reported score transcribed from the complete historical task table.","facets":{"areas":["protein-function"]},"id":"discovery-result-tape-stability-bepler-spearman-rho","kind":"result","links":[{"relation":"evaluation","target_id":"discovery-evaluation-tape-stability-bepler-leaderboard-evaluation"}],"name":"TAPE Stability Bepler Spearman rho","source_ids":["src-discovery-songlab-cal-tape"],"status":"source_checked"}
{"attributes":{"metric":"Spearman's rho","metric_direction":"higher","missing_metadata":{"denominator":"unextracted","seeds":"unreported","uncertainty":"unreported"},"numeric_value":"0.69","printed_value":"0.69","review":{"method":"automated table parse plus separate source read and manually enumerated row/value cross-check","notes":"All six rows checked for this task. Source checked, not experiment reproduced. Historical README does not resolve exact checkpoint or data release.","reviewed_at":"2026-09-16T10:35:25.806835+00:00","reviewer":"Codex research agent; no human review claimed"},"source_locator":"README.md > Leaderboard > Stability; row LSTM; column Spearman's rho","uncertainty":null,"unit":"correlation"},"description":"Reported score transcribed from the complete historical task table.","facets":{"areas":["protein-function"]},"id":"discovery-result-tape-stability-lstm-spearman-rho","kind":"result","links":[{"relation":"evaluation","target_id":"discovery-evaluation-tape-stability-lstm-leaderboard-evaluation"}],"name":"TAPE Stability LSTM Spearman rho","source_ids":["src-discovery-songlab-cal-tape"],"status":"source_checked"}
{"attributes":{"metric":"Spearman's rho","metric_direction":"higher","missing_metadata":{"denominator":"unextracted","seeds":"unreported","uncertainty":"unreported"},"numeric_value":"0.19","printed_value":"0.19","review":{"method":"automated table parse plus separate source read and manually enumerated row/value cross-check","notes":"All six rows checked for this task. Source checked, not experiment reproduced. Historical README does not resolve exact checkpoint or data release.","reviewed_at":"2026-09-16T10:35:25.806835+00:00","reviewer":"Codex research agent; no human review claimed"},"source_locator":"README.md > Leaderboard > Stability; row One Hot; column Spearman's rho","uncertainty":null,"unit":"correlation"},"description":"Reported score transcribed from the complete historical task table.","facets":{"areas":["protein-function"]},"id":"discovery-result-tape-stability-one-hot-spearman-rho","kind":"result","links":[{"relation":"evaluation","target_id":"discovery-evaluation-tape-stability-one-hot-leaderboard-evaluation"}],"name":"TAPE Stability One Hot Spearman rho","source_ids":["src-discovery-songlab-cal-tape"],"status":"source_checked"}
{"attributes":{"metric":"Spearman's rho","metric_direction":"higher","missing_metadata":{"denominator":"unextracted","seeds":"unreported","uncertainty":"unreported"},"numeric_value":"0.73","printed_value":"0.73","review":{"method":"automated table parse plus separate source read and manually enumerated row/value cross-check","notes":"All six rows checked for this task. Source checked, not experiment reproduced. Historical README does not resolve exact checkpoint or data release.","reviewed_at":"2026-09-16T10:35:25.806835+00:00","reviewer":"Codex research agent; no human review claimed"},"source_locator":"README.md > Leaderboard > Stability; row ResNet; column Spearman's rho","uncertainty":null,"unit":"correlation"},"description":"Reported score transcribed from the complete historical task table.","facets":{"areas":["protein-function"]},"id":"discovery-result-tape-stability-resnet-spearman-rho","kind":"result","links":[{"relation":"evaluation","target_id":"discovery-evaluation-tape-stability-resnet-leaderboard-evaluation"}],"name":"TAPE Stability ResNet Spearman rho","source_ids":["src-discovery-songlab-cal-tape"],"status":"source_checked"}
{"attributes":{"metric":"Spearman's rho","metric_direction":"higher","missing_metadata":{"denominator":"unextracted","seeds":"unreported","uncertainty":"unreported"},"numeric_value":"0.73","printed_value":"0.73","review":{"method":"automated table parse plus separate source read and manually enumerated row/value cross-check","notes":"All six rows checked for this task. Source checked, not experiment reproduced. Historical README does not resolve exact checkpoint or data release.","reviewed_at":"2026-09-16T10:35:25.806835+00:00","reviewer":"Codex research agent; no human review claimed"},"source_locator":"README.md > Leaderboard > Stability; row Transformer; column Spearman's rho","uncertainty":null,"unit":"correlation"},"description":"Reported score transcribed from the complete historical task table.","facets":{"areas":["protein-function"]},"id":"discovery-result-tape-stability-transformer-spearman-rho","kind":"result","links":[{"relation":"evaluation","target_id":"discovery-evaluation-tape-stability-transformer-leaderboard-evaluation"}],"name":"TAPE Stability Transformer Spearman rho","source_ids":["src-discovery-songlab-cal-tape"],"status":"source_checked"}
{"attributes":{"metric":"Spearman's rho","metric_direction":"higher","missing_metadata":{"denominator":"unextracted","seeds":"unreported","uncertainty":"unreported"},"numeric_value":"0.73","printed_value":"0.73","review":{"method":"automated table parse plus separate source read and manually enumerated row/value cross-check","notes":"All six rows checked for this task. Source checked, not experiment reproduced. Historical README does not resolve exact checkpoint or data release.","reviewed_at":"2026-09-16T10:35:25.806835+00:00","reviewer":"Codex research agent; no human review claimed"},"source_locator":"README.md > Leaderboard > Stability; row Unirep; column Spearman's rho","uncertainty":null,"unit":"correlation"},"description":"Reported score transcribed from the complete historical task table.","facets":{"areas":["protein-function"]},"id":"discovery-result-tape-stability-unirep-spearman-rho","kind":"result","links":[{"relation":"evaluation","target_id":"discovery-evaluation-tape-stability-unirep-leaderboard-evaluation"}],"name":"TAPE Stability Unirep Spearman rho","source_ids":["src-discovery-songlab-cal-tape"],"status":"source_checked"}
{"id":"dna-foundation-models-2025","kind":"source","name":"Benchmarking DNA foundation models for genomic and genetic tasks","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["dna-genomes"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12663285/","version":"PMC12663285.1","retrieved_at":"2026-09-15T23:25:00Z","doi":"10.1038/s41467-025-65823-8","publication_status":"peer_reviewed","year":2025,"artifact_sha256":"5d8ca9bcf88cc1b38ad667906a2e4699b1aefa6d31c6f49259784930353f3202","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12663285/fullTextXML","artifact_retrieved_at":"2026-09-16T10:33:35.327Z","legacy_paper":{"id":"dna-foundation-models-2025","title":"Benchmarking DNA foundation models for genomic and genetic tasks","year":2025,"publication_status":"peer_reviewed","version":"PMC12663285.1","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12663285/","primary_domain":"dna-genomes","retrieved_utc":"2026-09-15T23:25:00Z","notes":"Primary full text via Europe PMC XML; venue: Nature Communications; PMC ID: PMC12663285.","doi":"10.1038/s41467-025-65823-8"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"dnabert2-enhancer-2025","kind":"source","name":"Utilizing a deep learning model based on BERT for identifying enhancers and their strength","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["dna-genomes"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11981215/","version":"journal full text in PMC","retrieved_at":"2026-09-15T23:33:26Z","doi":"10.1371/journal.pone.0320085","publication_status":"peer_reviewed","year":2025,"artifact_sha256":"d052b80efe7bfc1380994ad28503a5575f04ef940f74d5c9c137cb4ba6827863","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11981215/fullTextXML","artifact_retrieved_at":"2026-09-16T10:38:57.558204+00:00","legacy_paper":{"id":"dnabert2-enhancer-2025","title":"Utilizing a deep learning model based on BERT for identifying enhancers and their strength","year":2025,"publication_status":"peer_reviewed","version":"journal full text in PMC","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11981215/","primary_domain":"dna-genomes","retrieved_utc":"2026-09-15T23:33:26Z","doi":"10.1371/journal.pone.0320085","notes":"Numeric result checked against Table 4 in primary full-text XML; journal/source: PLOS One."},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"dnalongbench-2025","kind":"source","name":"DNALongBench: A Benchmark Suite for Long-Range DNA Prediction Tasks","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["dna-genomes"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11741265/","version":"PMC11741265.1","retrieved_at":"2026-09-15T23:25:00Z","doi":"10.1101/2025.01.06.631595","publication_status":"preprint","year":2025,"artifact_sha256":"fa440a17cecf16a5d872d50a30910f7591b5f6f78e10a944c6bda5ea8d7e32dd","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11741265/fullTextXML","artifact_retrieved_at":"2026-09-16T10:41:16.492545+00:00","legacy_paper":{"id":"dnalongbench-2025","title":"DNALongBench: A Benchmark Suite for Long-Range DNA Prediction Tasks","year":2025,"publication_status":"preprint","version":"PMC11741265.1","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11741265/","primary_domain":"dna-genomes","retrieved_utc":"2026-09-15T23:25:00Z","notes":"Primary full text via Europe PMC XML; venue: bioRxiv; PMC ID: PMC11741265.","doi":"10.1101/2025.01.06.631595"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"eden-genomic-classification-2026","kind":"source","name":"EDEN: multiscale expected density of nucleotide encoding for enhanced DNA sequence classification with hybrid deep learning","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["dna-genomes"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12879454/","version":"journal full text in PMC","retrieved_at":"2026-09-15T23:33:26Z","doi":"10.1186/s12859-026-06367-6","publication_status":"peer_reviewed","year":2026,"artifact_sha256":"38a6e26b3caffe8e021a2b0b672218e783aca9ee42046765e323946813015e65","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12879454/fullTextXML","artifact_retrieved_at":"2026-09-16T10:33:37.531Z","legacy_paper":{"id":"eden-genomic-classification-2026","title":"EDEN: multiscale expected density of nucleotide encoding for enhanced DNA sequence classification with hybrid deep learning","year":2026,"publication_status":"peer_reviewed","version":"journal full text in PMC","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12879454/","primary_domain":"dna-genomes","retrieved_utc":"2026-09-15T23:33:26Z","doi":"10.1186/s12859-026-06367-6","notes":"DNABERT-2 comparator 70.52 is printed in Table 5. The article does not clearly document whether this comparator was independently rerun or consolidated from prior GUE results, so evaluation origin is conservatively marked paper_compilation."},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"enbed-2024","kind":"source","name":"Understanding the natural language of DNA using encoder–decoder foundation models with byte-level precision","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["dna-genomes"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11341122/","version":"version of record","retrieved_at":"2026-09-15T23:25:00Z","doi":"10.1093/bioadv/vbae117","publication_status":"peer_reviewed","year":2024,"artifact_sha256":"e95d4be70d32e61af5a92eda8ea66f25a2cc629e3f83e7b5241b13cde8bdb83b","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11341122/fullTextXML","artifact_retrieved_at":"2026-09-16T10:33:35.378Z","legacy_paper":{"id":"enbed-2024","title":"Understanding the natural language of DNA using encoder–decoder foundation models with byte-level precision","year":2024,"publication_status":"peer_reviewed","version":"version of record","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11341122/","primary_domain":"dna-genomes","retrieved_utc":"2026-09-15T23:25:00Z","notes":"Primary full text via Europe PMC XML; venue: Bioinformatics Advances; PMC ID: PMC11341122.","doi":"10.1093/bioadv/vbae117"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"enhancer-position-encoding-2024","kind":"source","name":"A deep learning model for DNA enhancer prediction based on nucleotide position aware feature encoding","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["dna-genomes"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11167433/","version":"version of record","retrieved_at":"2026-09-15T23:37:05Z","doi":"10.1016/j.isci.2024.110030","publication_status":"peer_reviewed","year":2024,"artifact_sha256":"0183b6a111b1b02344cad35a571a1fd2c56257e406c5be1df69f7902c5d06749","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11167433/fullTextXML","artifact_retrieved_at":"2026-09-16T10:41:06Z","legacy_paper":{"id":"enhancer-position-encoding-2024","title":"A deep learning model for DNA enhancer prediction based on nucleotide position aware feature encoding","year":2024,"publication_status":"peer_reviewed","version":"version of record","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11167433/","primary_domain":"dna-genomes","retrieved_utc":"2026-09-15T23:37:05Z","notes":"Primary full text verified via Europe PMC fullTextXML; venue: iScience; PMC ID: PMC11167433. Task-specific CNN baseline, included as a DNA benchmark protocol reference.","doi":"10.1016/j.isci.2024.110030"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"ensemble-idp-docking-2025","kind":"source","name":"Ensemble docking for intrinsically disordered proteins","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["molecular-interactions"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11785235/","version":"preprint archived 2025-01-26","retrieved_at":"2026-09-15T23:29:32Z","doi":"10.1101/2025.01.23.634614","publication_status":"preprint","year":2025,"artifact_sha256":"d02d91cdde41cb76ec5c86b532dffc564879c69e764a8c6b7752460fbbfd24b7","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11785235/fullTextXML","artifact_retrieved_at":"2026-09-16T10:33:56.275Z","legacy_paper":{"id":"ensemble-idp-docking-2025","title":"Ensemble docking for intrinsically disordered proteins","year":2025,"publication_status":"preprint","version":"preprint archived 2025-01-26","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11785235/","primary_domain":"molecular-interactions","retrieved_utc":"2026-09-15T23:29:32Z","notes":"Primary full text verified using Europe PMC XML; venue: bioRxiv; PMC ID: PMC11785235.","doi":"10.1101/2025.01.23.634614"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"ernie-rna-2025","kind":"source","name":"ERNIE-RNA: an RNA language model with structure-enhanced representations","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["rna-transcriptomes"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12627772/","version":"journal full text in PMC","retrieved_at":"2026-09-15T23:33:26Z","doi":"10.1038/s41467-025-64972-0","publication_status":"peer_reviewed","year":2025,"artifact_sha256":"0bd1d4b3cbf5d59d452cec4864614947861efcee050ba07e7de395cd90630047","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12627772/fullTextXML","artifact_retrieved_at":"2026-09-16T10:38:57.558206+00:00","legacy_paper":{"id":"ernie-rna-2025","title":"ERNIE-RNA: an RNA language model with structure-enhanced representations","year":2025,"publication_status":"peer_reviewed","version":"journal full text in PMC","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12627772/","primary_domain":"rna-transcriptomes","retrieved_utc":"2026-09-15T23:33:26Z","doi":"10.1038/s41467-025-64972-0","notes":"Numeric result checked against Table 2 in primary full-text XML; journal/source: Nature Communications."},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"esm2-amp-2025","kind":"source","name":"ESM2_AMP: an interpretable framework for protein–protein interactions prediction and biological mechanism discovery","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["molecular-interactions"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12392411/","version":"version of record","retrieved_at":"2026-09-15T23:37:05Z","doi":"10.1093/bib/bbaf434","publication_status":"peer_reviewed","year":2025,"artifact_sha256":"8e7ad6efb72ca28d73037cdf465b0e62f99cd6d0ee4ca9eaf96a4c48da22fd6c","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12392411/fullTextXML","artifact_retrieved_at":"2026-09-16T10:33:58.625Z","legacy_paper":{"id":"esm2-amp-2025","title":"ESM2_AMP: an interpretable framework for protein–protein interactions prediction and biological mechanism discovery","year":2025,"publication_status":"peer_reviewed","version":"version of record","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12392411/","primary_domain":"molecular-interactions","retrieved_utc":"2026-09-15T23:37:05Z","notes":"Primary full text verified via Europe PMC fullTextXML; venue: Briefings in Bioinformatics; PMC ID: PMC12392411. Paper has multiple model variants; selected named ESM2_AMPS variant only.","doi":"10.1093/bib/bbaf434"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"esm2-ofs-fitness-2025","kind":"source","name":"Pseudo-perplexity in One Fell Swoop for Protein Fitness Estimation","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["proteins-complexes"]},"source_ids":[],"links":[],"attributes":{"url":"https://journals.aps.org/prxlife/pdf/10.1103/zhx7-hcmm","version":"PRX Life 2025 journal article","retrieved_at":"2026-09-15T23:33:26Z","doi":"10.1103/zhx7-hcmm","publication_status":"peer_reviewed","year":2025,"artifact_sha256":"085ef646f11b8e5335c4b3d86b15fb6c7bf5edf4a80a8b622753ac69d9991a67","artifact_url":"https://harvest.aps.org/v2/journals/articles/10.1103/zhx7-hcmm/fulltext","artifact_retrieved_at":"2026-09-16T10:45:41.099916+00:00","legacy_paper":{"id":"esm2-ofs-fitness-2025","title":"Pseudo-perplexity in One Fell Swoop for Protein Fitness Estimation","year":2025,"publication_status":"peer_reviewed","version":"PRX Life 2025 journal article","source_url":"https://journals.aps.org/prxlife/pdf/10.1103/zhx7-hcmm","primary_domain":"proteins-complexes","retrieved_utc":"2026-09-15T23:33:26Z","doi":"10.1103/zhx7-hcmm","notes":"Final journal Table I, ESM2: OFS PP Aggregate Mean 0.403 checked directly; manuscript PMC11257618 printed the same value. Other models in the table are imported ProteinGym baselines; this row is the authors’ own evaluation."},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-b2-2ome-lm-2025","kind":"evaluation","name":"2OMe-LM: human RNA 2-prime-O-methylation site prediction","description":"","status":"needs_review","facets":{"areas":["rna-transcriptomes"],"tasks":["human RNA 2-prime-O-methylation site prediction"]},"source_ids":["2ome-lm-2025"],"links":[{"relation":"model","target_id":"reported-model-7f5b8234967c54"},{"relation":"benchmark","target_id":"reported-task-82fc7843f07324"},{"relation":"dataset","target_id":"reported-dataset-bd3d8e7d6cd196"}],"attributes":{"origin":"author_reported","protocol":"pretrained RNA language model predictor","version":"not stated in table","comparison":{"protocol_id":null,"dataset_version":null,"split":"5-fold cross-validation","population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-b2-antibody-deamidation-plm-2024","kind":"evaluation","name":"ESM-2 650M embeddings + classifier: antibody deamidation-site prediction","description":"","status":"needs_review","facets":{"areas":["proteins-complexes"],"tasks":["antibody deamidation-site prediction"]},"source_ids":["antibody-deamidation-plm-2024"],"links":[{"relation":"model","target_id":"reported-model-4921459942b45f"},{"relation":"benchmark","target_id":"reported-task-0647b0364def8f"},{"relation":"dataset","target_id":"reported-dataset-0edd8f724db696"}],"attributes":{"origin":"author_reported","protocol":"global contextual embeddings only","version":"esm2_t33_650m_UR50D","comparison":{"protocol_id":null,"dataset_version":null,"split":"fivefold stratified CV","population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-b2-barcodebert-2026","kind":"evaluation","name":"BarcodeBERT (4–4-4): unseen-species genus classification","description":"","status":"needs_review","facets":{"areas":["dna-genomes"],"tasks":["unseen-species genus classification"]},"source_ids":["barcodebert-2026"],"links":[{"relation":"model","target_id":"reported-model-05103f72325fe5"},{"relation":"benchmark","target_id":"reported-task-4a54ce01b5a855"},{"relation":"dataset","target_id":"reported-dataset-bc127dc9c441fe"}],"attributes":{"origin":"author_reported","protocol":"genus-level nearest-neighbor probe on species unseen in training","version":"4–4–4","comparison":{"protocol_id":null,"dataset_version":null,"split":"1-NN probe","population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-b2-birna-bert-2025","kind":"evaluation","name":"BiRNA-BERT: extremely long RNA species classification","description":"","status":"needs_review","facets":{"areas":["rna-transcriptomes"],"tasks":["extremely long RNA species classification"]},"source_ids":["birna-bert-2025"],"links":[{"relation":"model","target_id":"reported-model-d3fd83835a2d44"},{"relation":"benchmark","target_id":"reported-task-c40dac20d9af66"},{"relation":"dataset","target_id":"reported-dataset-ebc3f5fda43972"}],"attributes":{"origin":"author_reported","protocol":"adaptive tokenization on full-length long RNA sequences","version":"not stated in table","comparison":{"protocol_id":null,"dataset_version":null,"split":"paper evaluation","population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-b2-cathe2-2025","kind":"evaluation","name":"CATHe2 + ProstT5: CATH superfamily annotation","description":"","status":"needs_review","facets":{"areas":["proteins-complexes"],"tasks":["CATH superfamily annotation"]},"source_ids":["cathe2-2025"],"links":[{"relation":"model","target_id":"reported-model-49bc768f46b366"},{"relation":"benchmark","target_id":"reported-task-c98e91ffc7247d"},{"relation":"dataset","target_id":"reported-dataset-6e0c28dfde7337"}],"attributes":{"origin":"author_reported","protocol":"amino-acid and structural alphabet embedding classifier","version":"full ProstT5","comparison":{"protocol_id":null,"dataset_version":null,"split":"paper evaluation","population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-b2-clathrin-plm-2025","kind":"evaluation","name":"ESM-2 embedding + paper classifier: clathrin protein classification","description":"","status":"needs_review","facets":{"areas":["proteins-complexes"],"tasks":["clathrin protein classification"]},"source_ids":["clathrin-plm-2025"],"links":[{"relation":"model","target_id":"reported-model-e4710b1c3facf2"},{"relation":"benchmark","target_id":"reported-task-786c09824e9bf5"},{"relation":"dataset","target_id":"reported-dataset-0aab382ca2c063"}],"attributes":{"origin":"independent_paper","protocol":"single-feature ESM-2 embedding comparison","version":"not stated in table","comparison":{"protocol_id":null,"dataset_version":null,"split":"independent test","population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-b2-cobra-rna-binding-2026","kind":"evaluation","name":"ERNIE-RNA + CoBRA: RNA compound-binding site prediction","description":"","status":"needs_review","facets":{"areas":["rna-transcriptomes"],"tasks":["RNA compound-binding site prediction"]},"source_ids":["cobra-rna-binding-2026"],"links":[{"relation":"model","target_id":"reported-model-7ad28cd57f5f5b"},{"relation":"benchmark","target_id":"reported-task-3a3bff34cce634"},{"relation":"dataset","target_id":"reported-dataset-b1af840b76b351"}],"attributes":{"origin":"author_reported","protocol":"ERNIE-RNA embedding with TCL focal loss","version":"not stated in table","comparison":{"protocol_id":null,"dataset_version":null,"split":"test set","population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-b2-codonbert-vaccines-2024","kind":"evaluation","name":"CodonBERT: flu-vaccine mRNA property prediction","description":"","status":"needs_review","facets":{"areas":["rna-transcriptomes"],"tasks":["flu-vaccine mRNA property prediction"]},"source_ids":["codonbert-vaccines-2024"],"links":[{"relation":"model","target_id":"reported-model-cd246741c378db"},{"relation":"benchmark","target_id":"reported-task-1c74661df2c401"},{"relation":"dataset","target_id":"reported-dataset-54b9bc432928d6"}],"attributes":{"origin":"author_reported","protocol":"codon-based model fine-tuned for downstream regression","version":"not stated in table","comparison":{"protocol_id":null,"dataset_version":null,"split":"paper evaluation","population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-b2-dart-eval-regulatory-2024","kind":"evaluation","name":"DNABERT-2: regulatory element identification","description":"","status":"needs_review","facets":{"areas":["dna-genomes"],"tasks":["regulatory element identification"]},"source_ids":["dart-eval-regulatory-2024"],"links":[{"relation":"model","target_id":"reported-model-28413ae1766316"},{"relation":"benchmark","target_id":"reported-task-cdbee1c9285568"},{"relation":"dataset","target_id":"reported-dataset-b6ce37ba678d39"}],"attributes":{"origin":"independent_paper","protocol":"zero-shot likelihood ranking: higher likelihood for cCRE than matched control","version":"not stated in table","comparison":{"protocol_id":null,"dataset_version":null,"split":"paper evaluation","population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-b2-dnabert2-enhancer-2025","kind":"evaluation","name":"DNABERT2-Enhancer: enhancer recognition","description":"","status":"needs_review","facets":{"areas":["dna-genomes"],"tasks":["enhancer recognition"]},"source_ids":["dnabert2-enhancer-2025"],"links":[{"relation":"model","target_id":"reported-model-d0e594ec3c0430"},{"relation":"benchmark","target_id":"reported-task-86a628af87ff8f"},{"relation":"dataset","target_id":"reported-dataset-6212e779949708"}],"attributes":{"origin":"author_reported","protocol":"first-layer enhancer versus non-enhancer classifier","version":"not stated in table","comparison":{"protocol_id":null,"dataset_version":null,"split":"5-fold cross-validation","population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-b2-eden-genomic-classification-2026","kind":"evaluation","name":"DNABERT-2: human core-promoter classification","description":"","status":"needs_review","facets":{"areas":["dna-genomes"],"tasks":["human core-promoter classification"]},"source_ids":["eden-genomic-classification-2026"],"links":[{"relation":"model","target_id":"reported-model-2cb8118b4c77c0"},{"relation":"benchmark","target_id":"reported-task-9f62e739c6371e"},{"relation":"dataset","target_id":"reported-dataset-8e9488896becd4"}],"attributes":{"origin":"paper_compilation","protocol":"DNABERT-2 comparator in consolidated H-CPD table; rerun provenance not explicit","version":"not stated in table","comparison":{"protocol_id":null,"dataset_version":null,"split":"paper evaluation","population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract","original_evaluation":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-b2-ernie-rna-2025","kind":"evaluation","name":"ERNIE-RNA: RNA secondary-structure prediction","description":"","status":"needs_review","facets":{"areas":["rna-transcriptomes"],"tasks":["RNA secondary-structure prediction"]},"source_ids":["ernie-rna-2025"],"links":[{"relation":"model","target_id":"reported-model-d023fbe78bc4df"},{"relation":"benchmark","target_id":"reported-task-a2bf7ddbc71d23"},{"relation":"dataset","target_id":"reported-dataset-abdfba8cce7486"}],"attributes":{"origin":"author_reported","protocol":"zero-shot attention-derived base-pair prediction","version":"86M","comparison":{"protocol_id":null,"dataset_version":null,"split":"cross-family test","population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-b2-esm2-ofs-fitness-2025","kind":"evaluation","name":"ESM2 OFS pseudo-perplexity: protein variant fitness prediction","description":"","status":"needs_review","facets":{"areas":["proteins-complexes"],"tasks":["protein variant fitness prediction"]},"source_ids":["esm2-ofs-fitness-2025"],"links":[{"relation":"model","target_id":"reported-model-40ce004270dee4"},{"relation":"benchmark","target_id":"reported-task-c7a8a372f77886"},{"relation":"dataset","target_id":"reported-dataset-9c186c8f4ed3f4"}],"attributes":{"origin":"author_reported","protocol":"authors’ zero-shot ESM2 OFS pseudo-perplexity evaluation; aggregate mean across ProteinGym substitution assays","version":"not stated in table","comparison":{"protocol_id":null,"dataset_version":null,"split":"aggregate across assays","population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-b2-fusion-breakpoint-foundation-models-2026","kind":"evaluation","name":"Nucleotide Transformer + NN (middle): gene fusion breakpoint classification","description":"","status":"needs_review","facets":{"areas":["dna-genomes"],"tasks":["gene fusion breakpoint classification"]},"source_ids":["fusion-breakpoint-foundation-models-2026"],"links":[{"relation":"model","target_id":"reported-model-1a67087ac262c5"},{"relation":"benchmark","target_id":"reported-task-ee34721cf55590"},{"relation":"dataset","target_id":"reported-dataset-f6922a9744ba27"}],"attributes":{"origin":"independent_paper","protocol":"middle embedding with neural-network classifier","version":"not stated in table","comparison":{"protocol_id":null,"dataset_version":null,"split":"full test set","population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-b2-genomic-tokenizer-selection-2025","kind":"evaluation","name":"Caduceus (character tokens): regulatory sequence classification","description":"","status":"needs_review","facets":{"areas":["dna-genomes"],"tasks":["regulatory sequence classification"]},"source_ids":["genomic-tokenizer-selection-2025"],"links":[{"relation":"model","target_id":"reported-model-d5bc536ca6f0d3"},{"relation":"benchmark","target_id":"reported-task-cd127e56fb1f04"},{"relation":"dataset","target_id":"reported-dataset-0bba1c9a7ae410"}],"attributes":{"origin":"independent_paper","protocol":"task-category MCC across benchmark datasets","version":"3.9M parameter variant","comparison":{"protocol_id":null,"dataset_version":null,"split":"paper benchmark summary","population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-b2-gsmformer-ppi-2026","kind":"evaluation","name":"GSMFormer-PPI + ProstT5: protein-protein interaction prediction","description":"","status":"needs_review","facets":{"areas":["proteins-complexes"],"tasks":["protein-protein interaction prediction"]},"source_ids":["gsmformer-ppi-2026"],"links":[{"relation":"model","target_id":"reported-model-73ae07fb5be204"},{"relation":"benchmark","target_id":"reported-task-dfa8f2285dbfa5"},{"relation":"dataset","target_id":"reported-dataset-07d355c146be1f"}],"attributes":{"origin":"author_reported","protocol":"ProstT5 embeddings as graph node features","version":"not stated in table","comparison":{"protocol_id":null,"dataset_version":null,"split":"test set","population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-b2-megsite-2025","kind":"evaluation","name":"MegSite + ESM3: DNA-binding residue prediction","description":"","status":"needs_review","facets":{"areas":["proteins-complexes"],"tasks":["DNA-binding residue prediction"]},"source_ids":["megsite-2025"],"links":[{"relation":"model","target_id":"reported-model-86393c76dd8fa9"},{"relation":"benchmark","target_id":"reported-task-9917a0e69f33e7"},{"relation":"dataset","target_id":"reported-dataset-739aee3cf8d6f1"}],"attributes":{"origin":"author_reported","protocol":"ESM3 multimodal embedding ablation in MegSite","version":"not stated in table","comparison":{"protocol_id":null,"dataset_version":null,"split":"independent test","population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-b2-mrna-lm-2025","kind":"evaluation","name":"mRNA-LM: mRNA half-life prediction","description":"","status":"needs_review","facets":{"areas":["rna-transcriptomes"],"tasks":["mRNA half-life prediction"]},"source_ids":["mrna-lm-2025"],"links":[{"relation":"model","target_id":"reported-model-54d974e8e08043"},{"relation":"benchmark","target_id":"reported-task-5693847493f19f"},{"relation":"dataset","target_id":"reported-dataset-52f00ccaabf0d9"}],"attributes":{"origin":"author_reported","protocol":"average test performance across cross-validation splits","version":"not stated in table","comparison":{"protocol_id":null,"dataset_version":null,"split":"test set across CV splits","population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-b2-mrnabert-2025","kind":"evaluation","name":"mRNABERT: translation-efficiency prediction","description":"","status":"needs_review","facets":{"areas":["rna-transcriptomes"],"tasks":["translation-efficiency prediction"]},"source_ids":["mrnabert-2025"],"links":[{"relation":"model","target_id":"reported-model-13bd2a6c2d8178"},{"relation":"benchmark","target_id":"reported-task-f7142c3b3e0f3c"},{"relation":"dataset","target_id":"reported-dataset-1744719eef145b"}],"attributes":{"origin":"author_reported","protocol":"human translation-efficiency regression at 3066-nt input","version":"3066-nt input","comparison":{"protocol_id":null,"dataset_version":null,"split":"paper evaluation","population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-b2-mulan-2025","kind":"evaluation","name":"MULAN-ESM2 S: human protein-protein interaction prediction","description":"","status":"needs_review","facets":{"areas":["proteins-complexes"],"tasks":["human protein-protein interaction prediction"]},"source_ids":["mulan-2025"],"links":[{"relation":"model","target_id":"reported-model-a29203c09857ef"},{"relation":"benchmark","target_id":"reported-task-6e54c7452b2b81"},{"relation":"dataset","target_id":"reported-dataset-38151fa548e291"}],"attributes":{"origin":"author_reported","protocol":"MULAN sequence-structure model based on ESM2 8M","version":"small ESM2 backbone","comparison":{"protocol_id":null,"dataset_version":null,"split":"paper evaluation","population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-b2-phylogpn-2025","kind":"evaluation","name":"PhyloGPN: ClinVar 3-prime UTR variant classification","description":"","status":"needs_review","facets":{"areas":["dna-genomes"],"tasks":["ClinVar 3-prime UTR variant classification"]},"source_ids":["phylogpn-2025"],"links":[{"relation":"model","target_id":"reported-model-cbbe04b826ceff"},{"relation":"benchmark","target_id":"reported-task-ed3dd3b83c4505"},{"relation":"dataset","target_id":"reported-dataset-a28180d33f7a23"}],"attributes":{"origin":"author_reported","protocol":"log-likelihood-ratio scoring","version":"not stated in table","comparison":{"protocol_id":null,"dataset_version":null,"split":"paper evaluation","population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-b2-polya-glm-2025","kind":"evaluation","name":"HyenaDNA: polyadenylation site detection","description":"","status":"needs_review","facets":{"areas":["dna-genomes"],"tasks":["polyadenylation site detection"]},"source_ids":["polya-glm-2025"],"links":[{"relation":"model","target_id":"reported-model-953007693fb72a"},{"relation":"benchmark","target_id":"reported-task-13dfe6b33e71ed"},{"relation":"dataset","target_id":"reported-dataset-55f200c9481409"}],"attributes":{"origin":"independent_paper","protocol":"few-shot Gene-Gene negative-set comparison","version":"not stated in table","comparison":{"protocol_id":null,"dataset_version":null,"split":"5-fold cross-validation","population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-b2-rlsite-rna-binding-2025","kind":"evaluation","name":"RLsite: RNA-small-molecule binding-site prediction","description":"","status":"needs_review","facets":{"areas":["rna-transcriptomes"],"tasks":["RNA-small-molecule binding-site prediction"]},"source_ids":["rlsite-rna-binding-2025"],"links":[{"relation":"model","target_id":"reported-model-52b9c685d99290"},{"relation":"benchmark","target_id":"reported-task-b00a636d1ed8d9"},{"relation":"dataset","target_id":"reported-dataset-1c7f8ebb1968d9"}],"attributes":{"origin":"author_reported","protocol":"RNA language-model plus graph-attention classifier","version":"not stated in table","comparison":{"protocol_id":null,"dataset_version":null,"split":"paper evaluation","population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-b2-rnaret-2026","kind":"evaluation","name":"RNAret: miRNA-mRNA interaction prediction","description":"","status":"needs_review","facets":{"areas":["rna-transcriptomes"],"tasks":["miRNA-mRNA interaction prediction"]},"source_ids":["rnaret-2026"],"links":[{"relation":"model","target_id":"reported-model-7234658bc9c828"},{"relation":"benchmark","target_id":"reported-task-46e927bea10702"},{"relation":"dataset","target_id":"reported-dataset-99afd0c86b2954"}],"attributes":{"origin":"author_reported","protocol":"5-mer RNAret classifier; 72/8/20 train/validation/test split","version":"5-mer","comparison":{"protocol_id":null,"dataset_version":null,"split":"held-out test","population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-b2-spin-protein-function-2026","kind":"evaluation","name":"SPIN + ESM2-35M: protein function annotation","description":"","status":"needs_review","facets":{"areas":["proteins-complexes"],"tasks":["protein function annotation"]},"source_ids":["spin-protein-function-2026"],"links":[{"relation":"model","target_id":"reported-model-f8f0257b98749a"},{"relation":"benchmark","target_id":"reported-task-c4a578065f44b2"},{"relation":"dataset","target_id":"reported-dataset-dba1707164d296"}],"attributes":{"origin":"author_reported","protocol":"frozen ESM2-35M backbone in SPIN","version":"ESM2-35M frozen","comparison":{"protocol_id":null,"dataset_version":null,"split":"test set","population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-b2-structure-informed-plm-2025","kind":"evaluation","name":"structure-informed pLM: protein variant-effect classification","description":"","status":"needs_review","facets":{"areas":["proteins-complexes"],"tasks":["protein variant-effect classification"]},"source_ids":["structure-informed-plm-2025"],"links":[{"relation":"model","target_id":"reported-model-035a3ab36a3a6a"},{"relation":"benchmark","target_id":"reported-task-83be0998084c91"},{"relation":"dataset","target_id":"reported-dataset-2eaa2a051d45ee"}],"attributes":{"origin":"author_reported","protocol":"combined amino-acid, secondary structure, solvent accessibility and contact-map scoring","version":"not stated in table","comparison":{"protocol_id":null,"dataset_version":null,"split":"paper evaluation","population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-001","kind":"evaluation","name":"Caduceus-Ph: Human 5mC detection","description":"","status":"needs_review","facets":{"areas":["dna-genomes"],"tasks":["Human 5mC detection"]},"source_ids":["dna-foundation-models-2025"],"links":[{"relation":"model","target_id":"reported-model-47521865af7b04"},{"relation":"benchmark","target_id":"reported-task-988ff78f86471e"},{"relation":"dataset","target_id":"reported-dataset-463197d6a98b99"}],"attributes":{"origin":"independent_paper","protocol":"Binary epigenetic-modification classification as reported in the paper.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-002","kind":"evaluation","name":"NT-v2: Human 5mC detection","description":"","status":"needs_review","facets":{"areas":["dna-genomes"],"tasks":["Human 5mC detection"]},"source_ids":["dna-foundation-models-2025"],"links":[{"relation":"model","target_id":"reported-model-3af86cb274f658"},{"relation":"benchmark","target_id":"reported-task-988ff78f86471e"},{"relation":"dataset","target_id":"reported-dataset-463197d6a98b99"}],"attributes":{"origin":"independent_paper","protocol":"Binary epigenetic-modification classification as reported in the paper.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-003","kind":"evaluation","name":"ENBED: Enhancer classification","description":"","status":"needs_review","facets":{"areas":["dna-genomes"],"tasks":["Enhancer classification"]},"source_ids":["enbed-2024"],"links":[{"relation":"model","target_id":"reported-model-8db190bee6aae5"},{"relation":"benchmark","target_id":"reported-task-132da895d4c381"},{"relation":"dataset","target_id":"reported-dataset-f0bf60a62ad7c3"}],"attributes":{"origin":"author_reported","protocol":"Reported Genomic Benchmarks classification accuracy.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-004","kind":"evaluation","name":"ENBED (GRCh38): Enhancer classification","description":"","status":"needs_review","facets":{"areas":["dna-genomes"],"tasks":["Enhancer classification"]},"source_ids":["enbed-2024"],"links":[{"relation":"model","target_id":"reported-model-bdb1db16d3389d"},{"relation":"benchmark","target_id":"reported-task-132da895d4c381"},{"relation":"dataset","target_id":"reported-dataset-f0bf60a62ad7c3"}],"attributes":{"origin":"author_reported","protocol":"ENBED trained on GRCh38; reported Genomic Benchmarks classification accuracy.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-005","kind":"evaluation","name":"DNABERT-2: G-quadruplex classification","description":"","status":"needs_review","facets":{"areas":["dna-genomes"],"tasks":["G-quadruplex classification"]},"source_ids":["quadruplex-llm-benchmark-2025"],"links":[{"relation":"model","target_id":"reported-model-ade36035f58f27"},{"relation":"benchmark","target_id":"reported-task-c9d2a6435979e9"},{"relation":"dataset","target_id":"reported-dataset-9e9d18bc5bfb8b"}],"attributes":{"origin":"independent_paper","protocol":"Pretrained model evaluated on KEx as reported in Table 5.","version":"117M","comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-006","kind":"evaluation","name":"Caduceus: G-quadruplex classification","description":"","status":"needs_review","facets":{"areas":["dna-genomes"],"tasks":["G-quadruplex classification"]},"source_ids":["quadruplex-llm-benchmark-2025"],"links":[{"relation":"model","target_id":"reported-model-abc19288fe9009"},{"relation":"benchmark","target_id":"reported-task-c9d2a6435979e9"},{"relation":"dataset","target_id":"reported-dataset-9e9d18bc5bfb8b"}],"attributes":{"origin":"independent_paper","protocol":"Pretrained model evaluated on KEx as reported in Table 5.","version":"8M","comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-007","kind":"evaluation","name":"HyenaDNA: Enhancer-target gene prediction","description":"","status":"needs_review","facets":{"areas":["dna-genomes"],"tasks":["Enhancer-target gene prediction"]},"source_ids":["dnalongbench-2025"],"links":[{"relation":"model","target_id":"reported-model-9b3bc255532dd3"},{"relation":"benchmark","target_id":"reported-task-2cbac97dd849f5"},{"relation":"dataset","target_id":"reported-dataset-fcb5752d916d5d"}],"attributes":{"origin":"independent_paper","protocol":"Long-range ETGP benchmark; source table reports AUROC.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-008","kind":"evaluation","name":"Caduceus-Ph: Enhancer-target gene prediction","description":"","status":"needs_review","facets":{"areas":["dna-genomes"],"tasks":["Enhancer-target gene prediction"]},"source_ids":["dnalongbench-2025"],"links":[{"relation":"model","target_id":"reported-model-a7cfacf25d97ad"},{"relation":"benchmark","target_id":"reported-task-2cbac97dd849f5"},{"relation":"dataset","target_id":"reported-dataset-fcb5752d916d5d"}],"attributes":{"origin":"independent_paper","protocol":"Long-range ETGP benchmark; source table reports AUROC.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-009","kind":"evaluation","name":"RiNALMo: Mean ribosome load from MPRA","description":"","status":"needs_review","facets":{"areas":["rna-transcriptomes"],"tasks":["Mean ribosome load from MPRA"]},"source_ids":["mrnabench-2025"],"links":[{"relation":"model","target_id":"reported-model-3e58d0faf88d2e"},{"relation":"benchmark","target_id":"reported-task-57dc3dcdb67a81"},{"relation":"dataset","target_id":"reported-dataset-3a3e3880a3fed0"}],"attributes":{"origin":"independent_paper","protocol":"Linear probe; mean across ten random seeds.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-010","kind":"evaluation","name":"RNA-FM: Mean ribosome load from MPRA","description":"","status":"needs_review","facets":{"areas":["rna-transcriptomes"],"tasks":["Mean ribosome load from MPRA"]},"source_ids":["mrnabench-2025"],"links":[{"relation":"model","target_id":"reported-model-43cf51abca83d1"},{"relation":"benchmark","target_id":"reported-task-57dc3dcdb67a81"},{"relation":"dataset","target_id":"reported-dataset-3a3e3880a3fed0"}],"attributes":{"origin":"independent_paper","protocol":"Linear probe; mean across ten random seeds.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-011","kind":"evaluation","name":"BPfold: RNA secondary structure","description":"","status":"needs_review","facets":{"areas":["rna-transcriptomes"],"tasks":["RNA secondary structure"]},"source_ids":["bpfold-2025"],"links":[{"relation":"model","target_id":"reported-model-c464bface507ee"},{"relation":"benchmark","target_id":"reported-task-dc82fcbfb44935"},{"relation":"dataset","target_id":"reported-dataset-8317793f18b026"}],"attributes":{"origin":"author_reported","protocol":"Family-wise evaluation of canonical base-pair predictions.","version":null,"comparison":{"protocol_id":null,"dataset_version":"116 RNAs","split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-012","kind":"evaluation","name":"RNAfold: RNA secondary structure","description":"","status":"needs_review","facets":{"areas":["rna-transcriptomes"],"tasks":["RNA secondary structure"]},"source_ids":["bpfold-2025"],"links":[{"relation":"model","target_id":"reported-model-52eee4cc67ca26"},{"relation":"benchmark","target_id":"reported-task-dc82fcbfb44935"},{"relation":"dataset","target_id":"reported-dataset-8317793f18b026"}],"attributes":{"origin":"independent_paper","protocol":"Family-wise evaluation of canonical base-pair predictions.","version":null,"comparison":{"protocol_id":null,"dataset_version":"116 RNAs","split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-013","kind":"evaluation","name":"TU-Fold (aug): RNA secondary structure","description":"","status":"needs_review","facets":{"areas":["rna-transcriptomes"],"tasks":["RNA secondary structure"]},"source_ids":["tu-fold-2025"],"links":[{"relation":"model","target_id":"reported-model-d1cd9a425f9bbd"},{"relation":"benchmark","target_id":"reported-task-5ec7581b246ea6"},{"relation":"dataset","target_id":"reported-dataset-f2e729f333a333"}],"attributes":{"origin":"author_reported","protocol":"Three-fold training and evaluation; source reports mean and standard deviation.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-014","kind":"evaluation","name":"UFold: RNA secondary structure","description":"","status":"needs_review","facets":{"areas":["rna-transcriptomes"],"tasks":["RNA secondary structure"]},"source_ids":["tu-fold-2025"],"links":[{"relation":"model","target_id":"reported-model-e3abb0b9a2ec79"},{"relation":"benchmark","target_id":"reported-task-5ec7581b246ea6"},{"relation":"dataset","target_id":"reported-dataset-f2e729f333a333"}],"attributes":{"origin":"independent_paper","protocol":"Three-fold training and evaluation; source reports mean and standard deviation.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-015","kind":"evaluation","name":"DEBFold: RNA secondary structure","description":"","status":"needs_review","facets":{"areas":["rna-transcriptomes"],"tasks":["RNA secondary structure"]},"source_ids":["debfold-2024"],"links":[{"relation":"model","target_id":"reported-model-3f850c08d76410"},{"relation":"benchmark","target_id":"reported-task-016f70615f2cfc"},{"relation":"dataset","target_id":"reported-dataset-18ebde58579c2b"}],"attributes":{"origin":"author_reported","protocol":"Median F1 on the prepared TestSetβ.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-016","kind":"evaluation","name":"RNAfold: RNA secondary structure","description":"","status":"needs_review","facets":{"areas":["rna-transcriptomes"],"tasks":["RNA secondary structure"]},"source_ids":["debfold-2024"],"links":[{"relation":"model","target_id":"reported-model-ed7f0db85facb1"},{"relation":"benchmark","target_id":"reported-task-016f70615f2cfc"},{"relation":"dataset","target_id":"reported-dataset-18ebde58579c2b"}],"attributes":{"origin":"independent_paper","protocol":"Median F1 on the prepared TestSetβ.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-017","kind":"evaluation","name":"ESM-2: Zero-shot substitution mutation effects: stability","description":"","status":"needs_review","facets":{"areas":["proteins-complexes"],"tasks":["Zero-shot substitution mutation effects: stability"]},"source_ids":["proteingym-2023"],"links":[{"relation":"model","target_id":"reported-model-d326e3c4e3ba20"},{"relation":"benchmark","target_id":"reported-task-6243658a1bc215"},{"relation":"dataset","target_id":"reported-dataset-7cec655cd742f3"}],"attributes":{"origin":"independent_paper","protocol":"Zero-shot mutation scores; average Spearman across stability-category assays.","version":"15B","comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-018","kind":"evaluation","name":"ProteinMPNN: Zero-shot substitution mutation effects: stability","description":"","status":"needs_review","facets":{"areas":["proteins-complexes"],"tasks":["Zero-shot substitution mutation effects: stability"]},"source_ids":["proteingym-2023"],"links":[{"relation":"model","target_id":"reported-model-e0443048c6e110"},{"relation":"benchmark","target_id":"reported-task-6243658a1bc215"},{"relation":"dataset","target_id":"reported-dataset-7cec655cd742f3"}],"attributes":{"origin":"independent_paper","protocol":"Zero-shot mutation scores; average Spearman across stability-category assays.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-019","kind":"evaluation","name":"FUJISAN: Enzyme functional identity prediction","description":"","status":"needs_review","facets":{"areas":["proteins-complexes"],"tasks":["Enzyme functional identity prediction"]},"source_ids":["fujisan-2024"],"links":[{"relation":"model","target_id":"reported-model-9c10fbec02a365"},{"relation":"benchmark","target_id":"reported-task-1ebf9b408517f9"},{"relation":"dataset","target_id":"reported-dataset-5197cca532f89d"}],"attributes":{"origin":"author_reported","protocol":"Sequence and structural feature integration; paper-reported test sub-dataset.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-020","kind":"evaluation","name":"ESM2: Enzyme functional identity prediction","description":"","status":"needs_review","facets":{"areas":["proteins-complexes"],"tasks":["Enzyme functional identity prediction"]},"source_ids":["fujisan-2024"],"links":[{"relation":"model","target_id":"reported-model-ccd1160ad4ec27"},{"relation":"benchmark","target_id":"reported-task-1ebf9b408517f9"},{"relation":"dataset","target_id":"reported-dataset-5197cca532f89d"}],"attributes":{"origin":"independent_paper","protocol":"Comparator evaluated on the paper-reported test sub-dataset.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-021","kind":"evaluation","name":"ESM-2: Mutated RBD binding prediction","description":"","status":"needs_review","facets":{"areas":["proteins-complexes"],"tasks":["Mutated RBD binding prediction"]},"source_ids":["prime-2026"],"links":[{"relation":"model","target_id":"reported-model-d0d5df2beb02b2"},{"relation":"benchmark","target_id":"reported-task-00e594df6a182d"},{"relation":"dataset","target_id":"reported-dataset-becc215358afd0"}],"attributes":{"origin":"independent_paper","protocol":"Frozen mean-pooled representation with downstream regression; position-stratified split.","version":"8M","comparison":{"protocol_id":null,"dataset_version":null,"split":"position-stratified","population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-022","kind":"evaluation","name":"ESM-C: Mutated RBD binding prediction","description":"","status":"needs_review","facets":{"areas":["proteins-complexes"],"tasks":["Mutated RBD binding prediction"]},"source_ids":["prime-2026"],"links":[{"relation":"model","target_id":"reported-model-f83c0b833411a7"},{"relation":"benchmark","target_id":"reported-task-00e594df6a182d"},{"relation":"dataset","target_id":"reported-dataset-becc215358afd0"}],"attributes":{"origin":"independent_paper","protocol":"Frozen mean-pooled representation with downstream regression; position-stratified split.","version":"300M","comparison":{"protocol_id":null,"dataset_version":null,"split":"position-stratified","population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-023","kind":"evaluation","name":"PST: Zero-shot variant effect prediction","description":"","status":"needs_review","facets":{"areas":["proteins-complexes"],"tasks":["Zero-shot variant effect prediction"]},"source_ids":["pst-2025"],"links":[{"relation":"model","target_id":"reported-model-7dd5992188a868"},{"relation":"benchmark","target_id":"reported-task-a5141363b0ee45"},{"relation":"dataset","target_id":"reported-dataset-bd9255afb783d6"}],"attributes":{"origin":"author_reported","protocol":"Zero-shot VEP; paper averages absolute Spearman correlations.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-024","kind":"evaluation","name":"ESM-2: Zero-shot variant effect prediction","description":"","status":"needs_review","facets":{"areas":["proteins-complexes"],"tasks":["Zero-shot variant effect prediction"]},"source_ids":["pst-2025"],"links":[{"relation":"model","target_id":"reported-model-d25dab1a9c4fff"},{"relation":"benchmark","target_id":"reported-task-a5141363b0ee45"},{"relation":"dataset","target_id":"reported-dataset-bd9255afb783d6"}],"attributes":{"origin":"independent_paper","protocol":"Zero-shot VEP; paper averages absolute Spearman correlations.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-025","kind":"evaluation","name":"scGPT: Cell-type identification","description":"","status":"needs_review","facets":{"areas":["cells-tissues"],"tasks":["Cell-type identification"]},"source_ids":["single-cell-peft-2024"],"links":[{"relation":"model","target_id":"reported-model-3bdb3093e8d531"},{"relation":"benchmark","target_id":"reported-task-5b929593eefc76"},{"relation":"dataset","target_id":"reported-dataset-488d5de6bb9c1b"}],"attributes":{"origin":"independent_paper","protocol":"Native scLLM cell-type identification as reported in Table 2.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-026","kind":"evaluation","name":"Geneformer: Cell-type identification","description":"","status":"needs_review","facets":{"areas":["cells-tissues"],"tasks":["Cell-type identification"]},"source_ids":["single-cell-peft-2024"],"links":[{"relation":"model","target_id":"reported-model-b46ae14b9927ac"},{"relation":"benchmark","target_id":"reported-task-5b929593eefc76"},{"relation":"dataset","target_id":"reported-dataset-488d5de6bb9c1b"}],"attributes":{"origin":"independent_paper","protocol":"Native scLLM cell-type identification as reported in Table 2.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-027","kind":"evaluation","name":"C2S (GPT-2 Large): Combinatorial cell-label classification","description":"","status":"needs_review","facets":{"areas":["cells-tissues"],"tasks":["Combinatorial cell-label classification"]},"source_ids":["cell2sentence-2024"],"links":[{"relation":"model","target_id":"reported-model-ab02228f50a37c"},{"relation":"benchmark","target_id":"reported-task-7efe245cc94ee5"},{"relation":"dataset","target_id":"reported-dataset-87e91d9d6e6f4b"}],"attributes":{"origin":"author_reported","protocol":"Partial-credit labels including cell type, perturbation, and dose.","version":"GPT-2 Large","comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-028","kind":"evaluation","name":"Geneformer: Combinatorial cell-label classification","description":"","status":"needs_review","facets":{"areas":["cells-tissues"],"tasks":["Combinatorial cell-label classification"]},"source_ids":["cell2sentence-2024"],"links":[{"relation":"model","target_id":"reported-model-06816ce9073144"},{"relation":"benchmark","target_id":"reported-task-7efe245cc94ee5"},{"relation":"dataset","target_id":"reported-dataset-87e91d9d6e6f4b"}],"attributes":{"origin":"independent_paper","protocol":"Partial-credit labels including cell type, perturbation, and dose.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-029","kind":"evaluation","name":"scGPT: Cell-type annotation","description":"","status":"needs_review","facets":{"areas":["cells-tissues"],"tasks":["Cell-type annotation"]},"source_ids":["scelmo-2025"],"links":[{"relation":"model","target_id":"reported-model-77ad27d4098177"},{"relation":"benchmark","target_id":"reported-task-660753ec94e631"},{"relation":"dataset","target_id":"reported-dataset-8f123f006964ad"}],"attributes":{"origin":"paper_compilation","protocol":"Zero-shot setting; source caption says some comparator rows come from GenePT.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","original_evaluation":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-030","kind":"evaluation","name":"Geneformer: Cell-type annotation","description":"","status":"needs_review","facets":{"areas":["cells-tissues"],"tasks":["Cell-type annotation"]},"source_ids":["scelmo-2025"],"links":[{"relation":"model","target_id":"reported-model-d60f505aabb19c"},{"relation":"benchmark","target_id":"reported-task-660753ec94e631"},{"relation":"dataset","target_id":"reported-dataset-8f123f006964ad"}],"attributes":{"origin":"paper_compilation","protocol":"Zero-shot setting; source caption says some comparator rows come from GenePT.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","original_evaluation":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-031","kind":"evaluation","name":"scRegNet (Geneformer backbone): Gene-regulatory link prediction","description":"","status":"needs_review","facets":{"areas":["cells-tissues"],"tasks":["Gene-regulatory link prediction"]},"source_ids":["scregnet-2025"],"links":[{"relation":"model","target_id":"reported-model-60455ff7cc0c15"},{"relation":"benchmark","target_id":"reported-task-3063ed4da76b4b"},{"relation":"dataset","target_id":"reported-dataset-2ad2fad5e1cd0a"}],"attributes":{"origin":"author_reported","protocol":"TFs plus 500 variable genes; mean from 50 independent evaluations.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-032","kind":"evaluation","name":"scRegNet (scBERT backbone): Gene-regulatory link prediction","description":"","status":"needs_review","facets":{"areas":["cells-tissues"],"tasks":["Gene-regulatory link prediction"]},"source_ids":["scregnet-2025"],"links":[{"relation":"model","target_id":"reported-model-89f5a8f309fa18"},{"relation":"benchmark","target_id":"reported-task-3063ed4da76b4b"},{"relation":"dataset","target_id":"reported-dataset-2ad2fad5e1cd0a"}],"attributes":{"origin":"author_reported","protocol":"TFs plus 500 variable genes; mean from 50 independent evaluations.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-033","kind":"evaluation","name":"ProkBERT-mini: E. coli sigma70 promoter prediction","description":"","status":"needs_review","facets":{"areas":["microbes-communities"],"tasks":["E. coli sigma70 promoter prediction"]},"source_ids":["prokbert-2024"],"links":[{"relation":"model","target_id":"reported-model-4438513d9cd42c"},{"relation":"benchmark","target_id":"reported-task-3891811dcce8b3"},{"relation":"dataset","target_id":"reported-dataset-48def1da574597"}],"attributes":{"origin":"author_reported","protocol":"Promoter versus non-promoter classification.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-034","kind":"evaluation","name":"Promotech: E. coli sigma70 promoter prediction","description":"","status":"needs_review","facets":{"areas":["microbes-communities"],"tasks":["E. coli sigma70 promoter prediction"]},"source_ids":["prokbert-2024"],"links":[{"relation":"model","target_id":"reported-model-0d147487bf97be"},{"relation":"benchmark","target_id":"reported-task-3891811dcce8b3"},{"relation":"dataset","target_id":"reported-dataset-48def1da574597"}],"attributes":{"origin":"independent_paper","protocol":"Promoter versus non-promoter classification.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-035","kind":"evaluation","name":"Eco70PromBERT: E. coli sigma70 promoter prediction","description":"","status":"needs_review","facets":{"areas":["microbes-communities"],"tasks":["E. coli sigma70 promoter prediction"]},"source_ids":["cyaprombert-2022"],"links":[{"relation":"model","target_id":"reported-model-5b70fccb70bb70"},{"relation":"benchmark","target_id":"reported-task-e5c34f686ac403"},{"relation":"dataset","target_id":"reported-dataset-a1da4a37eb46a5"}],"attributes":{"origin":"author_reported","protocol":"BERT-base with 1bp tokenizer; 110 promoters and 108 non-promoters.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":"independent test","population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-036","kind":"evaluation","name":"iPro70-FMWin: E. coli sigma70 promoter prediction","description":"","status":"needs_review","facets":{"areas":["microbes-communities"],"tasks":["E. coli sigma70 promoter prediction"]},"source_ids":["cyaprombert-2022"],"links":[{"relation":"model","target_id":"reported-model-23cb15b93c00ff"},{"relation":"benchmark","target_id":"reported-task-e5c34f686ac403"},{"relation":"dataset","target_id":"reported-dataset-a1da4a37eb46a5"}],"attributes":{"origin":"independent_paper","protocol":"Compared on the same independent test dataset; 110 promoters and 108 non-promoters.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":"independent test","population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-037","kind":"evaluation","name":"EVO2: Genome-wide prophage detection","description":"","status":"needs_review","facets":{"areas":["microbes-communities"],"tasks":["Genome-wide prophage detection"]},"source_ids":["lambda-prophage-2026"],"links":[{"relation":"model","target_id":"reported-model-aa763db2cfdeff"},{"relation":"benchmark","target_id":"reported-task-dd001540e0f4ec"},{"relation":"dataset","target_id":"reported-dataset-1b4f6ea24c0587"}],"attributes":{"origin":"independent_paper","protocol":"Genomic language model fine-tuned for prophage detection; genome-wide evaluation.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-038","kind":"evaluation","name":"geNomad: Genome-wide prophage detection","description":"","status":"needs_review","facets":{"areas":["microbes-communities"],"tasks":["Genome-wide prophage detection"]},"source_ids":["lambda-prophage-2026"],"links":[{"relation":"model","target_id":"reported-model-d0677d52d2b9fd"},{"relation":"benchmark","target_id":"reported-task-dd001540e0f4ec"},{"relation":"dataset","target_id":"reported-dataset-1b4f6ea24c0587"}],"attributes":{"origin":"independent_paper","protocol":"Traditional specialist comparator; genome-wide evaluation.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-039","kind":"evaluation","name":"NABAS+: Metagenomic taxonomic classification","description":"","status":"needs_review","facets":{"areas":["microbes-communities"],"tasks":["Metagenomic taxonomic classification"]},"source_ids":["nabas-plus-2025"],"links":[{"relation":"model","target_id":"reported-model-e7d203bd99ca99"},{"relation":"benchmark","target_id":"reported-task-92137759a9e7b0"},{"relation":"dataset","target_id":"reported-dataset-b462aa24561fba"}],"attributes":{"origin":"author_reported","protocol":"Newly generated sample19 used for classifier comparison.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-040","kind":"evaluation","name":"MetaPhlAn3: Metagenomic taxonomic classification","description":"","status":"needs_review","facets":{"areas":["microbes-communities"],"tasks":["Metagenomic taxonomic classification"]},"source_ids":["nabas-plus-2025"],"links":[{"relation":"model","target_id":"reported-model-df4084611520b7"},{"relation":"benchmark","target_id":"reported-task-92137759a9e7b0"},{"relation":"dataset","target_id":"reported-dataset-b462aa24561fba"}],"attributes":{"origin":"independent_paper","protocol":"Newly generated sample19 used for classifier comparison.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-041","kind":"evaluation","name":"Chai-1: Lipid–protein binding pose","description":"","status":"needs_review","facets":{"areas":["molecular-interactions"],"tasks":["Lipid–protein binding pose"]},"source_ids":["lipp-2026"],"links":[{"relation":"model","target_id":"reported-model-eae60780097101"},{"relation":"benchmark","target_id":"reported-task-ff2dec63c5a3dd"},{"relation":"dataset","target_id":"reported-dataset-6a44f5946cd7ab"}],"attributes":{"origin":"independent_paper","protocol":"Top-scoring pose; all-atom lipid RMSD below 2 Å.","version":null,"comparison":{"protocol_id":null,"dataset_version":"331 complexes","split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-042","kind":"evaluation","name":"DiffDock-L: Lipid–protein binding pose","description":"","status":"needs_review","facets":{"areas":["molecular-interactions"],"tasks":["Lipid–protein binding pose"]},"source_ids":["lipp-2026"],"links":[{"relation":"model","target_id":"reported-model-51ed86132346a0"},{"relation":"benchmark","target_id":"reported-task-ff2dec63c5a3dd"},{"relation":"dataset","target_id":"reported-dataset-6a44f5946cd7ab"}],"attributes":{"origin":"independent_paper","protocol":"Top-scoring pose; all-atom lipid RMSD below 2 Å.","version":null,"comparison":{"protocol_id":null,"dataset_version":"331 complexes","split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-043","kind":"evaluation","name":"DiffDock-NMDN: Protein–ligand virtual screening","description":"","status":"needs_review","facets":{"areas":["molecular-interactions"],"tasks":["Protein–ligand virtual screening"]},"source_ids":["nmdn-2025"],"links":[{"relation":"model","target_id":"reported-model-6c0bc8d297cc7a"},{"relation":"benchmark","target_id":"reported-task-a7803ecf7708cc"},{"relation":"dataset","target_id":"reported-dataset-065b9fcc8da573"}],"attributes":{"origin":"author_reported","protocol":"NMDN scoring on DiffDock-NMDN blind docked poses; not ligand-pose RMSD.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-044","kind":"evaluation","name":"Vina: Protein–ligand virtual screening","description":"","status":"needs_review","facets":{"areas":["molecular-interactions"],"tasks":["Protein–ligand virtual screening"]},"source_ids":["nmdn-2025"],"links":[{"relation":"model","target_id":"reported-model-1e51ccbfd2de61"},{"relation":"benchmark","target_id":"reported-task-a7803ecf7708cc"},{"relation":"dataset","target_id":"reported-dataset-065b9fcc8da573"}],"attributes":{"origin":"independent_paper","protocol":"Vina scoring on the same DiffDock-NMDN blind docked poses; not ligand-pose RMSD.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-045","kind":"evaluation","name":"Boltz-1: Protein–ligand pose prediction","description":"","status":"needs_review","facets":{"areas":["molecular-interactions"],"tasks":["Protein–ligand pose prediction"]},"source_ids":["boltz-stereochemistry-2025"],"links":[{"relation":"model","target_id":"reported-model-d9a06805b36b8a"},{"relation":"benchmark","target_id":"reported-task-bf513ed6db92c5"},{"relation":"dataset","target_id":"reported-dataset-5afaefb87c8a94"}],"attributes":{"origin":"independent_paper","protocol":"All entries; authors note this dataset contains structures seen during model training.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-046","kind":"evaluation","name":"DiffDock: Protein–ligand pose prediction","description":"","status":"needs_review","facets":{"areas":["molecular-interactions"],"tasks":["Protein–ligand pose prediction"]},"source_ids":["boltz-stereochemistry-2025"],"links":[{"relation":"model","target_id":"reported-model-7f6ffd9e2a08be"},{"relation":"benchmark","target_id":"reported-task-bf513ed6db92c5"},{"relation":"dataset","target_id":"reported-dataset-5afaefb87c8a94"}],"attributes":{"origin":"independent_paper","protocol":"All entries; rigid-protein docking comparator; authors note this dataset contains structures seen during model training.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-047","kind":"evaluation","name":"Boltz-2: Ligand potency prediction using generated poses","description":"","status":"needs_review","facets":{"areas":["molecular-interactions"],"tasks":["Ligand potency prediction using generated poses"]},"source_ids":["mpro-pose-affinity-2025"],"links":[{"relation":"model","target_id":"reported-model-cdc9aabf4efc04"},{"relation":"benchmark","target_id":"reported-task-d5f897ab0f6f67"},{"relation":"dataset","target_id":"reported-dataset-235520c84b737f"}],"attributes":{"origin":"independent_paper","protocol":"Potency prediction using Boltz-2 ligand-pose generation protocol; see paper scoring pipeline.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-048","kind":"evaluation","name":"DiffDock: Ligand potency prediction using generated poses","description":"","status":"needs_review","facets":{"areas":["molecular-interactions"],"tasks":["Ligand potency prediction using generated poses"]},"source_ids":["mpro-pose-affinity-2025"],"links":[{"relation":"model","target_id":"reported-model-415ee22f46526c"},{"relation":"benchmark","target_id":"reported-task-d5f897ab0f6f67"},{"relation":"dataset","target_id":"reported-dataset-235520c84b737f"}],"attributes":{"origin":"independent_paper","protocol":"Potency prediction using DiffDock ligand-pose generation plus paper scoring pipeline; not a native DiffDock affinity score.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b3-003","kind":"evaluation","name":"Mouse-Geneformer: Human thymus cell-type classification","description":"","status":"needs_review","facets":{"areas":["cells-tissues"],"tasks":["Human thymus cell-type classification"]},"source_ids":["mouse-geneformer-2025"],"links":[{"relation":"model","target_id":"reported-model-e2f2f0d4830bb0"},{"relation":"benchmark","target_id":"reported-task-031186b57c62de"},{"relation":"dataset","target_id":"reported-dataset-477a9082515406"}],"attributes":{"origin":"author_reported","protocol":"Ortholog-based gene conversion; zero-shot mouse model on human cells.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b3-004","kind":"evaluation","name":"Human-Geneformer: Human thymus cell-type classification","description":"","status":"needs_review","facets":{"areas":["cells-tissues"],"tasks":["Human thymus cell-type classification"]},"source_ids":["mouse-geneformer-2025"],"links":[{"relation":"model","target_id":"reported-model-10d85f2a035720"},{"relation":"benchmark","target_id":"reported-task-031186b57c62de"},{"relation":"dataset","target_id":"reported-dataset-477a9082515406"}],"attributes":{"origin":"independent_paper","protocol":"Native human model; zero-shot setting.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b3-005","kind":"evaluation","name":"scLLMDA: Cross-platform scATAC cell-type annotation","description":"","status":"needs_review","facets":{"areas":["cells-tissues"],"tasks":["Cross-platform scATAC cell-type annotation"]},"source_ids":["scatac-llmda-2026"],"links":[{"relation":"model","target_id":"reported-model-a0db32ae53e5ed"},{"relation":"benchmark","target_id":"reported-task-d82b6284f3f431"},{"relation":"dataset","target_id":"reported-dataset-7fc59ce4c0ceaa"}],"attributes":{"origin":"author_reported","protocol":"Cross-platform reference-query cell-type annotation.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b3-006","kind":"evaluation","name":"MINGLE: Cross-platform scATAC cell-type annotation","description":"","status":"needs_review","facets":{"areas":["cells-tissues"],"tasks":["Cross-platform scATAC cell-type annotation"]},"source_ids":["scatac-llmda-2026"],"links":[{"relation":"model","target_id":"reported-model-f0c630d0565e64"},{"relation":"benchmark","target_id":"reported-task-d82b6284f3f431"},{"relation":"dataset","target_id":"reported-dataset-7fc59ce4c0ceaa"}],"attributes":{"origin":"independent_paper","protocol":"Cross-platform reference-query comparator.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b3-011","kind":"evaluation","name":"GenePT-w: Cell-type structure in frozen embeddings","description":"","status":"needs_review","facets":{"areas":["cells-tissues"],"tasks":["Cell-type structure in frozen embeddings"]},"source_ids":["genept-2024"],"links":[{"relation":"model","target_id":"reported-model-7c595040de69bc"},{"relation":"benchmark","target_id":"reported-task-4df1fb456d3deb"},{"relation":"dataset","target_id":"reported-dataset-eaa2965545c87b"}],"attributes":{"origin":"author_reported","protocol":"k-means on pretrained cell embeddings; agreement with original cell-type labels.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b3-012","kind":"evaluation","name":"scGPT: Cell-type structure in frozen embeddings","description":"","status":"needs_review","facets":{"areas":["cells-tissues"],"tasks":["Cell-type structure in frozen embeddings"]},"source_ids":["genept-2024"],"links":[{"relation":"model","target_id":"reported-model-399b1ce87a3f6d"},{"relation":"benchmark","target_id":"reported-task-4df1fb456d3deb"},{"relation":"dataset","target_id":"reported-dataset-eaa2965545c87b"}],"attributes":{"origin":"independent_paper","protocol":"k-means on pretrained cell embeddings; agreement with original cell-type labels.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b3-013","kind":"evaluation","name":"Best frozen single-cell foundation model: Donor-aware age-class prediction","description":"","status":"needs_review","facets":{"areas":["cells-tissues"],"tasks":["Donor-aware age-class prediction"]},"source_ids":["single-cell-aging-probes-2026"],"links":[{"relation":"model","target_id":"reported-model-148b613975b6eb"},{"relation":"benchmark","target_id":"reported-task-d6018ca598e525"},{"relation":"dataset","target_id":"reported-dataset-571ce000cd74b9"}],"attributes":{"origin":"independent_paper","protocol":"Same donor-aware splits and logistic-regression probe as expression PCA; text names Geneformer as best model on AIDA v2.","version":null,"comparison":{"protocol_id":null,"dataset_version":"622 donors","split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b3-014","kind":"evaluation","name":"Gene-expression PCA: Donor-aware age-class prediction","description":"","status":"needs_review","facets":{"areas":["cells-tissues"],"tasks":["Donor-aware age-class prediction"]},"source_ids":["single-cell-aging-probes-2026"],"links":[{"relation":"model","target_id":"reported-model-177f32ce8189a0"},{"relation":"benchmark","target_id":"reported-task-d6018ca598e525"},{"relation":"dataset","target_id":"reported-dataset-571ce000cd74b9"}],"attributes":{"origin":"independent_paper","protocol":"Fifty-component gene-expression PCA with the same donor-aware probe splits.","version":null,"comparison":{"protocol_id":null,"dataset_version":"622 donors","split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b3-015","kind":"evaluation","name":"scaLR: PBMC cell-type classification","description":"","status":"needs_review","facets":{"areas":["cells-tissues"],"tasks":["PBMC cell-type classification"]},"source_ids":["scalr-2025"],"links":[{"relation":"model","target_id":"reported-model-7cf2f9951e1dba"},{"relation":"benchmark","target_id":"reported-task-b46b7b839bff93"},{"relation":"dataset","target_id":"reported-dataset-33a41fe5fc66cf"}],"attributes":{"origin":"author_reported","protocol":"All features and samples from PBMCs-BS.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b3-016","kind":"evaluation","name":"scVI + scANVI: PBMC cell-type classification","description":"","status":"needs_review","facets":{"areas":["cells-tissues"],"tasks":["PBMC cell-type classification"]},"source_ids":["scalr-2025"],"links":[{"relation":"model","target_id":"reported-model-1be5c4b7c52a41"},{"relation":"benchmark","target_id":"reported-task-b46b7b839bff93"},{"relation":"dataset","target_id":"reported-dataset-33a41fe5fc66cf"}],"attributes":{"origin":"independent_paper","protocol":"All features and samples from PBMCs-BS; comparison pipeline combines scVI and scANVI.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b3-017","kind":"evaluation","name":"scXDR: Cross-dataset single-cell drug response transfer","description":"","status":"needs_review","facets":{"areas":["cells-tissues"],"tasks":["Cross-dataset single-cell drug response transfer"]},"source_ids":["scxdr-2026"],"links":[{"relation":"model","target_id":"reported-model-9f39de53f7a139"},{"relation":"benchmark","target_id":"reported-task-167f08013c270e"},{"relation":"dataset","target_id":"reported-dataset-f7210686a78474"}],"attributes":{"origin":"author_reported","protocol":"Single-cell-to-single-cell transfer; source scenario 2.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b3-018","kind":"evaluation","name":"scVI: Cross-dataset single-cell drug response transfer","description":"","status":"needs_review","facets":{"areas":["cells-tissues"],"tasks":["Cross-dataset single-cell drug response transfer"]},"source_ids":["scxdr-2026"],"links":[{"relation":"model","target_id":"reported-model-f23306b94dc7b6"},{"relation":"benchmark","target_id":"reported-task-167f08013c270e"},{"relation":"dataset","target_id":"reported-dataset-f7210686a78474"}],"attributes":{"origin":"independent_paper","protocol":"Single-cell-to-single-cell transfer; source scenario 2.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b3-019","kind":"evaluation","name":"CAMMiQ: Strain-level abundance quantification","description":"","status":"needs_review","facets":{"areas":["microbes-communities"],"tasks":["Strain-level abundance quantification"]},"source_ids":["cammiq-2022"],"links":[{"relation":"model","target_id":"reported-model-32a19f43a4c254"},{"relation":"benchmark","target_id":"reported-task-571f0a2e7faed3"},{"relation":"dataset","target_id":"reported-dataset-72e837e5b97041"}],"attributes":{"origin":"author_reported","protocol":"Strain-level quantification on the HumanGut-all synthetic query.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b3-020","kind":"evaluation","name":"Kraken2: Strain-level abundance quantification","description":"","status":"needs_review","facets":{"areas":["microbes-communities"],"tasks":["Strain-level abundance quantification"]},"source_ids":["cammiq-2022"],"links":[{"relation":"model","target_id":"reported-model-70f57ebb163a5c"},{"relation":"benchmark","target_id":"reported-task-571f0a2e7faed3"},{"relation":"dataset","target_id":"reported-dataset-72e837e5b97041"}],"attributes":{"origin":"independent_paper","protocol":"Strain-level quantification on the HumanGut-all synthetic query.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b3-021","kind":"evaluation","name":"Lazypipe-nt: Simulated metagenome virus-taxon retrieval","description":"","status":"needs_review","facets":{"areas":["microbes-communities"],"tasks":["Simulated metagenome virus-taxon retrieval"]},"source_ids":["lazypipe-2020"],"links":[{"relation":"model","target_id":"reported-model-8cb3dd4e9f5b10"},{"relation":"benchmark","target_id":"reported-task-369dcfef14c4a9"},{"relation":"dataset","target_id":"reported-dataset-450c1af18cc623"}],"attributes":{"origin":"author_reported","protocol":"Genus-rank viral taxon retrieval.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b3-022","kind":"evaluation","name":"Kraken2: Simulated metagenome virus-taxon retrieval","description":"","status":"needs_review","facets":{"areas":["microbes-communities"],"tasks":["Simulated metagenome virus-taxon retrieval"]},"source_ids":["lazypipe-2020"],"links":[{"relation":"model","target_id":"reported-model-673b8f46361000"},{"relation":"benchmark","target_id":"reported-task-369dcfef14c4a9"},{"relation":"dataset","target_id":"reported-dataset-450c1af18cc623"}],"attributes":{"origin":"independent_paper","protocol":"Genus-rank viral taxon retrieval.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b3-023","kind":"evaluation","name":"NCD-gzip: CAMI II superkingdom read classification","description":"","status":"needs_review","facets":{"areas":["microbes-communities"],"tasks":["CAMI II superkingdom read classification"]},"source_ids":["ncd-metagenomics-2026"],"links":[{"relation":"model","target_id":"reported-model-7b052acf17b5ba"},{"relation":"benchmark","target_id":"reported-task-a2c37b8c420bc3"},{"relation":"dataset","target_id":"reported-dataset-beb4f5da29da0a"}],"attributes":{"origin":"author_reported","protocol":"Superkingdom-level macro-averaged F1; NCD assigns every read.","version":null,"comparison":{"protocol_id":null,"dataset_version":"10,000 reads","split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b3-024","kind":"evaluation","name":"NCD-gzip: CAMI II phylum read classification","description":"","status":"needs_review","facets":{"areas":["microbes-communities"],"tasks":["CAMI II phylum read classification"]},"source_ids":["ncd-metagenomics-2026"],"links":[{"relation":"model","target_id":"reported-model-7b052acf17b5ba"},{"relation":"benchmark","target_id":"reported-task-45105e1c486251"},{"relation":"dataset","target_id":"reported-dataset-beb4f5da29da0a"}],"attributes":{"origin":"author_reported","protocol":"Phylum-level macro-averaged F1; distinct taxonomic rank from the other row.","version":null,"comparison":{"protocol_id":null,"dataset_version":"10,000 reads","split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b3-025","kind":"evaluation","name":"VIBRANT: Simulated prophage-contig detection","description":"","status":"needs_review","facets":{"areas":["microbes-communities"],"tasks":["Simulated prophage-contig detection"]},"source_ids":["viral-contig-simulation-2021"],"links":[{"relation":"model","target_id":"reported-model-27dca28a87cf3c"},{"relation":"benchmark","target_id":"reported-task-53e3d216eef6db"},{"relation":"dataset","target_id":"reported-dataset-cd51026cdb6a7a"}],"attributes":{"origin":"independent_paper","protocol":"Average across twenty medium- and high-complexity simulated communities.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b3-026","kind":"evaluation","name":"VirSorter: Simulated prophage-contig detection","description":"","status":"needs_review","facets":{"areas":["microbes-communities"],"tasks":["Simulated prophage-contig detection"]},"source_ids":["viral-contig-simulation-2021"],"links":[{"relation":"model","target_id":"reported-model-e78e3886df0d3a"},{"relation":"benchmark","target_id":"reported-task-53e3d216eef6db"},{"relation":"dataset","target_id":"reported-dataset-cd51026cdb6a7a"}],"attributes":{"origin":"independent_paper","protocol":"Average across twenty medium- and high-complexity simulated communities.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b3-027","kind":"evaluation","name":"GenomeOcean: Natural vs artificial microbial genome sequence","description":"","status":"needs_review","facets":{"areas":["microbes-communities"],"tasks":["Natural vs artificial microbial genome sequence"]},"source_ids":["genomeocean-2025"],"links":[{"relation":"model","target_id":"reported-model-2df975e60d16d1"},{"relation":"benchmark","target_id":"reported-task-9f9ab0090f6522"},{"relation":"dataset","target_id":"reported-dataset-0c3ac7efe99c37"}],"attributes":{"origin":"author_reported","protocol":"Source reports natural-versus-artificial sequence classification.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b3-028","kind":"evaluation","name":"DNABERT-2: Natural vs artificial microbial genome sequence","description":"","status":"needs_review","facets":{"areas":["microbes-communities"],"tasks":["Natural vs artificial microbial genome sequence"]},"source_ids":["genomeocean-2025"],"links":[{"relation":"model","target_id":"reported-model-7f1165b35f10e2"},{"relation":"benchmark","target_id":"reported-task-9f9ab0090f6522"},{"relation":"dataset","target_id":"reported-dataset-0c3ac7efe99c37"}],"attributes":{"origin":"independent_paper","protocol":"Source reports natural-versus-artificial sequence classification.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b3-029","kind":"evaluation","name":"kMetaShot: Mock-community MAG taxonomy classification","description":"","status":"needs_review","facets":{"areas":["microbes-communities"],"tasks":["Mock-community MAG taxonomy classification"]},"source_ids":["kmetashot-2025"],"links":[{"relation":"model","target_id":"reported-model-790768ed581685"},{"relation":"benchmark","target_id":"reported-task-8406b6aabfb8c0"},{"relation":"dataset","target_id":"reported-dataset-8cad416ddc80dc"}],"attributes":{"origin":"author_reported","protocol":"Genus classification of MAGs from MegaHIT contigs; uncorrected kMetaShot.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b3-030","kind":"evaluation","name":"GTDB-Tk: Mock-community MAG taxonomy classification","description":"","status":"needs_review","facets":{"areas":["microbes-communities"],"tasks":["Mock-community MAG taxonomy classification"]},"source_ids":["kmetashot-2025"],"links":[{"relation":"model","target_id":"reported-model-3fd1e9f6c573b2"},{"relation":"benchmark","target_id":"reported-task-8406b6aabfb8c0"},{"relation":"dataset","target_id":"reported-dataset-8cad416ddc80dc"}],"attributes":{"origin":"independent_paper","protocol":"Genus classification of the same MAG set.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b3-031","kind":"evaluation","name":"Lemur: Long-read taxonomic profiling","description":"","status":"needs_review","facets":{"areas":["microbes-communities"],"tasks":["Long-read taxonomic profiling"]},"source_ids":["lemur-magnet-2024"],"links":[{"relation":"model","target_id":"reported-model-6ac0730e8481de"},{"relation":"benchmark","target_id":"reported-task-6330d593980b5b"},{"relation":"dataset","target_id":"reported-dataset-768a7ff5bac414"}],"attributes":{"origin":"author_reported","protocol":"Mean across five replicate runs on Zymo LOG 10%.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b3-032","kind":"evaluation","name":"Kraken 2: Long-read taxonomic profiling","description":"","status":"needs_review","facets":{"areas":["microbes-communities"],"tasks":["Long-read taxonomic profiling"]},"source_ids":["lemur-magnet-2024"],"links":[{"relation":"model","target_id":"reported-model-62bc5e5ba13e7d"},{"relation":"benchmark","target_id":"reported-task-6330d593980b5b"},{"relation":"dataset","target_id":"reported-dataset-768a7ff5bac414"}],"attributes":{"origin":"independent_paper","protocol":"Mean across five replicate runs on Zymo LOG 10%.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b3-033","kind":"evaluation","name":"iPro-MP: Multi-species prokaryotic promoter detection","description":"","status":"needs_review","facets":{"areas":["microbes-communities"],"tasks":["Multi-species prokaryotic promoter detection"]},"source_ids":["ipromp-2025"],"links":[{"relation":"model","target_id":"reported-model-9a200c55b0e03e"},{"relation":"benchmark","target_id":"reported-task-a1151e386a3d3f"},{"relation":"dataset","target_id":"reported-dataset-e0f34dcaa1ba3b"}],"attributes":{"origin":"author_reported","protocol":"Average over independent testing sets.","version":null,"comparison":{"protocol_id":null,"dataset_version":"23 test sets","split":"independent test","population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b3-034","kind":"evaluation","name":"Prompt: Multi-species prokaryotic promoter detection","description":"","status":"needs_review","facets":{"areas":["microbes-communities"],"tasks":["Multi-species prokaryotic promoter detection"]},"source_ids":["ipromp-2025"],"links":[{"relation":"model","target_id":"reported-model-03080a5289c07e"},{"relation":"benchmark","target_id":"reported-task-a1151e386a3d3f"},{"relation":"dataset","target_id":"reported-dataset-e0f34dcaa1ba3b"}],"attributes":{"origin":"independent_paper","protocol":"Average over the same independent testing sets.","version":null,"comparison":{"protocol_id":null,"dataset_version":"23 test sets","split":"independent test","population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b3-035","kind":"evaluation","name":"ICCTax: Hierarchical metagenomic taxonomy classification","description":"","status":"needs_review","facets":{"areas":["microbes-communities"],"tasks":["Hierarchical metagenomic taxonomy classification"]},"source_ids":["icctax-2025"],"links":[{"relation":"model","target_id":"reported-model-67eaf766fa9877"},{"relation":"benchmark","target_id":"reported-task-f4b1c9373f0929"},{"relation":"dataset","target_id":"reported-dataset-561834dfa1682c"}],"attributes":{"origin":"author_reported","protocol":"Macro average precision at genus rank on Complete dataset.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b3-036","kind":"evaluation","name":"Kraken2: Hierarchical metagenomic taxonomy classification","description":"","status":"needs_review","facets":{"areas":["microbes-communities"],"tasks":["Hierarchical metagenomic taxonomy classification"]},"source_ids":["icctax-2025"],"links":[{"relation":"model","target_id":"reported-model-8861b9ad9b9c9b"},{"relation":"benchmark","target_id":"reported-task-f4b1c9373f0929"},{"relation":"dataset","target_id":"reported-dataset-561834dfa1682c"}],"attributes":{"origin":"independent_paper","protocol":"Macro average precision at genus rank on Complete dataset.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b3-037","kind":"evaluation","name":"Chai-1: Antibody–antigen interaction prediction using folded complexes","description":"","status":"needs_review","facets":{"areas":["molecular-interactions"],"tasks":["Antibody–antigen interaction prediction using folded complexes"]},"source_ids":["antibody-flexibility-2025"],"links":[{"relation":"model","target_id":"reported-model-b3fdf259d51533"},{"relation":"benchmark","target_id":"reported-task-0c92cda11228c4"},{"relation":"dataset","target_id":"reported-dataset-ee26acbd6e8cf7"}],"attributes":{"origin":"independent_paper","protocol":"Interaction classifier evaluated using Chai-1-folded input complexes; this is pipeline AUC, not DockQ.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b3-038","kind":"evaluation","name":"Boltz-1: Antibody–antigen interaction prediction using folded complexes","description":"","status":"needs_review","facets":{"areas":["molecular-interactions"],"tasks":["Antibody–antigen interaction prediction using folded complexes"]},"source_ids":["antibody-flexibility-2025"],"links":[{"relation":"model","target_id":"reported-model-884582fb0c70dc"},{"relation":"benchmark","target_id":"reported-task-0c92cda11228c4"},{"relation":"dataset","target_id":"reported-dataset-ee26acbd6e8cf7"}],"attributes":{"origin":"independent_paper","protocol":"Interaction classifier evaluated using Boltz-1-folded input complexes; this is pipeline AUC, not DockQ.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b3-039","kind":"evaluation","name":"Boltz-1: Protein–ligand pose prediction","description":"","status":"needs_review","facets":{"areas":["molecular-interactions"],"tasks":["Protein–ligand pose prediction"]},"source_ids":["boltz1-2025"],"links":[{"relation":"model","target_id":"reported-model-4058c43eb73b90"},{"relation":"benchmark","target_id":"reported-task-c04bb5ee6ecea6"},{"relation":"dataset","target_id":"reported-dataset-e45a5a140888ee"}],"attributes":{"origin":"author_reported","protocol":"Highest-confidence pose from five samples; precomputed MSAs up to 4,096 sequences.","version":"3 recycling rounds; 200 diffusion steps","comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b3-040","kind":"evaluation","name":"Ibex: Antibody loop structure prediction","description":"","status":"needs_review","facets":{"areas":["molecular-interactions"],"tasks":["Antibody loop structure prediction"]},"source_ids":["ibex-2025"],"links":[{"relation":"model","target_id":"reported-model-2ae5fb0c147618"},{"relation":"benchmark","target_id":"reported-task-f3a12dbc0e0439"},{"relation":"dataset","target_id":"reported-dataset-b7204b005bd476"}],"attributes":{"origin":"author_reported","protocol":"Backbone RMSD after framework alignment; average over antibody test structures.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b3-041","kind":"evaluation","name":"Chai-1: Antibody loop structure prediction","description":"","status":"needs_review","facets":{"areas":["molecular-interactions"],"tasks":["Antibody loop structure prediction"]},"source_ids":["ibex-2025"],"links":[{"relation":"model","target_id":"reported-model-70c732770a200f"},{"relation":"benchmark","target_id":"reported-task-f3a12dbc0e0439"},{"relation":"dataset","target_id":"reported-dataset-b7204b005bd476"}],"attributes":{"origin":"independent_paper","protocol":"Backbone RMSD after framework alignment; one seed and one diffusion trajectory.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b3-042","kind":"evaluation","name":"DEELIG: Protein–ligand binding affinity prediction","description":"","status":"needs_review","facets":{"areas":["molecular-interactions"],"tasks":["Protein–ligand binding affinity prediction"]},"source_ids":["deelig-2021"],"links":[{"relation":"model","target_id":"reported-model-fa2da404b4d08e"},{"relation":"benchmark","target_id":"reported-task-d81be76396e644"},{"relation":"dataset","target_id":"reported-dataset-327cfcdae0c937"}],"attributes":{"origin":"author_reported","protocol":"Source paper reports DEELIG on PDBbind core set.","version":null,"comparison":{"protocol_id":null,"dataset_version":"v2016","split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b3-043","kind":"evaluation","name":"TOPBP (Complex): Protein–ligand binding affinity prediction","description":"","status":"needs_review","facets":{"areas":["molecular-interactions"],"tasks":["Protein–ligand binding affinity prediction"]},"source_ids":["deelig-2021"],"links":[{"relation":"model","target_id":"reported-model-0068c3eff1bf7b"},{"relation":"benchmark","target_id":"reported-task-d81be76396e644"},{"relation":"dataset","target_id":"reported-dataset-327cfcdae0c937"}],"attributes":{"origin":"paper_compilation","protocol":"Source table compiles a previously published comparator; protocol equivalence is not established.","version":null,"comparison":{"protocol_id":null,"dataset_version":"v2016","split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","original_evaluation":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b3-044","kind":"evaluation","name":"MolAS: Physically valid protein–ligand pose selection","description":"","status":"needs_review","facets":{"areas":["molecular-interactions"],"tasks":["Physically valid protein–ligand pose selection"]},"source_ids":["molas-2026"],"links":[{"relation":"model","target_id":"reported-model-0eb4b0535b58e3"},{"relation":"benchmark","target_id":"reported-task-d1c46526c39983"},{"relation":"dataset","target_id":"reported-dataset-4b6c13924d4256"}],"attributes":{"origin":"author_reported","protocol":"Averaged five-fold algorithm-selection performance on PoseBusters; joint RMSD and validity criterion.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b3-045","kind":"evaluation","name":"Single best solver: Physically valid protein–ligand pose selection","description":"","status":"needs_review","facets":{"areas":["molecular-interactions"],"tasks":["Physically valid protein–ligand pose selection"]},"source_ids":["molas-2026"],"links":[{"relation":"model","target_id":"reported-model-028e4bb9baa074"},{"relation":"benchmark","target_id":"reported-task-d1c46526c39983"},{"relation":"dataset","target_id":"reported-dataset-4b6c13924d4256"}],"attributes":{"origin":"independent_paper","protocol":"Single best solver baseline under the same averaged five-fold selection test.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b3-046","kind":"evaluation","name":"AutoDock Vina holo: Intrinsically disordered protein ensemble docking","description":"","status":"needs_review","facets":{"areas":["molecular-interactions"],"tasks":["Intrinsically disordered protein ensemble docking"]},"source_ids":["ensemble-idp-docking-2025"],"links":[{"relation":"model","target_id":"reported-model-1587ab674d30a2"},{"relation":"benchmark","target_id":"reported-task-dec9e0f5e3da2a"},{"relation":"dataset","target_id":"reported-dataset-00201f65c32f6d"}],"attributes":{"origin":"independent_paper","protocol":"Fraction of docked frames within 3 Å of MD-observed bound pose; holo protein ensemble.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b3-047","kind":"evaluation","name":"DiffDock holo: Intrinsically disordered protein ensemble docking","description":"","status":"needs_review","facets":{"areas":["molecular-interactions"],"tasks":["Intrinsically disordered protein ensemble docking"]},"source_ids":["ensemble-idp-docking-2025"],"links":[{"relation":"model","target_id":"reported-model-75e4e5e5965320"},{"relation":"benchmark","target_id":"reported-task-dec9e0f5e3da2a"},{"relation":"dataset","target_id":"reported-dataset-00201f65c32f6d"}],"attributes":{"origin":"independent_paper","protocol":"Fraction of docked frames within 3 Å of MD-observed bound pose; holo protein ensemble.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b3-048","kind":"evaluation","name":"AK-score-ensemble: Protein–ligand binding affinity scoring","description":"","status":"needs_review","facets":{"areas":["molecular-interactions"],"tasks":["Protein–ligand binding affinity scoring"]},"source_ids":["akscore-2020"],"links":[{"relation":"model","target_id":"reported-model-54be8a811c206e"},{"relation":"benchmark","target_id":"reported-task-a78312d5df6dad"},{"relation":"dataset","target_id":"reported-dataset-f18fcc23dfa798"}],"attributes":{"origin":"author_reported","protocol":"CASF-2016 scoring-power evaluation.","version":"ensemble; learning rate 0.0007","comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b3-049","kind":"evaluation","name":"AK-score-single: Protein–ligand binding affinity scoring","description":"","status":"needs_review","facets":{"areas":["molecular-interactions"],"tasks":["Protein–ligand binding affinity scoring"]},"source_ids":["akscore-2020"],"links":[{"relation":"model","target_id":"reported-model-de89576d8d316b"},{"relation":"benchmark","target_id":"reported-task-a78312d5df6dad"},{"relation":"dataset","target_id":"reported-dataset-f18fcc23dfa798"}],"attributes":{"origin":"author_reported","protocol":"CASF-2016 scoring-power evaluation.","version":"single; learning rate 0.0007","comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b3-050","kind":"evaluation","name":"PMF + ECFP + PF (LightGBM): Protein–ligand binding energy prediction","description":"","status":"needs_review","facets":{"areas":["molecular-interactions"],"tasks":["Protein–ligand binding energy prediction"]},"source_ids":["fingerprint-scoring-2022"],"links":[{"relation":"model","target_id":"reported-model-3c196326586fa7"},{"relation":"benchmark","target_id":"reported-task-94802534b7026d"},{"relation":"dataset","target_id":"reported-dataset-16d01b5ef88e84"}],"attributes":{"origin":"author_reported","protocol":"Binding-energy model using ligand and protein fingerprints with LightGBM.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b3-051","kind":"evaluation","name":"PMF (LASSO): Protein–ligand binding energy prediction","description":"","status":"needs_review","facets":{"areas":["molecular-interactions"],"tasks":["Protein–ligand binding energy prediction"]},"source_ids":["fingerprint-scoring-2022"],"links":[{"relation":"model","target_id":"reported-model-8100b3de6c7811"},{"relation":"benchmark","target_id":"reported-task-94802534b7026d"},{"relation":"dataset","target_id":"reported-dataset-16d01b5ef88e84"}],"attributes":{"origin":"author_reported","protocol":"PMF-only LASSO baseline evaluated by the same authors.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b4-001","kind":"evaluation","name":"ARSENAL+ChromBPNet: regulatory-variant scoring","description":"","status":"needs_review","facets":{"areas":["dna-genomes"],"tasks":["regulatory-variant scoring"]},"source_ids":["arsenal-regulatory-dna-2026"],"links":[{"relation":"model","target_id":"reported-model-ee1ae8162c7d67"},{"relation":"benchmark","target_id":"reported-task-b9199a30a0bcb2"},{"relation":"dataset","target_id":"reported-dataset-158b121281b650"}],"attributes":{"origin":"author_reported","protocol":"Supervised ChromBPNet variant scoring with ARSENAL motif-discovery regularization","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b4-002","kind":"evaluation","name":"PlantCAD2: cross-species conservation prediction","description":"","status":"needs_review","facets":{"areas":["dna-genomes"],"tasks":["cross-species conservation prediction"]},"source_ids":["plantcad2-2025"],"links":[{"relation":"model","target_id":"reported-model-65059c3a806306"},{"relation":"benchmark","target_id":"reported-task-3109f8d0f2b7b5"},{"relation":"dataset","target_id":"reported-dataset-b6d0ebaca196a6"}],"attributes":{"origin":"author_reported","protocol":"Zero-shot score for conserved versus non-conserved sites from alignments of 35 Andropogoneae genomes","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b4-003","kind":"evaluation","name":"Stacking-Auto: enhancer prediction","description":"","status":"needs_review","facets":{"areas":["dna-genomes"],"tasks":["enhancer prediction"]},"source_ids":["hi-enhancer-2025"],"links":[{"relation":"model","target_id":"reported-model-0829aff5471d4b"},{"relation":"benchmark","target_id":"reported-task-22024610c4d658"},{"relation":"dataset","target_id":"reported-dataset-a8610f2b80cdf0"}],"attributes":{"origin":"author_reported","protocol":"Two-stage Hi-Enhancer system; paper Table 2 method comparison","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b4-004","kind":"evaluation","name":"position-aware CNN: enhancer prediction","description":"","status":"needs_review","facets":{"areas":["dna-genomes"],"tasks":["enhancer prediction"]},"source_ids":["enhancer-position-encoding-2024"],"links":[{"relation":"model","target_id":"reported-model-d2c81acf1c42c4"},{"relation":"benchmark","target_id":"reported-task-64607443a9ba15"},{"relation":"dataset","target_id":"reported-dataset-a03b8e9efde37b"}],"attributes":{"origin":"author_reported","protocol":"Nucleotide position-aware feature encoding; average assessment of CNN classifier","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b4-005","kind":"evaluation","name":"ADAR-GPT continual: A-to-I RNA editing site prediction","description":"","status":"needs_review","facets":{"areas":["rna-transcriptomes"],"tasks":["A-to-I RNA editing site prediction"]},"source_ids":["adar-gpt-editing-2026"],"links":[{"relation":"model","target_id":"reported-model-1d2aa9880a1c77"},{"relation":"benchmark","target_id":"reported-task-d635fc6c281a27"},{"relation":"dataset","target_id":"reported-dataset-236eaa4e55147f"}],"attributes":{"origin":"author_reported","protocol":"Curriculum plus 15% fine-tuning; 201-nt sequence windows; decision threshold 0.5","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":"15% validation set","population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b4-006","kind":"evaluation","name":"R3Design: RNA sequence design","description":"","status":"needs_review","facets":{"areas":["rna-transcriptomes"],"tasks":["RNA sequence design"]},"source_ids":["r3design-2025"],"links":[{"relation":"model","target_id":"reported-model-1b5fa066945d3d"},{"relation":"benchmark","target_id":"reported-task-df18c710f45213"},{"relation":"dataset","target_id":"reported-dataset-71614d99b3099f"}],"attributes":{"origin":"author_reported","protocol":"Tertiary-structure-conditioned RNA sequence design; external Rfam assessment","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":"external","population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b4-007","kind":"evaluation","name":"CUPID Data-aug-Avg: non-coding RNA pairwise interaction prediction","description":"","status":"needs_review","facets":{"areas":["rna-transcriptomes"],"tasks":["non-coding RNA pairwise interaction prediction"]},"source_ids":["cupid-rna-interactions-2026"],"links":[{"relation":"model","target_id":"reported-model-2894d253c5e8a8"},{"relation":"benchmark","target_id":"reported-task-c7ce06b753b8b6"},{"relation":"dataset","target_id":"reported-dataset-32ccef507a1dd7"}],"attributes":{"origin":"author_reported","protocol":"Data augmentation with average pooling for molecule-level ncRNA embeddings","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b4-008","kind":"evaluation","name":"ProteinBERT LLM-encoding model: mRNA-protein interaction prediction","description":"","status":"needs_review","facets":{"areas":["rna-transcriptomes"],"tasks":["mRNA-protein interaction prediction"]},"source_ids":["mrna-protein-diversity-2026"],"links":[{"relation":"model","target_id":"reported-model-41ae49bb40ed8e"},{"relation":"benchmark","target_id":"reported-task-d7e6274011946e"},{"relation":"dataset","target_id":"reported-dataset-2e87449871ca47"}],"attributes":{"origin":"author_reported","protocol":"LLM encoding of protein partner; RBP-aware partition tests generalization to unseen protein diversity","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":"RBP-aware test set","population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b4-009","kind":"evaluation","name":"ESM2 650M: human-versus-viral protein classification","description":"","status":"needs_review","facets":{"areas":["proteins-complexes"],"tasks":["human-versus-viral protein classification"]},"source_ids":["viral-immune-mimicry-2025"],"links":[{"relation":"model","target_id":"reported-model-4c73500c39e9d0"},{"relation":"benchmark","target_id":"reported-task-53506fe386e4a1"},{"relation":"dataset","target_id":"reported-dataset-43f24c4dfb7351"}],"attributes":{"origin":"author_reported","protocol":"ESM2 650M embedding-based human-virus classifier","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b4-010","kind":"evaluation","name":"ProtT5 embeddings + ensemble classifier: protein-protein binding-site prediction","description":"","status":"needs_review","facets":{"areas":["proteins-complexes"],"tasks":["protein-protein binding-site prediction"]},"source_ids":["protein-binding-sites-2023"],"links":[{"relation":"model","target_id":"reported-model-fdac4c1ec8a433"},{"relation":"benchmark","target_id":"reported-task-f0ed5188dbb6d4"},{"relation":"dataset","target_id":"reported-dataset-f08b1a60aebeeb"}],"attributes":{"origin":"author_reported","protocol":"Explainable ensemble binding-site predictor using ProtT5 features","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b4-011","kind":"evaluation","name":"CLAPE-SMB with ESM-2: protein-small molecule binding-site prediction","description":"","status":"needs_review","facets":{"areas":["proteins-complexes"],"tasks":["protein-small molecule binding-site prediction"]},"source_ids":["clape-smb-2024"],"links":[{"relation":"model","target_id":"reported-model-57dbab30462150"},{"relation":"benchmark","target_id":"reported-task-b181ed450cdd41"},{"relation":"dataset","target_id":"reported-dataset-701d910b02d25c"}],"attributes":{"origin":"author_reported","protocol":"Contrastive CLAPE-SMB binding-site predictor with ESM-2 feature extractor","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b4-012","kind":"evaluation","name":"Vaxign-DL + ESM: vaccine-antigen candidate prediction","description":"","status":"needs_review","facets":{"areas":["proteins-complexes"],"tasks":["vaccine-antigen candidate prediction"]},"source_ids":["vaxign-esm-2024"],"links":[{"relation":"model","target_id":"reported-model-8ad3e0cefde796"},{"relation":"benchmark","target_id":"reported-task-47465954d606e6"},{"relation":"dataset","target_id":"reported-dataset-b91c871eb7740a"}],"attributes":{"origin":"author_reported","protocol":"Combined skip architecture, four layers, ESM-generated sequence features","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b4-013","kind":"evaluation","name":"scGPT + residual geometry: gene-regulatory signal prediction","description":"","status":"needs_review","facets":{"areas":["cells-tissues"],"tasks":["gene-regulatory signal prediction"]},"source_ids":["single-cell-residual-geometry-2026"],"links":[{"relation":"model","target_id":"reported-model-d6a7fa854437e8"},{"relation":"benchmark","target_id":"reported-task-99afd88cb12895"},{"relation":"dataset","target_id":"reported-dataset-d9fdd8dc7a0184"}],"attributes":{"origin":"author_reported","protocol":"Asymmetric extraction, PCA-64 centered cosine geometry added to scGPT baseline","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b4-014","kind":"evaluation","name":"GREmLN: cell-type annotation","description":"","status":"needs_review","facets":{"areas":["cells-tissues"],"tasks":["cell-type annotation"]},"source_ids":["gremln-2026"],"links":[{"relation":"model","target_id":"reported-model-53d6515bcc1f39"},{"relation":"benchmark","target_id":"reported-task-6312c8a7ac045e"},{"relation":"dataset","target_id":"reported-dataset-59def895fbdbb4"}],"attributes":{"origin":"author_reported","protocol":"Zero-shot cell-type annotation using pre-trained cellular graph foundation model","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":"zero-shot","population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b4-015","kind":"evaluation","name":"Cell-DINO ViT-L: protein localization classification","description":"","status":"needs_review","facets":{"areas":["cells-tissues"],"tasks":["protein localization classification"]},"source_ids":["cell-dino-2025"],"links":[{"relation":"model","target_id":"reported-model-808b23c65fbc89"},{"relation":"benchmark","target_id":"reported-task-7621fa1be55362"},{"relation":"dataset","target_id":"reported-dataset-d356eac961cb69"}],"attributes":{"origin":"author_reported","protocol":"Self-supervised microscopy embedding pre-trained on HPA-FoV; downstream protein-localization classifier. Dataset-specific pretraining; the paper does not claim a general-purpose foundation model that generalizes beyond these benchmarks.","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b4-016","kind":"evaluation","name":"scGen: differentially expressed gene identification","description":"","status":"needs_review","facets":{"areas":["cells-tissues"],"tasks":["differentially expressed gene identification"]},"source_ids":["insilico-perturbation-auprc-2025"],"links":[{"relation":"model","target_id":"reported-model-fcf2cd29a81aae"},{"relation":"benchmark","target_id":"reported-task-003d746a129c9b"},{"relation":"dataset","target_id":"reported-dataset-7bf2cf7d2b2d01"}],"attributes":{"origin":"independent_paper","protocol":"In-silico perturbation assessment with precision sampled at fixed 50% recall","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":"CD14+Mono","population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b4-017","kind":"evaluation","name":"TCINet + HTRS: pathogen detection","description":"","status":"needs_review","facets":{"areas":["microbes-communities"],"tasks":["pathogen detection"]},"source_ids":["metagenomic-pathogens-2025"],"links":[{"relation":"model","target_id":"reported-model-4ce8cae0f2eafc"},{"relation":"benchmark","target_id":"reported-task-d3fd502fdc2b38"},{"relation":"dataset","target_id":"reported-dataset-1c4c71078ffe01"}],"attributes":{"origin":"author_reported","protocol":"Taxonomy-constrained inference network with hierarchical taxonomy representation","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b4-018","kind":"evaluation","name":"DETIRE: viral sequence detection","description":"","status":"needs_review","facets":{"areas":["microbes-communities"],"tasks":["viral sequence detection"]},"source_ids":["detire-viral-metagenomes-2023"],"links":[{"relation":"model","target_id":"reported-model-6d9dbac97852d8"},{"relation":"benchmark","target_id":"reported-task-3d4dec23120fef"},{"relation":"dataset","target_id":"reported-dataset-0b54f42a987b1d"}],"attributes":{"origin":"author_reported","protocol":"Hybrid deep learning virus-fragment classifier on paper testing dataset","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":"test","population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b4-019","kind":"evaluation","name":"PC-mer + LR: metagenomic genus classification","description":"","status":"needs_review","facets":{"areas":["microbes-communities"],"tasks":["metagenomic genus classification"]},"source_ids":["pc-mer-2024"],"links":[{"relation":"model","target_id":"reported-model-688eb780ef7d2e"},{"relation":"benchmark","target_id":"reported-task-4420dcdfe8338d"},{"relation":"dataset","target_id":"reported-dataset-d28955d5872903"}],"attributes":{"origin":"author_reported","protocol":"k=8 PC-mer feature extraction with logistic regression on AMP genus-classification dataset","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":"genus-level","population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b4-020","kind":"evaluation","name":"MDL4Microbiome: microbiome disease-state classification","description":"","status":"needs_review","facets":{"areas":["microbes-communities"],"tasks":["microbiome disease-state classification"]},"source_ids":["mdl4microbiome-2022"],"links":[{"relation":"model","target_id":"reported-model-e6ba198c2ac996"},{"relation":"benchmark","target_id":"reported-task-e2009c35eabd69"},{"relation":"dataset","target_id":"reported-dataset-bd3f98e2eeb5d3"}],"attributes":{"origin":"author_reported","protocol":"Multimodal deep learning model on colorectal-cancer versus healthy microbiome samples","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b4-021","kind":"evaluation","name":"binding-affinity meta-model: protein-ligand binding affinity prediction","description":"","status":"needs_review","facets":{"areas":["molecular-interactions"],"tasks":["protein-ligand binding affinity prediction"]},"source_ids":["ligand-affinity-meta-model-2024"],"links":[{"relation":"model","target_id":"reported-model-df0efcc0346224"},{"relation":"benchmark","target_id":"reported-task-77a32496ce8fe6"},{"relation":"dataset","target_id":"reported-dataset-17132fbabd7683"}],"attributes":{"origin":"author_reported","protocol":"Sequence-or-structure meta-model; predicts ln(Kd/Ki) using docked and deep-learning components","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":"core benchmark","population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b4-022","kind":"evaluation","name":"DeepInterAware: antigen-antibody HIV neutralization prediction","description":"","status":"needs_review","facets":{"areas":["molecular-interactions"],"tasks":["antigen-antibody HIV neutralization prediction"]},"source_ids":["deepinteraware-2025"],"links":[{"relation":"model","target_id":"reported-model-8d2c291733dfe1"},{"relation":"benchmark","target_id":"reported-task-45ead9a1eddf8d"},{"relation":"dataset","target_id":"reported-dataset-50f0bdb7cf9ca4"}],"attributes":{"origin":"author_reported","protocol":"Sequence-based interface-aware model, antibody-unseen split","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":"antibody-unseen","population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b4-023","kind":"evaluation","name":"TransBind: transcription-factor DNA binding-site prediction","description":"","status":"needs_review","facets":{"areas":["molecular-interactions"],"tasks":["transcription-factor DNA binding-site prediction"]},"source_ids":["transbind-2026"],"links":[{"relation":"model","target_id":"reported-model-81b0394d5ac3e8"},{"relation":"benchmark","target_id":"reported-task-ac191e878dff5e"},{"relation":"dataset","target_id":"reported-dataset-034c60a2dabc73"}],"attributes":{"origin":"author_reported","protocol":"Integrates protein and DNA embeddings for TFBS prediction on paper test dataset","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":"test","population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract"}}}
{"id":"evaluation-lit-b4-024","kind":"evaluation","name":"ESM2_AMPS: protein-protein interaction prediction","description":"","status":"needs_review","facets":{"areas":["molecular-interactions"],"tasks":["protein-protein interaction prediction"]},"source_ids":["esm2-amp-2025"],"links":[{"relation":"model","target_id":"reported-model-67ea6bd77b2ed1"},{"relation":"benchmark","target_id":"reported-task-09c3100b77dcc5"},{"relation":"dataset","target_id":"reported-dataset-9135087a16af1c"}],"attributes":{"origin":"author_reported","protocol":"ESM2-derived embeddings plus paper interaction predictor","version":null,"comparison":{"protocol_id":null,"dataset_version":null,"split":null,"population":null,"inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"evidence-alphafold-docs-input","kind":"source","name":"AlphaFold 3 input specification","description":"Primary source inspected for the AlphaFold 3 profile; source checking is not experimental reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/google-deepmind/alphafold3/blob/c0f97eda2f1f482fd94d3a38bece18c7069b4a5c/docs/input.md","version":"c0f97eda2f1f482fd94d3a38bece18c7069b4a5c","retrieved_at":"2026-09-16T19:56:24.856478+00:00","artifact_url":"https://raw.githubusercontent.com/google-deepmind/alphafold3/c0f97eda2f1f482fd94d3a38bece18c7069b4a5c/docs/input.md","artifact_sha256":"e75407473d8c6975a91bd40605b7c8075008426a15a1a952826996e3387226c1","source_locator":"docs/input.md","extraction_method":"primary_artifact_review"}}
{"id":"evidence-alphafold-docs-installation","kind":"source","name":"AlphaFold 3 installation documentation","description":"Primary source inspected for the AlphaFold 3 profile; source checking is not experimental reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/google-deepmind/alphafold3/blob/c0f97eda2f1f482fd94d3a38bece18c7069b4a5c/docs/installation.md","version":"c0f97eda2f1f482fd94d3a38bece18c7069b4a5c","retrieved_at":"2026-09-16T19:56:24.856478+00:00","artifact_url":"https://raw.githubusercontent.com/google-deepmind/alphafold3/c0f97eda2f1f482fd94d3a38bece18c7069b4a5c/docs/installation.md","artifact_sha256":"4e91ac50393d9579316ca13cf438d516a17cea2f73c2ef1a0a1066b37c797c34","source_locator":"docs/installation.md","extraction_method":"primary_artifact_review"}}
{"id":"evidence-alphafold-docs-output","kind":"source","name":"AlphaFold 3 output specification","description":"Primary source inspected for the AlphaFold 3 profile; source checking is not experimental reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/google-deepmind/alphafold3/blob/c0f97eda2f1f482fd94d3a38bece18c7069b4a5c/docs/output.md","version":"c0f97eda2f1f482fd94d3a38bece18c7069b4a5c","retrieved_at":"2026-09-16T19:56:24.856478+00:00","artifact_url":"https://raw.githubusercontent.com/google-deepmind/alphafold3/c0f97eda2f1f482fd94d3a38bece18c7069b4a5c/docs/output.md","artifact_sha256":"bb792ca2564e5a1f48512388efc4ffda00535f8847211e095f11ce15111362fe","source_locator":"docs/output.md","extraction_method":"primary_artifact_review"}}
{"id":"evidence-alphafold-docs-performance","kind":"source","name":"AlphaFold 3 performance documentation","description":"Primary source inspected for the AlphaFold 3 profile; source checking is not experimental reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/google-deepmind/alphafold3/blob/c0f97eda2f1f482fd94d3a38bece18c7069b4a5c/docs/performance.md","version":"c0f97eda2f1f482fd94d3a38bece18c7069b4a5c","retrieved_at":"2026-09-16T19:56:24.856478+00:00","artifact_url":"https://raw.githubusercontent.com/google-deepmind/alphafold3/c0f97eda2f1f482fd94d3a38bece18c7069b4a5c/docs/performance.md","artifact_sha256":"5fb82c4be3c91f6d488196a57be03dab6a3685d629c87c72d719bc868fc9e38a","source_locator":"docs/performance.md","extraction_method":"primary_artifact_review"}}
{"id":"evidence-alphafold-license","kind":"source","name":"AlphaFold 3 code licence","description":"Primary source inspected for the AlphaFold 3 profile; source checking is not experimental reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/google-deepmind/alphafold3/blob/c0f97eda2f1f482fd94d3a38bece18c7069b4a5c/LICENSE","version":"c0f97eda2f1f482fd94d3a38bece18c7069b4a5c","retrieved_at":"2026-09-16T19:56:24.856478+00:00","artifact_url":"https://raw.githubusercontent.com/google-deepmind/alphafold3/c0f97eda2f1f482fd94d3a38bece18c7069b4a5c/LICENSE","artifact_sha256":"cfc7749b96f63bd31c3c42b5c471bf756814053e847c10f3eb003417bc523d30","source_locator":"LICENSE","extraction_method":"primary_artifact_review"}}
{"id":"evidence-alphafold-paper","kind":"source","name":"AlphaFold 3 paper","description":"Primary source inspected for the AlphaFold 3 profile; source checking is not experimental reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11168924/","version":"Nature 2024; DOI 10.1038/s41586-024-07487-w; XML retrieved 2026-09-16","retrieved_at":"2026-09-16T19:56:24.856478+00:00","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11168924/fullTextXML","artifact_sha256":"e627083f74d990b275e96a7503eee3e3eae9e3035b8efb28321277fcac6f1d6e","source_locator":"Main text, Fig. 1; Model limitations; Methods","extraction_method":"primary_artifact_review"}}
{"id":"evidence-alphafold-readme","kind":"source","name":"AlphaFold 3 README","description":"Primary source inspected for the AlphaFold 3 profile; source checking is not experimental reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/google-deepmind/alphafold3/blob/c0f97eda2f1f482fd94d3a38bece18c7069b4a5c/README.md","version":"c0f97eda2f1f482fd94d3a38bece18c7069b4a5c","retrieved_at":"2026-09-16T19:56:24.856478+00:00","artifact_url":"https://raw.githubusercontent.com/google-deepmind/alphafold3/c0f97eda2f1f482fd94d3a38bece18c7069b4a5c/README.md","artifact_sha256":"cc16ba436ea8a967c764ac034a655d632b4c1ef8b4065868b8d07914c7cb88be","source_locator":"README.md","extraction_method":"primary_artifact_review"}}
{"id":"evidence-alphafold-server-faq","kind":"source","name":"AlphaFold Server FAQ","description":"Primary source inspected for the AlphaFold 3 profile; source checking is not experimental reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://alphafoldserver.com/faq","version":"Rendered public FAQ, 2026-09-16","retrieved_at":"2026-09-16T19:56:24.856478+00:00","artifact_url":"https://alphafoldserver.com/faq","artifact_sha256":"66df2a77d57bbf01afc72d6ac110bc195d85a259d01d3f532c2e6e576e652eb0","source_locator":"Named FAQ questions; browser-rendered body text","extraction_method":"rendered_primary_page_review"}}
{"id":"evidence-alphafold-server-output-terms","kind":"source","name":"AlphaFold Server output terms","description":"Primary source inspected for the AlphaFold 3 profile; source checking is not experimental reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://alphafoldserver.com/output-terms","version":"Last modified 2024-05-08; rendered 2026-09-16","retrieved_at":"2026-09-16T19:56:24.856478+00:00","artifact_url":"https://alphafoldserver.com/output-terms","artifact_sha256":"9832af4e56499e0f373584b95327dfc805b834e90186f85ee2387f6e1509c126","source_locator":"Use restrictions; Miscellaneous","extraction_method":"rendered_primary_page_review"}}
{"id":"evidence-alphafold-server-terms","kind":"source","name":"AlphaFold Server terms","description":"Primary source inspected for the AlphaFold 3 profile; source checking is not experimental reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://alphafoldserver.com/terms","version":"Last modified 2024-05-08; rendered 2026-09-16","retrieved_at":"2026-09-16T19:56:24.856478+00:00","artifact_url":"https://alphafoldserver.com/terms","artifact_sha256":"5861ed425310fcda7e2a7807d20fbfd4e6bf6dc00801d082cd8b4f2507dacfae","source_locator":"Key things to know; Overview","extraction_method":"rendered_primary_page_review"}}
{"id":"evidence-alphafold-supplement","kind":"source","name":"AlphaFold 3 supplementary information","description":"Original supplementary PDF inspected for architecture and training-data provenance; table layout visually checked.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://www.nature.com/articles/s41586-024-07487-w#Sec23","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11168924/supplementaryFiles","archive_member":"41586_2024_7487_MOESM1_ESM.pdf","artifact_sha256":"9f19a51ea050e6f77b6c07e2c06aa871d40050fd07da22ae09e92d0887a83130","archive_sha256":"b175fcf15fecf03fbf2ec39e47bd59a504af601080b35c32f3cdee3507ff4e88","retrieved_at":"2026-09-16T20:31:11.456977+00:00","version":"Supplement distributed with DOI 10.1038/s41586-024-07487-w, retrieved 2026-09-16","locator":"Sections 2.2, 2.5, 3 and 5.2; Tables 3 and 6","review_method":"automated_source_review","hash_scope":"SHA-256 of original PDF bytes extracted from the Europe PMC supplement archive"}}
{"id":"evidence-alphafold-weights-terms-of-use","kind":"source","name":"AlphaFold 3 weights terms","description":"Primary source inspected for the AlphaFold 3 profile; source checking is not experimental reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/google-deepmind/alphafold3/blob/c0f97eda2f1f482fd94d3a38bece18c7069b4a5c/WEIGHTS_TERMS_OF_USE.md","version":"c0f97eda2f1f482fd94d3a38bece18c7069b4a5c","retrieved_at":"2026-09-16T19:56:24.856478+00:00","artifact_url":"https://raw.githubusercontent.com/google-deepmind/alphafold3/c0f97eda2f1f482fd94d3a38bece18c7069b4a5c/WEIGHTS_TERMS_OF_USE.md","artifact_sha256":"41adf62ff5eabc58831c828793988537948663c139f8b87d8d413851b150b6e5","source_locator":"WEIGHTS_TERMS_OF_USE.md","extraction_method":"primary_artifact_review"}}
{"id":"evidence-benchmark-cafa-20260916","kind":"source","name":"CAFA official description — reviewed snapshot 2026-09-16","description":"Official CAFA page snapshot inspected in this completion pass; the cached bytes differ from the prior source snapshot.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"d7a5fa76bea98551f8322aab9da965c383104c79d83992e723c0c5836b2fdcd1","licence":null,"locator":"HTML main page","missing_metadata":{"immutable_source_version":"unavailable","licence":"unextracted","version":"unreported"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T19:47:27.547276+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","scope_note":"Mutable official page: content hash and retrieval time recorded, but the original page is not redistributed.","url":"https://biofunctionprediction.org/cafa/","version":"2026-09-16 website snapshot sha256:d7a5fa76bea98551f8322aab9da965c383104c79d83992e723c0c5836b2fdcd1","artifact_url":"https://biofunctionprediction.org/cafa/"}}
{"id":"evidence-benchmark-cami-snapshot","kind":"source","name":"cami official source","description":"Primary project documentation or project-maintained evidence.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"17825bf33280f40b596a104c547b57fae5ee5c07f8d60b396d0f4780d47ef9a5","licence":null,"locator":"HTML main page","missing_metadata":{"immutable_source_version":"unavailable","licence":"unextracted","version":"unreported"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:31:54.339676+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","scope_note":"Mutable official page: content hash and retrieval time recorded, but the original page is not redistributed.","url":"https://cami-challenge.org/","version":"Retrieved website snapshot sha256:17825bf33280f40b596a104c547b57fae5ee5c07f8d60b396d0f4780d47ef9a5","artifact_url":"https://cami-challenge.org/"}}
{"id":"evidence-benchmark-capri-snapshot","kind":"source","name":"capri official source","description":"Primary project documentation or project-maintained evidence.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"5e22606895cf0c30565ed4456bc680ca97c10c8c5d38054f82a6e4677c8d8c24","licence":null,"locator":"HTML main page","missing_metadata":{"immutable_source_version":"unavailable","licence":"unextracted","version":"unreported"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:31:54.812767+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","scope_note":"Mutable official page: content hash and retrieval time recorded, but the original page is not redistributed.","url":"https://www.capri-docking.org/","version":"Retrieved website snapshot sha256:5e22606895cf0c30565ed4456bc680ca97c10c8c5d38054f82a6e4677c8d8c24","artifact_url":"https://www.capri-docking.org/"}}
{"id":"evidence-benchmark-casp-snapshot","kind":"source","name":"casp official source","description":"Primary project documentation or project-maintained evidence.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"de391b68636462ddb78d0659d8784128909d23e9c015832cca94d883f404a3f7","licence":null,"locator":"HTML main page","missing_metadata":{"immutable_source_version":"unavailable","licence":"unextracted","version":"unreported"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:31:54.981498+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","scope_note":"Mutable official page: content hash and retrieval time recorded, but the original page is not redistributed.","url":"https://predictioncenter.org/","version":"Retrieved website snapshot sha256:de391b68636462ddb78d0659d8784128909d23e9c015832cca94d883f404a3f7","artifact_url":"https://predictioncenter.org/"}}
{"id":"evidence-benchmark-flip2-snapshot","kind":"source","name":"flip2 official source","description":"Primary project documentation or project-maintained evidence.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"cd991ae6e76a5e84ea5449f91c4ed86ba4f57942682dc3c50d872c366bbbd4b7","licence":null,"locator":"HTML main page","missing_metadata":{"immutable_source_version":"unavailable","licence":"unextracted","version":"unreported"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:31:54.381289+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","scope_note":"Mutable official page: content hash and retrieval time recorded, but the original page is not redistributed.","url":"https://flip.protein.properties/","version":"Retrieved website snapshot sha256:cd991ae6e76a5e84ea5449f91c4ed86ba4f57942682dc3c50d872c366bbbd4b7","artifact_url":"https://flip.protein.properties/"}}
{"id":"evidence-benchmark-glycan-classification-dataset","kind":"source","name":"GlycanML/GlycanML module/custom_datasets/glycan_classification.py","description":"Pinned computational evaluator or dataset/configuration source inspected for benchmark metadata.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://raw.githubusercontent.com/GlycanML/GlycanML/9f392aa6f9c6d74a296a250199beb347923d04e0/module/custom_datasets/glycan_classification.py","artifact_url":"https://raw.githubusercontent.com/GlycanML/GlycanML/9f392aa6f9c6d74a296a250199beb347923d04e0/module/custom_datasets/glycan_classification.py","artifact_sha256":"1a385c4afcbfc3d2f1f319cf03fc0f4be5a1e9848cd51f9df4239cbd39771530","version":"9f392aa6f9c6d74a296a250199beb347923d04e0","retrieved_at":"2026-09-16T20:43:24.344407+00:00"}}
{"id":"evidence-benchmark-glycan-immunogenicity-dataset","kind":"source","name":"GlycanML/GlycanML module/custom_datasets/glycan_immunogenicity.py","description":"Pinned computational evaluator or dataset/configuration source inspected for benchmark metadata.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://raw.githubusercontent.com/GlycanML/GlycanML/9f392aa6f9c6d74a296a250199beb347923d04e0/module/custom_datasets/glycan_immunogenicity.py","artifact_url":"https://raw.githubusercontent.com/GlycanML/GlycanML/9f392aa6f9c6d74a296a250199beb347923d04e0/module/custom_datasets/glycan_immunogenicity.py","artifact_sha256":"73bb768deff2b22b4cf57199353267dc0778b42ccbbaee8035e91ec67b3a3a1f","version":"9f392aa6f9c6d74a296a250199beb347923d04e0","retrieved_at":"2026-09-16T20:43:24.817696+00:00"}}
{"id":"evidence-benchmark-glycan-interaction-dataset","kind":"source","name":"GlycanML/GlycanML module/custom_datasets/glycan_interaction.py","description":"Pinned computational evaluator or dataset/configuration source inspected for benchmark metadata.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://raw.githubusercontent.com/GlycanML/GlycanML/9f392aa6f9c6d74a296a250199beb347923d04e0/module/custom_datasets/glycan_interaction.py","artifact_url":"https://raw.githubusercontent.com/GlycanML/GlycanML/9f392aa6f9c6d74a296a250199beb347923d04e0/module/custom_datasets/glycan_interaction.py","artifact_sha256":"e7b1fcae29941ffef3cc6cefc5df373a05fce4d6cf9ffe46cde594503fb2b7f5","version":"9f392aa6f9c6d74a296a250199beb347923d04e0","retrieved_at":"2026-09-16T20:43:24.579180+00:00"}}
{"id":"evidence-benchmark-glycan-link-dataset","kind":"source","name":"GlycanML/GlycanML module/custom_datasets/glycan_link.py","description":"Pinned computational evaluator or dataset/configuration source inspected for benchmark metadata.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://raw.githubusercontent.com/GlycanML/GlycanML/9f392aa6f9c6d74a296a250199beb347923d04e0/module/custom_datasets/glycan_link.py","artifact_url":"https://raw.githubusercontent.com/GlycanML/GlycanML/9f392aa6f9c6d74a296a250199beb347923d04e0/module/custom_datasets/glycan_link.py","artifact_sha256":"eab9f95d0007d0c4cc903530aa6677c7d8f73734d7826f6d73ba1b0f5a9319d8","version":"9f392aa6f9c6d74a296a250199beb347923d04e0","retrieved_at":"2026-09-16T20:43:24.095380+00:00"}}
{"id":"evidence-benchmark-glycanml-bert-immunogenicity-bert-yaml","kind":"source","name":"GlycanML/GlycanML configs/single_task/BERT/immunogenicity_BERT.yaml","description":"Pinned computational evaluator or dataset/configuration source inspected for benchmark metadata.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://raw.githubusercontent.com/GlycanML/GlycanML/9f392aa6f9c6d74a296a250199beb347923d04e0/configs/single_task/BERT/immunogenicity_BERT.yaml","artifact_url":"https://raw.githubusercontent.com/GlycanML/GlycanML/9f392aa6f9c6d74a296a250199beb347923d04e0/configs/single_task/BERT/immunogenicity_BERT.yaml","artifact_sha256":"5c92f79c530629642450ca4536cd864cccc3a626f8ecdfeaaf6fcd2ece356901","version":"9f392aa6f9c6d74a296a250199beb347923d04e0","retrieved_at":"2026-09-16T20:42:46.930612+00:00"}}
{"id":"evidence-benchmark-glycanml-bert-interaction-bert-yaml","kind":"source","name":"GlycanML/GlycanML configs/single_task/BERT/interaction_BERT.yaml","description":"Pinned computational evaluator or dataset/configuration source inspected for benchmark metadata.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://raw.githubusercontent.com/GlycanML/GlycanML/9f392aa6f9c6d74a296a250199beb347923d04e0/configs/single_task/BERT/interaction_BERT.yaml","artifact_url":"https://raw.githubusercontent.com/GlycanML/GlycanML/9f392aa6f9c6d74a296a250199beb347923d04e0/configs/single_task/BERT/interaction_BERT.yaml","artifact_sha256":"7cb8f9e774ab0f9d66536b2db47bb236214275ee08d4de95540e4e51d0129639","version":"9f392aa6f9c6d74a296a250199beb347923d04e0","retrieved_at":"2026-09-16T20:42:47.417430+00:00"}}
{"id":"evidence-benchmark-glycanml-bert-link-bert-yaml","kind":"source","name":"GlycanML/GlycanML configs/single_task/BERT/link_BERT.yaml","description":"Pinned computational evaluator or dataset/configuration source inspected for benchmark metadata.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://raw.githubusercontent.com/GlycanML/GlycanML/9f392aa6f9c6d74a296a250199beb347923d04e0/configs/single_task/BERT/link_BERT.yaml","artifact_url":"https://raw.githubusercontent.com/GlycanML/GlycanML/9f392aa6f9c6d74a296a250199beb347923d04e0/configs/single_task/BERT/link_BERT.yaml","artifact_sha256":"82f43f0d45085132ecd6e6dff1b95bf7eb20ad5650713fc8ad0a42d0bbba5a94","version":"9f392aa6f9c6d74a296a250199beb347923d04e0","retrieved_at":"2026-09-16T20:42:47.161322+00:00"}}
{"id":"evidence-benchmark-glycanml-bert-species-bert-yaml","kind":"source","name":"GlycanML/GlycanML configs/single_task/BERT/species_BERT.yaml","description":"Pinned computational evaluator or dataset/configuration source inspected for benchmark metadata.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://raw.githubusercontent.com/GlycanML/GlycanML/9f392aa6f9c6d74a296a250199beb347923d04e0/configs/single_task/BERT/species_BERT.yaml","artifact_url":"https://raw.githubusercontent.com/GlycanML/GlycanML/9f392aa6f9c6d74a296a250199beb347923d04e0/configs/single_task/BERT/species_BERT.yaml","artifact_sha256":"474995bb150c76479b0c99b6f0b2de49f5b6ad626eae1751e2f53774a4fbf38e","version":"9f392aa6f9c6d74a296a250199beb347923d04e0","retrieved_at":"2026-09-16T20:42:46.675170+00:00"}}
{"id":"evidence-benchmark-massspecgym-de-novo-base-py","kind":"source","name":"pluskal-lab/MassSpecGym massspecgym/models/de_novo/base.py","description":"Pinned computational evaluator or dataset/configuration source inspected for benchmark metadata.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://raw.githubusercontent.com/pluskal-lab/MassSpecGym/f259fe3780d5bd227fc6ece36ce6f397c2eef716/massspecgym/models/de_novo/base.py","artifact_url":"https://raw.githubusercontent.com/pluskal-lab/MassSpecGym/f259fe3780d5bd227fc6ece36ce6f397c2eef716/massspecgym/models/de_novo/base.py","artifact_sha256":"5d2d61d7ac3c6503b3869d5cec687e7e82d438900df3d01c6f7cf4c340e5a004","version":"f259fe3780d5bd227fc6ece36ce6f397c2eef716","retrieved_at":"2026-09-16T20:42:47.661606+00:00"}}
{"id":"evidence-benchmark-massspecgym-retrieval-base-py","kind":"source","name":"pluskal-lab/MassSpecGym massspecgym/models/retrieval/base.py","description":"Pinned computational evaluator or dataset/configuration source inspected for benchmark metadata.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://raw.githubusercontent.com/pluskal-lab/MassSpecGym/f259fe3780d5bd227fc6ece36ce6f397c2eef716/massspecgym/models/retrieval/base.py","artifact_url":"https://raw.githubusercontent.com/pluskal-lab/MassSpecGym/f259fe3780d5bd227fc6ece36ce6f397c2eef716/massspecgym/models/retrieval/base.py","artifact_sha256":"2ae08f58bd6db0e00430b54c078b45ae91b9252aac02ee358fef9d8ba1b3f81f","version":"f259fe3780d5bd227fc6ece36ce6f397c2eef716","retrieved_at":"2026-09-16T20:42:47.889112+00:00"}}
{"id":"evidence-benchmark-massspecgym-simulation-base-py","kind":"source","name":"pluskal-lab/MassSpecGym massspecgym/models/simulation/base.py","description":"Pinned computational evaluator or dataset/configuration source inspected for benchmark metadata.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://raw.githubusercontent.com/pluskal-lab/MassSpecGym/f259fe3780d5bd227fc6ece36ce6f397c2eef716/massspecgym/models/simulation/base.py","artifact_url":"https://raw.githubusercontent.com/pluskal-lab/MassSpecGym/f259fe3780d5bd227fc6ece36ce6f397c2eef716/massspecgym/models/simulation/base.py","artifact_sha256":"5208cc352c144e24ea8249124483856aeb1e9821c0cea142d6a907ac248b6295","version":"f259fe3780d5bd227fc6ece36ce6f397c2eef716","retrieved_at":"2026-09-16T20:42:48.138101+00:00"}}
{"id":"evidence-benchmark-mfass-pinned-readme","kind":"source","name":"MFASS computational protocol README at bee9133b83f3aedaf2bbb9013f1875515845607e","description":"","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/timini/rewire-benchmarks/blob/bee9133b83f3aedaf2bbb9013f1875515845607e/benchmarks/mfass/README.md","version":"bee9133b83f3aedaf2bbb9013f1875515845607e","retrieved_at":"2026-09-16T20:55:04.446365+00:00","artifact_sha256":"62a9381484fd2360767e78b114d9aa6cd8a4fc6487a981881a4c4999f0ab1b23","artifact_url":"https://raw.githubusercontent.com/timini/rewire-benchmarks/bee9133b83f3aedaf2bbb9013f1875515845607e/benchmarks/mfass/README.md"}}
{"id":"evidence-benchmark-proteinbench-snapshot","kind":"source","name":"proteinbench official source","description":"Primary project documentation or project-maintained evidence.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"2e488850a6557bb57407615f2df9194351718b3dc0298a03c0c97d8e93460712","licence":null,"locator":"HTML main page","missing_metadata":{"immutable_source_version":"unavailable","licence":"unextracted","version":"unreported"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:31:54.349409+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","scope_note":"Mutable official page: content hash and retrieval time recorded, but the original page is not redistributed.","url":"https://proteinbench.github.io/","version":"Retrieved website snapshot sha256:2e488850a6557bb57407615f2df9194351718b3dc0298a03c0c97d8e93460712","artifact_url":"https://proteinbench.github.io/"}}
{"id":"evidence-benchmark-structure-informed-html","kind":"source","name":"Structure-Informed Protein Language Models are Robust Predictors for Variant Effects (reviewed HTML snapshot)","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"6686d2646e7b1b1203a7ef6d48bacf966b4cd6b0895fd2855551937c6bb87111","artifact_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12068927/","retrieved_at":"2026-09-16T19:58:11.209175+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12068927/","version":"Human Genetics 2025 journal article (online 2024)"}}
{"id":"evidence-benchmark-vcc2026-snapshot","kind":"source","name":"vcc2026 official source","description":"Primary project documentation or project-maintained evidence.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"57dd9cbd5e6b62ed665eebd909d82ffec580a55c3bd2f48a928f807a8c983b1e","licence":null,"locator":"HTML main page","missing_metadata":{"immutable_source_version":"unavailable","licence":"unextracted","version":"unreported"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:31:54.531595+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","scope_note":"Mutable official page: content hash and retrieval time recorded, but the original page is not redistributed.","url":"https://arcinstitute.org/news/virtual-cell-challenge-2026","version":"Retrieved website snapshot sha256:57dd9cbd5e6b62ed665eebd909d82ffec580a55c3bd2f48a928f807a8c983b1e","artifact_url":"https://arcinstitute.org/news/virtual-cell-challenge-2026"}}
{"id":"evidence-discovery-final-amber","kind":"source","name":"amber primary benchmark evidence","description":"Primary paper or official task implementation inspected for the named methodology claims.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"2a75501cbe44396ec0d103d896026d19d823336d131e1fa2aada8555f495a7d0","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC6022608/fullTextXML","locator":"Full primary artifact; individual claims specify their section.","retrieved_at":"2026-09-16T21:04:55.738801+00:00","review_method":"automated_source_review","review_scope":"Benchmark methodology metadata only; no independent reproduction or new numerical result extraction.","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC6022608/fullTextXML","version":"PMC6022608"}}
{"id":"evidence-discovery-final-atom3d","kind":"source","name":"atom3d primary benchmark evidence","description":"Primary paper or official task implementation inspected for the named methodology claims.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"92656c20a15311c32bed9edc7f465bb26eb30338f40324fe43bec4b1fc6a7890","artifact_url":"https://arxiv.org/pdf/2012.04035v4","locator":"Full primary artifact; individual claims specify their section.","retrieved_at":"2026-09-16T21:05:49.738972+00:00","review_method":"automated_source_review","review_scope":"Benchmark methodology metadata only; no independent reproduction or new numerical result extraction.","url":"https://arxiv.org/abs/2012.04035v4","version":"arXiv:2012.04035v4 (15 January 2022)"}}
{"id":"evidence-discovery-final-beacon","kind":"source","name":"beacon primary benchmark evidence","description":"Primary paper or official task implementation inspected for the named methodology claims.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"c2496bff164b87e635ba253ea5a5edc55c94c4673e81c071bc8f0db1e288d006","artifact_url":"https://arxiv.org/pdf/2406.10391v1","locator":"Full primary artifact; individual claims specify their section.","retrieved_at":"2026-09-16T21:04:55.172530+00:00","review_method":"automated_source_review","review_scope":"Benchmark methodology metadata only; no independent reproduction or new numerical result extraction.","url":"https://arxiv.org/pdf/2406.10391v1","version":"2406.10391v1"}}
{"id":"evidence-discovery-final-beeline","kind":"source","name":"beeline primary benchmark evidence","description":"Primary paper or official task implementation inspected for the named methodology claims.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"99b59a6941878779b34ab8eed9ea27c8d8967c24fb67b92fce418e7f512d3a11","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC7098173/fullTextXML","locator":"Full primary artifact; individual claims specify their section.","retrieved_at":"2026-09-16T21:04:55.226418+00:00","review_method":"automated_source_review","review_scope":"Benchmark methodology metadata only; no independent reproduction or new numerical result extraction.","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC7098173/fullTextXML","version":"PMC7098173"}}
{"id":"evidence-discovery-final-bend","kind":"source","name":"bend primary benchmark evidence","description":"Primary paper or official task implementation inspected for the named methodology claims.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"6ff9f6dc19fc831200e241e3279a06c44d0fcd0aac566da81058e9db7fb70334","artifact_url":"https://arxiv.org/pdf/2311.12570v1","locator":"Full primary artifact; individual claims specify their section.","retrieved_at":"2026-09-16T21:04:56.127991+00:00","review_method":"automated_source_review","review_scope":"Benchmark methodology metadata only; no independent reproduction or new numerical result extraction.","url":"https://arxiv.org/pdf/2311.12570v1","version":"2311.12570v1"}}
{"id":"evidence-discovery-final-cafa3","kind":"source","name":"cafa3 primary benchmark evidence","description":"Primary paper or official task implementation inspected for the named methodology claims.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"3802f37548d77addac829127bf03322cbe6c69eba944d88a775b534311804fa4","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC6864930/fullTextXML","locator":"Full primary artifact; individual claims specify their section.","retrieved_at":"2026-09-16T21:05:50.876204+00:00","review_method":"automated_source_review","review_scope":"Benchmark methodology metadata only; no independent reproduction or new numerical result extraction.","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC6864930/fullTextXML","version":"PMC6864930"}}
{"id":"evidence-discovery-final-cami2","kind":"source","name":"cami2 primary benchmark evidence","description":"Primary paper or official task implementation inspected for the named methodology claims.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"da932c1cde8b290e1694fc3cf44d98ed527be1ec7cf41542dd5e9aee75c38704","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC9007738/fullTextXML","locator":"Full primary artifact; individual claims specify their section.","retrieved_at":"2026-09-16T21:04:55.691966+00:00","review_method":"automated_source_review","review_scope":"Benchmark methodology metadata only; no independent reproduction or new numerical result extraction.","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC9007738/fullTextXML","version":"PMC9007738"}}
{"id":"evidence-discovery-final-capri","kind":"source","name":"capri primary benchmark evidence","description":"Primary paper or official task implementation inspected for the named methodology claims.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"194e32f06903caee54706cb61570cb1d5bc6b47911a8a427d9244789ce5a0165","artifact_url":"https://www.capri-docking.org/assessment/","locator":"Full primary artifact; individual claims specify their section.","retrieved_at":"2026-09-16T21:11:45.067183+00:00","review_method":"automated_source_review","review_scope":"Benchmark methodology metadata only; no independent reproduction or new numerical result extraction.","url":"https://www.capri-docking.org/assessment/","version":"Retrieved 2026-09-16 assessment page"}}
{"id":"evidence-discovery-final-casp16","kind":"source","name":"casp16 primary benchmark evidence","description":"Primary paper or official task implementation inspected for the named methodology claims.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"32ef3bdfac658a0191e2939699b68e23dcaf465568c5d1c9f75743e886bd76cc","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12750037/fullTextXML","locator":"Full primary artifact; individual claims specify their section.","retrieved_at":"2026-09-16T21:11:45.148220+00:00","review_method":"automated_source_review","review_scope":"Benchmark methodology metadata only; no independent reproduction or new numerical result extraction.","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12750037/fullTextXML","version":"PMC12750037"}}
{"id":"evidence-discovery-final-dart","kind":"source","name":"dart primary benchmark evidence","description":"Primary paper or official task implementation inspected for the named methodology claims.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"4194b137ba55c9a2c269d119a9afec6ae1bb0feaf17d91433ae483c41221a56b","artifact_url":"https://arxiv.org/html/2412.05430v1","locator":"Full primary artifact; individual claims specify their section.","retrieved_at":"2026-09-16T21:06:29.715960+00:00","review_method":"automated_source_review","review_scope":"Benchmark methodology metadata only; no independent reproduction or new numerical result extraction.","url":"https://arxiv.org/html/2412.05430v1","version":"2412.05430v1"}}
{"id":"evidence-discovery-final-dockq","kind":"source","name":"dockq primary benchmark evidence","description":"Primary paper or official task implementation inspected for the named methodology claims.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"013ceed0e054f971a81caf7cdcac10c3f4241ecec0532e326a1fca124e69dbe9","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC4999177/fullTextXML","locator":"Full primary artifact; individual claims specify their section.","retrieved_at":"2026-09-16T21:11:45.113639+00:00","review_method":"automated_source_review","review_scope":"Benchmark methodology metadata only; no independent reproduction or new numerical result extraction.","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC4999177/fullTextXML","version":"PMC4999177"}}
{"id":"evidence-discovery-final-flip","kind":"source","name":"flip primary benchmark evidence","description":"Primary paper or official task implementation inspected for the named methodology claims.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"afcf360c88a7a4ae153b3c2d8d4fa6d4ac0abe84f2131b94409abff6447ce363","artifact_url":"https://flip.protein.properties/assets/FLIP_2021_manuscript.pdf","locator":"Full primary artifact; individual claims specify their section.","retrieved_at":"2026-09-16T21:07:13.889041+00:00","review_method":"automated_source_review","review_scope":"Benchmark methodology metadata only; no independent reproduction or new numerical result extraction.","url":"https://flip.protein.properties/assets/FLIP_2021_manuscript.pdf","version":"2021 manuscript"}}
{"id":"evidence-discovery-final-flip2","kind":"source","name":"flip2 primary benchmark evidence","description":"Primary paper or official task implementation inspected for the named methodology claims.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"d0e61ca27863c023adddcc225a3e4af1718f2e3fdfe3642bef89c24c93621a17","artifact_url":"https://flip.protein.properties/assets/FLIP_manuscipt.pdf","locator":"Full primary artifact; individual claims specify their section.","retrieved_at":"2026-09-16T21:05:54.225877+00:00","review_method":"automated_source_review","review_scope":"Benchmark methodology metadata only; no independent reproduction or new numerical result extraction.","url":"https://flip.protein.properties/assets/FLIP_manuscipt.pdf","version":"ICML2026 manuscript"}}
{"id":"evidence-discovery-final-geneb","kind":"source","name":"geneb primary benchmark evidence","description":"Primary paper or official task implementation inspected for the named methodology claims.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"47975089c0ca738d5e2d6e6ea91cd4e7b80498c3f175803d8ec39ca77aa41629","artifact_url":"https://arxiv.org/html/2606.04525v1","locator":"Full primary artifact; individual claims specify their section.","retrieved_at":"2026-09-16T21:06:29.746844+00:00","review_method":"automated_source_review","review_scope":"Benchmark methodology metadata only; no independent reproduction or new numerical result extraction.","url":"https://arxiv.org/html/2606.04525v1","version":"2606.04525v1"}}
{"id":"evidence-discovery-final-genomic-benchmarks","kind":"source","name":"genomic-benchmarks primary benchmark evidence","description":"Primary paper or official task implementation inspected for the named methodology claims.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"bda6fe51e3363a5d2fc8d265ca536897d3e83eb76458fc95066c21933e3bd0c0","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC10150520/fullTextXML","locator":"Full primary artifact; individual claims specify their section.","retrieved_at":"2026-09-16T21:07:13.681402+00:00","review_method":"automated_source_review","review_scope":"Benchmark methodology metadata only; no independent reproduction or new numerical result extraction.","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC10150520/fullTextXML","version":"PMC10150520"}}
{"id":"evidence-discovery-final-glycanml","kind":"source","name":"glycanml primary benchmark evidence","description":"Primary paper or official task implementation inspected for the named methodology claims.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"9ba3db678b4550898a612b42e8832bf9dee40935090fc4d969ba8f6ac5106979","artifact_url":"https://arxiv.org/pdf/2405.16206v1","locator":"Full primary artifact; individual claims specify their section.","retrieved_at":"2026-09-16T21:04:56.017006+00:00","review_method":"automated_source_review","review_scope":"Benchmark methodology metadata only; no independent reproduction or new numerical result extraction.","url":"https://arxiv.org/pdf/2405.16206v1","version":"2405.16206v1"}}
{"id":"evidence-discovery-final-gue","kind":"source","name":"gue primary benchmark evidence","description":"Primary paper or official task implementation inspected for the named methodology claims.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"49300acee3e4afd44bebc3de9893c3bc310d331bd4805374e0952fdfbf366f06","artifact_url":"https://arxiv.org/pdf/2306.15006","locator":"Full primary artifact; individual claims specify their section.","retrieved_at":"2026-09-16T20:00:00+00:00","review_method":"automated_source_review","review_scope":"Benchmark methodology metadata only; no independent reproduction or new numerical result extraction.","url":"https://arxiv.org/pdf/2306.15006","version":"2306.15006 retrieved PDF"}}
{"id":"evidence-discovery-final-hest","kind":"source","name":"hest primary benchmark evidence","description":"Primary paper or official task implementation inspected for the named methodology claims.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"636099a73dee8337f60e6e9120230b914605b35553872ddf76e4661bbe14be9b","artifact_url":"https://arxiv.org/pdf/2406.16192v1","locator":"Full primary artifact; individual claims specify their section.","retrieved_at":"2026-09-16T21:05:00.540640+00:00","review_method":"automated_source_review","review_scope":"Benchmark methodology metadata only; no independent reproduction or new numerical result extraction.","url":"https://arxiv.org/pdf/2406.16192v1","version":"2406.16192v1"}}
{"id":"evidence-discovery-final-massspecgym","kind":"source","name":"massspecgym primary benchmark evidence","description":"Primary paper or official task implementation inspected for the named methodology claims.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"82176d50e8947c8b9baa2a0d91493f5680c0e4c7e25a2266ff7879f48a58c40c","artifact_url":"https://arxiv.org/pdf/2410.23326v1","locator":"Full primary artifact; individual claims specify their section.","retrieved_at":"2026-09-16T21:04:58.775040+00:00","review_method":"automated_source_review","review_scope":"Benchmark methodology metadata only; no independent reproduction or new numerical result extraction.","url":"https://arxiv.org/pdf/2410.23326v1","version":"2410.23326v1"}}
{"id":"evidence-discovery-final-mrnabench","kind":"source","name":"mrnabench primary benchmark evidence","description":"Primary paper or official task implementation inspected for the named methodology claims.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"79f6264ee883535203c63a313547e7c57baa85585f76b42f8d899eb17fb7e600","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12265608/fullTextXML","locator":"Full primary artifact; individual claims specify their section.","retrieved_at":"2026-09-16T10:41:16.497221+00:00","review_method":"automated_source_review","review_scope":"Benchmark methodology metadata only; no independent reproduction or new numerical result extraction.","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12265608/fullTextXML","version":"preprint archived 2025-07-08"}}
{"id":"evidence-discovery-final-nabench","kind":"source","name":"nabench primary benchmark evidence","description":"Primary paper or official task implementation inspected for the named methodology claims.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"fefd48d53b1a7eadf9e14db96adacc8e646304c1b592562d9f136f4508941350","artifact_url":"https://arxiv.org/html/2511.02888v1","locator":"Full primary artifact; individual claims specify their section.","retrieved_at":"2026-09-16T21:06:29.579095+00:00","review_method":"automated_source_review","review_scope":"Benchmark methodology metadata only; no independent reproduction or new numerical result extraction.","url":"https://arxiv.org/html/2511.02888v1","version":"2511.02888v1"}}
{"id":"evidence-discovery-final-opal","kind":"source","name":"opal primary benchmark evidence","description":"Primary paper or official task implementation inspected for the named methodology claims.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"a4503ef32e99127d55d0af8ebf0c39baa298b320e4032fcacdb9aa5d33821a09","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC6398228/fullTextXML","locator":"Full primary artifact; individual claims specify their section.","retrieved_at":"2026-09-16T21:04:56.586502+00:00","review_method":"automated_source_review","review_scope":"Benchmark methodology metadata only; no independent reproduction or new numerical result extraction.","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC6398228/fullTextXML","version":"PMC6398228"}}
{"id":"evidence-discovery-final-openproblems-label","kind":"source","name":"openproblems-label primary benchmark evidence","description":"Primary paper or official task implementation inspected for the named methodology claims.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"e223ab712ff55997a3abe659f280d4ea2952700e767b87e02c434701da9833c1","artifact_url":"https://www.openproblems.bio/benchmarks/label_projection/v1.0.0/","locator":"Full primary artifact; individual claims specify their section.","retrieved_at":"2026-09-16T21:16:30.026457+00:00","review_method":"automated_source_review","review_scope":"Benchmark methodology metadata only; no independent reproduction or new numerical result extraction.","url":"https://www.openproblems.bio/benchmarks/label_projection/v1.0.0/","version":"v1.0.0"}}
{"id":"evidence-discovery-final-perturbench","kind":"source","name":"perturbench primary benchmark evidence","description":"Primary paper or official task implementation inspected for the named methodology claims.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"5c4804565dd9faa17a11853a79e9847dcb6da73715b4c91f62e5b274cc79f186","artifact_url":"https://arxiv.org/html/2408.10609v1","locator":"Full primary artifact; individual claims specify their section.","retrieved_at":"2026-09-16T21:07:13.159027+00:00","review_method":"automated_source_review","review_scope":"Benchmark methodology metadata only; no independent reproduction or new numerical result extraction.","url":"https://arxiv.org/html/2408.10609v1","version":"2408.10609v1"}}
{"id":"evidence-discovery-final-petab","kind":"source","name":"petab primary benchmark evidence","description":"Primary paper or official task implementation inspected for the named methodology claims.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"9a9f73410331cb6d0ee148f6872f191393cbaa334f3437ed7fd0fe289476e7a7","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC6735869/fullTextXML","locator":"Full primary artifact; individual claims specify their section.","retrieved_at":"2026-09-16T21:04:56.573248+00:00","review_method":"automated_source_review","review_scope":"Benchmark methodology metadata only; no independent reproduction or new numerical result extraction.","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC6735869/fullTextXML","version":"PMC6735869"}}
{"id":"evidence-discovery-final-pfmbench","kind":"source","name":"pfmbench primary benchmark evidence","description":"Primary paper or official task implementation inspected for the named methodology claims.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"59c7bbb888e8e91f33c1e2cabfde062c32381d6ca727d23b6655c977aabf97a2","artifact_url":"https://arxiv.org/html/2506.14796v1","locator":"Full primary artifact; individual claims specify their section.","retrieved_at":"2026-09-16T21:06:29.716004+00:00","review_method":"automated_source_review","review_scope":"Benchmark methodology metadata only; no independent reproduction or new numerical result extraction.","url":"https://arxiv.org/html/2506.14796v1","version":"2506.14796v1"}}
{"id":"evidence-discovery-final-plinder-config0","kind":"source","name":"plinder-config0 primary benchmark evidence","description":"Primary paper or official task implementation inspected for the named methodology claims.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"23ad37a97187942840416e240d7ee99d0a677a4964a75beed3e8be0082b9375d","artifact_url":"https://raw.githubusercontent.com/plinder-org/plinder/85b3f1cb1763530a6cfd934f4263a1777c41afa4/docs/evaluation.md","locator":"Full primary artifact; individual claims specify their section.","retrieved_at":"2026-09-16T21:11:45.036654+00:00","review_method":"automated_source_review","review_scope":"Benchmark methodology metadata only; no independent reproduction or new numerical result extraction.","url":"https://raw.githubusercontent.com/plinder-org/plinder/85b3f1cb1763530a6cfd934f4263a1777c41afa4/docs/evaluation.md","version":"85b3f1cb1763530a6cfd934f4263a1777c41afa4:docs/evaluation.md"}}
{"id":"evidence-discovery-final-plinder-readme","kind":"source","name":"plinder-readme primary benchmark evidence","description":"Primary paper or official task implementation inspected for the named methodology claims.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"1d53c3b89030dc4651d3e7bf4749256b7e579330fc7a660cffaa992d646da34a","artifact_url":"https://github.com/plinder-org/plinder/blob/85b3f1cb1763530a6cfd934f4263a1777c41afa4/README.md","locator":"Full primary artifact; individual claims specify their section.","retrieved_at":"2026-09-16T10:30:23.659162+00:00","review_method":"automated_source_review","review_scope":"Benchmark methodology metadata only; no independent reproduction or new numerical result extraction.","url":"https://github.com/plinder-org/plinder/blob/85b3f1cb1763530a6cfd934f4263a1777c41afa4/README.md","version":"85b3f1cb1763530a6cfd934f4263a1777c41afa4"}}
{"id":"evidence-discovery-final-posebusters","kind":"source","name":"posebusters primary benchmark evidence","description":"Primary paper or official task implementation inspected for the named methodology claims.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"8bbc6eadc59d33f7433b610d91c0b0d2d3094cacb3c0bdfa83bf89239580c763","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC10901501/fullTextXML","locator":"Full primary artifact; individual claims specify their section.","retrieved_at":"2026-09-16T21:05:51.089997+00:00","review_method":"automated_source_review","review_scope":"Benchmark methodology metadata only; no independent reproduction or new numerical result extraction.","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC10901501/fullTextXML","version":"PMC10901501"}}
{"id":"evidence-discovery-final-proteinbench","kind":"source","name":"proteinbench primary benchmark evidence","description":"Primary paper or official task implementation inspected for the named methodology claims.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"4334d636223ad42bfb9ae68aae03f5a255c29ba1cebe7b8f9588fb3b9b5453b2","artifact_url":"https://arxiv.org/html/2409.06744v1","locator":"Full primary artifact; individual claims specify their section.","retrieved_at":"2026-09-16T21:07:13.231727+00:00","review_method":"automated_source_review","review_scope":"Benchmark methodology metadata only; no independent reproduction or new numerical result extraction.","url":"https://arxiv.org/html/2409.06744v1","version":"2409.06744v1"}}
{"id":"evidence-discovery-final-proteingym","kind":"source","name":"proteingym primary benchmark evidence","description":"Primary paper or official task implementation inspected for the named methodology claims.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"4519641f13271bdd09b166e7d93232f22542489bc52a25b5a1628c3df8badce1","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC10723403/fullTextXML","locator":"Full primary artifact; individual claims specify their section.","retrieved_at":"2026-09-16T10:41:16.517323+00:00","review_method":"automated_source_review","review_scope":"Benchmark methodology metadata only; no independent reproduction or new numerical result extraction.","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC10723403/fullTextXML","version":"PMC10723403.1"}}
{"id":"evidence-discovery-final-scib","kind":"source","name":"scib primary benchmark evidence","description":"Primary paper or official task implementation inspected for the named methodology claims.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"f65dd8b63336ff1a5045dad3cb9c5905ec3bb8b494a221f67dfbffb9a6f612db","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC8748196/fullTextXML","locator":"Full primary artifact; individual claims specify their section.","retrieved_at":"2026-09-16T21:04:56.581902+00:00","review_method":"automated_source_review","review_scope":"Benchmark methodology metadata only; no independent reproduction or new numerical result extraction.","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC8748196/fullTextXML","version":"PMC8748196"}}
{"id":"evidence-discovery-final-scperteval0","kind":"source","name":"scperteval0 primary benchmark evidence","description":"Primary paper or official task implementation inspected for the named methodology claims.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"c4dbfdbc539ba78350ddba68ca4c03edfe89cf3625c6c699884bd3a85e47d929","artifact_url":"https://raw.githubusercontent.com/Virtual-Cell-Research-Community/scPertEval/4685f11927e887745737600170da7a655b727553/src/scperteval/protocols/metrics.py","locator":"Full primary artifact; individual claims specify their section.","retrieved_at":"2026-09-16T21:11:45.124583+00:00","review_method":"automated_source_review","review_scope":"Benchmark methodology metadata only; no independent reproduction or new numerical result extraction.","url":"https://raw.githubusercontent.com/Virtual-Cell-Research-Community/scPertEval/4685f11927e887745737600170da7a655b727553/src/scperteval/protocols/metrics.py","version":"4685f11927e887745737600170da7a655b727553:src/scperteval/protocols/metrics.py"}}
{"id":"evidence-discovery-final-tape","kind":"source","name":"tape primary benchmark evidence","description":"Primary paper or official task implementation inspected for the named methodology claims.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"6ee0c3e6e870635cba8fa67e0a4abc5598c0ab2a10ba127a46b67ac450ae0168","artifact_url":"https://arxiv.org/pdf/1906.08230v1","locator":"Full primary artifact; individual claims specify their section.","retrieved_at":"2026-09-16T20:23:48.238777+00:00","review_method":"automated_source_review","review_scope":"Benchmark methodology metadata only; no independent reproduction or new numerical result extraction.","url":"https://arxiv.org/pdf/1906.08230v1","version":"1906.08230v1"}}
{"id":"evidence-discovery-final-tdc","kind":"source","name":"tdc primary benchmark evidence","description":"Primary paper or official task implementation inspected for the named methodology claims.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"aaa5526f6f100093bc06c9921a8564ad08cfd5332270d614d7681051c2dd66bc","artifact_url":"https://arxiv.org/pdf/2102.09548v1","locator":"Full primary artifact; individual claims specify their section.","retrieved_at":"2026-09-16T21:04:55.709744+00:00","review_method":"automated_source_review","review_scope":"Benchmark methodology metadata only; no independent reproduction or new numerical result extraction.","url":"https://arxiv.org/pdf/2102.09548v1","version":"2102.09548v1"}}
{"id":"evidence-discovery-final-vcc-guide","kind":"source","name":"vcc-guide primary benchmark evidence","description":"Primary paper or official task implementation inspected for the named methodology claims.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"a2a085fd104ecf6b97e125ebc5714c642e8140db4f549ff84abd3bce1d2a2c80","artifact_url":"https://vcc-cli-wiki.virtualcellchallenge.org/","locator":"Full primary artifact; individual claims specify their section.","retrieved_at":"2026-09-16T21:16:29.975126+00:00","review_method":"automated_source_review","review_scope":"Benchmark methodology metadata only; no independent reproduction or new numerical result extraction.","url":"https://vcc-cli-wiki.virtualcellchallenge.org/","version":"Retrieved 2026-09-16"}}
{"id":"evidence-final-model-geneformer-tree","kind":"source","name":"Geneformer official repository file inventory","description":"Primary-source follow-up resolving earlier evidence retrieval gaps.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://huggingface.co/ctheodoris/Geneformer/tree/1f7fbae4e469a5f4f1af8c111a529cfe1b3829f5","artifact_url":"https://huggingface.co/ctheodoris/Geneformer/tree/1f7fbae4e469a5f4f1af8c111a529cfe1b3829f5","version":"1f7fbae4e469a5f4f1af8c111a529cfe1b3829f5","artifact_sha256":"d1e85b25927c06241c2240f8272223838ab279f5072ae0a270a34efc8185d7dc","artifact_format":"extracted_file_path_inventory","retrieved_at":"2026-09-16T21:10:17.420034+00:00"}}
{"id":"evidence-final-model-rna-fm-paper","kind":"source","name":"RNA-FM original methods, arXiv v5","description":"Primary-source follow-up resolving earlier evidence retrieval gaps.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://arxiv.org/pdf/2204.00300v5","artifact_url":"https://arxiv.org/pdf/2204.00300v5","version":"2204.00300v5","artifact_sha256":"b3945c283c5ca6e8a346cfeb3e8d6609dac6e4a3874fd38dc2436c35eeb6074a","artifact_format":"original_pdf","retrieved_at":"2026-09-16T21:10:17.419803+00:00"}}
{"id":"evidence-mfass-final-geo","kind":"source","name":"MFASS original assay accession GSE120695","description":"Primary artifact inspected for profile metadata; historical numerical results are not changed.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://ftp.ncbi.nlm.nih.gov/geo/series/GSE120nnn/GSE120695/soft/GSE120695_family.soft.gz","artifact_url":"https://ftp.ncbi.nlm.nih.gov/geo/series/GSE120nnn/GSE120695/soft/GSE120695_family.soft.gz","artifact_sha256":"a85331f962b962747ba39c8d0518b0f14d90ca8db3d8396f8fe2f2ffc3e48fa3","version":"GSE120695 SOFT retrieved 2026-09-16","retrieved_at":"2026-09-16T21:02:58.608027+00:00"}}
{"id":"evidence-mfass-final-v1-comparison","kind":"source","name":"Archived MFASS v1 SpliceAI comparison","description":"Primary artifact inspected for profile metadata; historical numerical results are not changed.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://raw.githubusercontent.com/rewire-bio/rewire-benchmarks/bee9133b83f3aedaf2bbb9013f1875515845607e/benchmarks/mfass/results/compare-baseline-vs-spliceai.json","artifact_url":"https://raw.githubusercontent.com/rewire-bio/rewire-benchmarks/bee9133b83f3aedaf2bbb9013f1875515845607e/benchmarks/mfass/results/compare-baseline-vs-spliceai.json","artifact_sha256":"4856394ecb59c7163afe5d3ef50d5ab9c2cec5b20ed3b4f3b39ef8dfc02b5778","version":"bee9133b83f3aedaf2bbb9013f1875515845607e","retrieved_at":"2026-09-16T21:02:58.608188+00:00"}}
{"id":"evidence-official-00ee97a5586392ae2348","kind":"source","name":"soedinglab/hh-suite: LICENSE","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/soedinglab/hh-suite/blob/43095e46ada4ec2a8a47d47ef5ad7e38b1429f7b/LICENSE","artifact_url":"https://raw.githubusercontent.com/soedinglab/hh-suite/43095e46ada4ec2a8a47d47ef5ad7e38b1429f7b/LICENSE","version":"43095e46ada4ec2a8a47d47ef5ad7e38b1429f7b","retrieved_at":"2026-09-16T19:46:20.402052+00:00","artifact_sha256":"589ed823e9a84c56feb95ac58e7cf384626b9cbf4fda2a907bc36e103de1bad2","locator":"LICENSE","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-019aff235ae2c3cf29d6","kind":"source","name":"dauparas/ProteinMPNN: protein_mpnn_utils.py","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/dauparas/ProteinMPNN/blob/8907e6671bfbfc92303b5f79c4b5e6ce47cdef57/protein_mpnn_utils.py","artifact_url":"https://raw.githubusercontent.com/dauparas/ProteinMPNN/8907e6671bfbfc92303b5f79c4b5e6ce47cdef57/protein_mpnn_utils.py","version":"8907e6671bfbfc92303b5f79c4b5e6ce47cdef57","retrieved_at":"2026-09-16T19:46:18.948060+00:00","artifact_sha256":"74c8f9b7553422a7a0bbd705874844ee103c8926c2c96f154a87e0b824071e1b","locator":"protein_mpnn_utils.py","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-028db5eeb224def23456","kind":"source","name":"songlab-cal/tape: README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/songlab-cal/tape/blob/6d345c2b2bbf52cd32cf179325c222afd92aec7e/README.md","artifact_url":"https://raw.githubusercontent.com/songlab-cal/tape/6d345c2b2bbf52cd32cf179325c222afd92aec7e/README.md","version":"6d345c2b2bbf52cd32cf179325c222afd92aec7e","retrieved_at":"2026-09-16T19:46:20.570146+00:00","artifact_sha256":"b28c74fe3cd6b69a8ba6d84891d0539e54dfef882abd5ed4d11ed0b029bb477a","locator":"README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-03106594cde7719aa359","kind":"source","name":"nbrg-ppcu/prokbert: LICENSE","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/nbrg-ppcu/prokbert/blob/8670ae92b816cff158a0b85647a8dea122e251eb/LICENSE","artifact_url":"https://raw.githubusercontent.com/nbrg-ppcu/prokbert/8670ae92b816cff158a0b85647a8dea122e251eb/LICENSE","version":"8670ae92b816cff158a0b85647a8dea122e251eb","retrieved_at":"2026-09-16T19:46:19.913949+00:00","artifact_sha256":"66e17f3a57ace034ab422667c02d75134bb531619d9491f3abc7ea42bb8a6643","locator":"LICENSE","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-0437be6150a3da9f8efc","kind":"source","name":"evolutionaryscale/esm: _assets/ESM3_README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/evolutionaryscale/esm/blob/bf343ba264b650dff7a073643725f9aaa1fdbe8d/_assets/ESM3_README.md","artifact_url":"https://raw.githubusercontent.com/evolutionaryscale/esm/bf343ba264b650dff7a073643725f9aaa1fdbe8d/_assets/ESM3_README.md","version":"bf343ba264b650dff7a073643725f9aaa1fdbe8d","retrieved_at":"2026-09-16T19:46:18.983347+00:00","artifact_sha256":"2f564153712e32ad668f1586b6e78599d1a7fd6a3b3b7049718d85f9b33f79a6","locator":"_assets/ESM3_README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-044ad0df4e4fe48a2125","kind":"source","name":"songlab-cal/tape: tape/models/modeling_unirep.py","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/songlab-cal/tape/blob/6d345c2b2bbf52cd32cf179325c222afd92aec7e/tape/models/modeling_unirep.py","artifact_url":"https://raw.githubusercontent.com/songlab-cal/tape/6d345c2b2bbf52cd32cf179325c222afd92aec7e/tape/models/modeling_unirep.py","version":"6d345c2b2bbf52cd32cf179325c222afd92aec7e","retrieved_at":"2026-09-16T19:46:20.570146+00:00","artifact_sha256":"f6a0da726a8bcdd40640fab39c0814a56d6188531b44a3ce6e7e7137755c9a02","locator":"tape/models/modeling_unirep.py","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-04ac37076f906b8470cb","kind":"source","name":"InstaDeepAI/agro-nucleotide-transformer-1b: config.json","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://huggingface.co/InstaDeepAI/agro-nucleotide-transformer-1b/blob/b0e1ea1f53a2bf5bb29f8eab7a7e553bf06c1ab1/config.json","artifact_url":"https://huggingface.co/InstaDeepAI/agro-nucleotide-transformer-1b/blob/b0e1ea1f53a2bf5bb29f8eab7a7e553bf06c1ab1/config.json","version":"b0e1ea1f53a2bf5bb29f8eab7a7e553bf06c1ab1","retrieved_at":"2026-09-16T20:12:20.913125+00:00","artifact_sha256":"4063f8250f32d922611d8b36f0def1bb53b7ae129d6439c50c8b6c340e8eb0bd","locator":"config.json","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-082d60e1af5af7ed12ca","kind":"source","name":"aqlaboratory/genie3: README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/aqlaboratory/genie3/blob/d77ae5ac04212ff1e8b29b585859a3244c614804/README.md","artifact_url":"https://raw.githubusercontent.com/aqlaboratory/genie3/d77ae5ac04212ff1e8b29b585859a3244c614804/README.md","version":"d77ae5ac04212ff1e8b29b585859a3244c614804","retrieved_at":"2026-09-16T19:46:18.306022+00:00","artifact_sha256":"50a634fe3236b6272b70928ac41bdebd4a230467b151f59597742ccd56ac8909","locator":"README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-099d109417c7fae96e96","kind":"source","name":"aertslab/GRNBoost: README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/aertslab/GRNBoost/blob/26c836b3dcbb85852d3c6f4b8340e8655434da02/README.md","artifact_url":"https://raw.githubusercontent.com/aertslab/GRNBoost/26c836b3dcbb85852d3c6f4b8340e8655434da02/README.md","version":"26c836b3dcbb85852d3c6f4b8340e8655434da02","retrieved_at":"2026-09-16T19:46:18.179520+00:00","artifact_sha256":"f2befd99acf59576a22b8a44abd2345e8ed7304cf470609f27d311e08ed3f066","locator":"README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-0b8003691827eb09b0a4","kind":"source","name":"proteinmpnn-supp: Publisher supplementary archive","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC9997061/supplementaryFiles","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC9997061/supplementaryFiles","version":"Retrieved page snapshot; no immutable publisher revision supplied","retrieved_at":"2026-09-16T20:43:11.978593+00:00","artifact_sha256":"002269144962619673b7b187bdda20bd250e5f3cfd9f72a529eac53e7a1ac082","locator":"Publisher supplementary archive","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-0cbab80c2d1769287a1f","kind":"source","name":"scverse/scvi-tools: docs/user_guide/models/scvi.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/scverse/scvi-tools/blob/73b28e44223621470e582a81a102c107bb22678b/docs/user_guide/models/scvi.md","artifact_url":"https://raw.githubusercontent.com/scverse/scvi-tools/73b28e44223621470e582a81a102c107bb22678b/docs/user_guide/models/scvi.md","version":"73b28e44223621470e582a81a102c107bb22678b","retrieved_at":"2026-09-16T19:46:20.137176+00:00","artifact_sha256":"c81a5cfb30db1eca8c03af3d0db1a36a447a1bb5a27b270de5633ba73ebeeeef","locator":"docs/user_guide/models/scvi.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-0e4dac85aa0e0ff1d4b2","kind":"source","name":"bowang-lab/scGPT: README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/bowang-lab/scGPT/blob/cebd6fae655b9c585a4807daa3ac31bb764f06b4/README.md","artifact_url":"https://raw.githubusercontent.com/bowang-lab/scGPT/cebd6fae655b9c585a4807daa3ac31bb764f06b4/README.md","version":"cebd6fae655b9c585a4807daa3ac31bb764f06b4","retrieved_at":"2026-09-16T19:46:18.594429+00:00","artifact_sha256":"b0503e8ca789f19f1fc2350c5aaf57b1b323bbae43b354655231b5f4a1586c83","locator":"README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-0ffe1031786a71400ffd","kind":"source","name":"mimic: Primary paper PDF","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://arxiv.org/pdf/2604.24506","artifact_url":"https://arxiv.org/pdf/2604.24506","version":"2604.24506v1","retrieved_at":"2026-09-16T20:04:04.776851+00:00","artifact_sha256":"0ba8639b263f0e7f736bbd4564a8b1acca1f66acc4300cffd3c26e6bf2e93ffe","locator":"Primary paper PDF","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-1005311cb2f598fe63de","kind":"source","name":"CAMI-challenge/CAMISIM: LICENSE.txt","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/CAMI-challenge/CAMISIM/blob/7ce6013c6d5a0fac8ba8a80e52a03560d3546fa6/LICENSE.txt","artifact_url":"https://raw.githubusercontent.com/CAMI-challenge/CAMISIM/7ce6013c6d5a0fac8ba8a80e52a03560d3546fa6/LICENSE.txt","version":"7ce6013c6d5a0fac8ba8a80e52a03560d3546fa6","retrieved_at":"2026-09-16T19:46:17.767933+00:00","artifact_sha256":"b40930bbcf80744c86c46a12bc9da056641d722716c378f5659b9e555ef833e1","locator":"LICENSE.txt","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-10a6550ca70fdeaf33cb","kind":"source","name":"matsui-lab/GlycanGT: README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/matsui-lab/GlycanGT/blob/96611518c971deb89215ca163deaf9de3a59fa32/README.md","artifact_url":"https://raw.githubusercontent.com/matsui-lab/GlycanGT/96611518c971deb89215ca163deaf9de3a59fa32/README.md","version":"96611518c971deb89215ca163deaf9de3a59fa32","retrieved_at":"2026-09-16T19:46:19.737544+00:00","artifact_sha256":"fa39bffca31211baedb3b63df5dede775fcffff7b411d4261d5b18f526ae1153","locator":"README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-10f58e320879def8b102","kind":"source","name":"biobakery/humann: readme.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/biobakery/humann/blob/e07b3a34d0b94c09a8ac5d28ff95009611178be2/readme.md","artifact_url":"https://raw.githubusercontent.com/biobakery/humann/e07b3a34d0b94c09a8ac5d28ff95009611178be2/readme.md","version":"e07b3a34d0b94c09a8ac5d28ff95009611178be2","retrieved_at":"2026-09-16T19:46:18.580457+00:00","artifact_sha256":"96260519b594ae22cba9f28d1f64622de001f2abf11d406c9da572bfaf145727","locator":"readme.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-116050b81d03a38e3b6f","kind":"source","name":"dreams: Journal full-text XML","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC13090125/","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC13090125/fullTextXML","version":"Retrieved page snapshot; no immutable publisher revision supplied","retrieved_at":"2026-09-16T19:53:03.200000+00:00","artifact_sha256":"4bbe2ecff0ad75944b3aec5131369c1b2f75ff6f4b5dc257487a274fdd7b0fef","locator":"Journal full-text XML","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-116dfd8f61e04b51c148","kind":"source","name":"tbepler/protein-sequence-embedding-iclr2019: README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/tbepler/protein-sequence-embedding-iclr2019/blob/be32cffeec26431bdf87438eb5f07ddd6fc5d7dd/README.md","artifact_url":"https://raw.githubusercontent.com/tbepler/protein-sequence-embedding-iclr2019/be32cffeec26431bdf87438eb5f07ddd6fc5d7dd/README.md","version":"be32cffeec26431bdf87438eb5f07ddd6fc5d7dd","retrieved_at":"2026-09-16T19:57:33.688522+00:00","artifact_sha256":"ecb193485e89d65558a509bae571057862dd079d29afc877c8c9812c4c525415","locator":"README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-12d43bf3076a93fc4ee7","kind":"source","name":"songlab-cal/tape: tape/models/modeling_onehot.py","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/songlab-cal/tape/blob/6d345c2b2bbf52cd32cf179325c222afd92aec7e/tape/models/modeling_onehot.py","artifact_url":"https://raw.githubusercontent.com/songlab-cal/tape/6d345c2b2bbf52cd32cf179325c222afd92aec7e/tape/models/modeling_onehot.py","version":"6d345c2b2bbf52cd32cf179325c222afd92aec7e","retrieved_at":"2026-09-16T19:46:20.570146+00:00","artifact_sha256":"546fab3931d2ae7b6a0ee06ff64c6bcedf2d58b460e5aa90be623b0895befe43","locator":"tape/models/modeling_onehot.py","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-137a3a936ffa4673b3d3","kind":"source","name":"facebook/esm1v_t33_650M_UR90S_1: config.json","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://huggingface.co/facebook/esm1v_t33_650M_UR90S_1/blob/8bfdb1892536cc77bd0760b9c25ddced2cd0b4c8/config.json","artifact_url":"https://huggingface.co/facebook/esm1v_t33_650M_UR90S_1/blob/8bfdb1892536cc77bd0760b9c25ddced2cd0b4c8/config.json","version":"8bfdb1892536cc77bd0760b9c25ddced2cd0b4c8","retrieved_at":"2026-09-16T20:12:20.912868+00:00","artifact_sha256":"6c4b576e2e73ad85fc51a4fa75d4e350532b773e04551307a74643c3fc3c5d59","locator":"config.json","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-18d8f4d5fc3f6616922d","kind":"source","name":"instadeepai/nucleotide-transformer: README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/instadeepai/nucleotide-transformer/blob/2dc37b86e16a6970fbc731751f7719d9f676f7f9/README.md","artifact_url":"https://raw.githubusercontent.com/instadeepai/nucleotide-transformer/2dc37b86e16a6970fbc731751f7719d9f676f7f9/README.md","version":"2dc37b86e16a6970fbc731751f7719d9f676f7f9","retrieved_at":"2026-09-16T19:46:19.364532+00:00","artifact_sha256":"9f51bbb20c4c5c36e77fb03ca1c5c36236e287c48a1ee31f53150545d421ec25","locator":"README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-1a0775cac85be75449ef","kind":"source","name":"biobakery/MetaPhlAn: license.txt","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/biobakery/MetaPhlAn/blob/424f3e6e30618266404353e1083c6405a9f02f48/license.txt","artifact_url":"https://raw.githubusercontent.com/biobakery/MetaPhlAn/424f3e6e30618266404353e1083c6405a9f02f48/license.txt","version":"424f3e6e30618266404353e1083c6405a9f02f48","retrieved_at":"2026-09-16T19:46:18.561509+00:00","artifact_sha256":"ecf18c2928e49997ba1f098f0f0da0958d257d69fffb48875162dbd24f3d7762","locator":"license.txt","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-1aae5efd1b27ae9c761a","kind":"source","name":"songlab-cal/tape: tape/models/modeling_lstm.py","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/songlab-cal/tape/blob/6d345c2b2bbf52cd32cf179325c222afd92aec7e/tape/models/modeling_lstm.py","artifact_url":"https://raw.githubusercontent.com/songlab-cal/tape/6d345c2b2bbf52cd32cf179325c222afd92aec7e/tape/models/modeling_lstm.py","version":"6d345c2b2bbf52cd32cf179325c222afd92aec7e","retrieved_at":"2026-09-16T19:46:20.570146+00:00","artifact_sha256":"e857bc4d6d551915a375116208f7f74b286e7fc88d0d54f54b2746b2fb02e1ca","locator":"tape/models/modeling_lstm.py","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-1b2aa7207e3a05f297ab","kind":"source","name":"ViennaRNA/ViennaRNA: README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/ViennaRNA/ViennaRNA/blob/1ffec79f5e258896160f7362ced8263450f371dc/README.md","artifact_url":"https://raw.githubusercontent.com/ViennaRNA/ViennaRNA/1ffec79f5e258896160f7362ced8263450f371dc/README.md","version":"1ffec79f5e258896160f7362ced8263450f371dc","retrieved_at":"2026-09-16T19:46:18.166113+00:00","artifact_sha256":"d37146b01e4273062a5230c496a9af8414ee7ef14fcd905cf415959256d59c4e","locator":"README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-1d81a48c6dcc0e6dc14d","kind":"source","name":"biohub/ESMC-6B: README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://huggingface.co/biohub/ESMC-6B/blob/af1602ba7406f521b11bf8f81d52af378cde09e4/README.md","artifact_url":"https://huggingface.co/biohub/ESMC-6B/blob/af1602ba7406f521b11bf8f81d52af378cde09e4/README.md","version":"af1602ba7406f521b11bf8f81d52af378cde09e4","retrieved_at":"2026-09-16T20:22:27.767420+00:00","artifact_sha256":"64096baa2f7345c1badd5be2c97c0998fc46d4c3d017c2a3f3ce83418a20f147","locator":"README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-1df5b1865861177c0c75","kind":"source","name":"ArcInstitute/state: README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/ArcInstitute/state/blob/9bbfe78a434a55205e4de834e1ea99f85f7a3add/README.md","artifact_url":"https://raw.githubusercontent.com/ArcInstitute/state/9bbfe78a434a55205e4de834e1ea99f85f7a3add/README.md","version":"9bbfe78a434a55205e4de834e1ea99f85f7a3add","retrieved_at":"2026-09-16T19:46:17.767683+00:00","artifact_sha256":"568c0b4f9d93374f7ebc7467c226fdde0050abfd60147bce0b316163bf4199d6","locator":"README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-1f027f8ecc53decdeec8","kind":"source","name":"evolutionaryscale/esm: THIRD_PARTY_NOTICE.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/evolutionaryscale/esm/blob/bf343ba264b650dff7a073643725f9aaa1fdbe8d/THIRD_PARTY_NOTICE.md","artifact_url":"https://raw.githubusercontent.com/evolutionaryscale/esm/bf343ba264b650dff7a073643725f9aaa1fdbe8d/THIRD_PARTY_NOTICE.md","version":"bf343ba264b650dff7a073643725f9aaa1fdbe8d","retrieved_at":"2026-09-16T19:46:18.983347+00:00","artifact_sha256":"5bff8515ba4e0f53abdc43714c180b79c5b606160497d98de741a369cb9b6a23","locator":"THIRD_PARTY_NOTICE.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-1f756fefa823dd402b0f","kind":"source","name":"kraken2: Journal full-text XML","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC6883579/","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC6883579/fullTextXML","version":"Retrieved page snapshot; no immutable publisher revision supplied","retrieved_at":"2026-09-16T19:53:03.009224+00:00","artifact_sha256":"2fd1d7785b45855deacb566acc79336e292c4c2d4975095b2ff7b1200c406869","locator":"Journal full-text XML","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-23e72f7e93a0a633ab2b","kind":"source","name":"ccsb-scripps/AutoDock-Vina: README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/ccsb-scripps/AutoDock-Vina/blob/3c65c0b3e6c2c1d183f6a175ecb65e3c5ba91645/README.md","artifact_url":"https://raw.githubusercontent.com/ccsb-scripps/AutoDock-Vina/3c65c0b3e6c2c1d183f6a175ecb65e3c5ba91645/README.md","version":"3c65c0b3e6c2c1d183f6a175ecb65e3c5ba91645","retrieved_at":"2026-09-16T19:46:18.718558+00:00","artifact_sha256":"4f1728521ab79de1c33e1cf8605b31037effed5de2a2fbbccba58d7b0a005ae7","locator":"README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-25b222d11900e0e88a51","kind":"source","name":"MAGICS-LAB/DNABERT_2: LICENSE","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/MAGICS-LAB/DNABERT_2/blob/f25bed9ee20db966dff39e5c1571249d04e36404/LICENSE","artifact_url":"https://raw.githubusercontent.com/MAGICS-LAB/DNABERT_2/f25bed9ee20db966dff39e5c1571249d04e36404/LICENSE","version":"f25bed9ee20db966dff39e5c1571249d04e36404","retrieved_at":"2026-09-16T19:46:17.892989+00:00","artifact_sha256":"c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4","locator":"LICENSE","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-28c1720afecf4240f3f7","kind":"source","name":"rfdiffusion-supp: Publisher supplementary archive","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC10468394/supplementaryFiles","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC10468394/supplementaryFiles","version":"Retrieved page snapshot; no immutable publisher revision supplied","retrieved_at":"2026-09-16T20:43:11.978663+00:00","artifact_sha256":"eb2ee3da27da262aa90fd5c4ee371581e429039b2b976c135463fbe4b2dbb199","locator":"Publisher supplementary archive","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-28fc17fbe0c1d99da219","kind":"source","name":"facebook/esm2_t33_650M_UR50D: config.json","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://huggingface.co/facebook/esm2_t33_650M_UR50D/blob/08e4846e537177426273712802403f7ba8261b6c/config.json","artifact_url":"https://huggingface.co/facebook/esm2_t33_650M_UR50D/blob/08e4846e537177426273712802403f7ba8261b6c/config.json","version":"08e4846e537177426273712802403f7ba8261b6c","retrieved_at":"2026-09-16T20:04:02.230771+00:00","artifact_sha256":"539095c22efc52a09d6147074ba4ca119f76a890df5901213b2b55f7d2f96b2b","locator":"config.json","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-290b60bb932abe9929a2","kind":"source","name":"chaidiscovery/chai-lab: LICENSE","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/chaidiscovery/chai-lab/blob/66c38d1fe5c6756a89ff8596b1dea87d305ec06f/LICENSE","artifact_url":"https://raw.githubusercontent.com/chaidiscovery/chai-lab/66c38d1fe5c6756a89ff8596b1dea87d305ec06f/LICENSE","version":"66c38d1fe5c6756a89ff8596b1dea87d305ec06f","retrieved_at":"2026-09-16T19:46:18.824295+00:00","artifact_sha256":"511edf51c5c6f47bae9ae19c59d98666a682e0ccc98e90c2de2a9a897c44c003","locator":"LICENSE","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-29d4b229a3aa426a6dfb","kind":"source","name":"facebookresearch/esm: README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/facebookresearch/esm/blob/2b369911bb5b4b0dda914521b9475cad1656b2ac/README.md","artifact_url":"https://raw.githubusercontent.com/facebookresearch/esm/2b369911bb5b4b0dda914521b9475cad1656b2ac/README.md","version":"2b369911bb5b4b0dda914521b9475cad1656b2ac","retrieved_at":"2026-09-16T19:46:19.090082+00:00","artifact_sha256":"8b273c21a322fc9473d1b68d0dd40c8166ab2f89e4a190aa26ca87251b97cba9","locator":"README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-2bf115f7d072744af00a","kind":"source","name":"BojarLab/glycowork: glycowork/ml/models.py","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/BojarLab/glycowork/blob/3d63f1ec25c850da3cde4d25cb602d50b6b5732b/glycowork/ml/models.py","artifact_url":"https://raw.githubusercontent.com/BojarLab/glycowork/3d63f1ec25c850da3cde4d25cb602d50b6b5732b/glycowork/ml/models.py","version":"3d63f1ec25c850da3cde4d25cb602d50b6b5732b","retrieved_at":"2026-09-16T19:46:17.767851+00:00","artifact_sha256":"40222af745e6ec5795ff03d86ca33f9d40eaf331330cd677d9013fce01e88bfb","locator":"glycowork/ml/models.py","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-2cf0a41fec83ee9c5cc9","kind":"source","name":"zhihan1996/DNABERT-2-117M: config.json","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://huggingface.co/zhihan1996/DNABERT-2-117M/blob/7bce263b15377fc15361f52cfab88f8b586abda0/config.json","artifact_url":"https://huggingface.co/zhihan1996/DNABERT-2-117M/blob/7bce263b15377fc15361f52cfab88f8b586abda0/config.json","version":"7bce263b15377fc15361f52cfab88f8b586abda0","retrieved_at":"2026-09-16T19:46:20.874956+00:00","artifact_sha256":"ba9bdafaff0cc3e30556927474d4a179519a9864012bed2628e9f1bc23c84bfd","locator":"config.json","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-2e7c7649620407f50f6b","kind":"source","name":"facebookresearch/esm: LICENSE","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/facebookresearch/esm/blob/2b369911bb5b4b0dda914521b9475cad1656b2ac/LICENSE","artifact_url":"https://raw.githubusercontent.com/facebookresearch/esm/2b369911bb5b4b0dda914521b9475cad1656b2ac/LICENSE","version":"2b369911bb5b4b0dda914521b9475cad1656b2ac","retrieved_at":"2026-09-16T19:46:19.090082+00:00","artifact_sha256":"da6d3703ed11cbe42bd212c725957c98da23cbff1998c05fa4b3d976d1a58e93","locator":"LICENSE","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-2efbfaba5a1c09f4f7fa","kind":"source","name":"nbrg-ppcu/prokbert: README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/nbrg-ppcu/prokbert/blob/8670ae92b816cff158a0b85647a8dea122e251eb/README.md","artifact_url":"https://raw.githubusercontent.com/nbrg-ppcu/prokbert/8670ae92b816cff158a0b85647a8dea122e251eb/README.md","version":"8670ae92b816cff158a0b85647a8dea122e251eb","retrieved_at":"2026-09-16T19:46:19.913949+00:00","artifact_sha256":"29de39c6ad006ce704ab14240cfd97af93da411fb63ec89ebe499c2646928cfc","locator":"README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-2f4711ecd64b0162e85b","kind":"source","name":"evolutionaryscale/esm: LICENSE.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/evolutionaryscale/esm/blob/bf343ba264b650dff7a073643725f9aaa1fdbe8d/LICENSE.md","artifact_url":"https://raw.githubusercontent.com/evolutionaryscale/esm/bf343ba264b650dff7a073643725f9aaa1fdbe8d/LICENSE.md","version":"bf343ba264b650dff7a073643725f9aaa1fdbe8d","retrieved_at":"2026-09-16T19:46:18.983347+00:00","artifact_sha256":"b63df9ca1dd96b3b21eec226b51b236d0bd152ac20eafc43aad46bf832b48d8a","locator":"LICENSE.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-3076c52b2fa7cb48ba7b","kind":"source","name":"https://zenodo.org/api/records/10997887: page.html","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://zenodo.org/api/records/10997887","artifact_url":"https://zenodo.org/api/records/10997887","version":"Retrieved page snapshot; no immutable publisher revision supplied","retrieved_at":"2026-09-16T20:22:27.767327+00:00","artifact_sha256":"bfd7a787f3a7f1f900ca62b763a7161a486653011a43b50c39911a3b82d46c9e","locator":"page.html","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-33fe0374c8ebbc2c7294","kind":"source","name":"DerrickWood/kraken2: docs/MANUAL.markdown","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/DerrickWood/kraken2/blob/8c190b1b668825935dbf6dee5f969227dc8269bb/docs/MANUAL.markdown","artifact_url":"https://raw.githubusercontent.com/DerrickWood/kraken2/8c190b1b668825935dbf6dee5f969227dc8269bb/docs/MANUAL.markdown","version":"8c190b1b668825935dbf6dee5f969227dc8269bb","retrieved_at":"2026-09-16T19:46:17.767980+00:00","artifact_sha256":"182565cb02f3958b39b8303749e58cd8829c6c6bf4734e39533d86e7980071f7","locator":"docs/MANUAL.markdown","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-34faacf1ad53d6aa6bef","kind":"source","name":"PolymathicAI/MIMIC: LICENSE","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/PolymathicAI/MIMIC/blob/9e652f16491e6c2e3881111e24b285c288554275/LICENSE","artifact_url":"https://raw.githubusercontent.com/PolymathicAI/MIMIC/9e652f16491e6c2e3881111e24b285c288554275/LICENSE","version":"9e652f16491e6c2e3881111e24b285c288554275","retrieved_at":"2026-09-16T20:04:02.231308+00:00","artifact_sha256":"cb4951b3d04c3153a6192951a520d7042c927d1f0574978d503596d6c14e36c9","locator":"LICENSE","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-362b9cef071111dfe343","kind":"source","name":"tape: Primary paper PDF","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://arxiv.org/pdf/1906.08230","artifact_url":"https://arxiv.org/pdf/1906.08230","version":"1906.08230v1","retrieved_at":"2026-09-16T20:23:48.238777+00:00","artifact_sha256":"6ee0c3e6e870635cba8fa67e0a4abc5598c0ab2a10ba127a46b67ac450ae0168","locator":"Primary paper PDF","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-3759a2f34cb49263eb43","kind":"source","name":"ODonnell-Lipidomics/LipidFinder: README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/ODonnell-Lipidomics/LipidFinder/blob/8306ca014c2e6b34ce6ef3ec6cab01fa4c666a09/README.md","artifact_url":"https://raw.githubusercontent.com/ODonnell-Lipidomics/LipidFinder/8306ca014c2e6b34ce6ef3ec6cab01fa4c666a09/README.md","version":"8306ca014c2e6b34ce6ef3ec6cab01fa4c666a09","retrieved_at":"2026-09-16T19:57:35.576108+00:00","artifact_sha256":"0449bf076d0d37a2690f0e66f655a03d5e3577207e166f2c963d93d3afcbcd1a","locator":"README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-37f825d1226e59cbbfbf","kind":"source","name":"metaphlan: Journal full-text XML","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC10635831/","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC10635831/fullTextXML","version":"Retrieved page snapshot; no immutable publisher revision supplied","retrieved_at":"2026-09-16T19:53:03.009254+00:00","artifact_sha256":"547306680d52f70865906db567dee93323ecdec3b2e162608dff5057e057aa4b","locator":"Journal full-text XML","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-39a15ed8072aeea14a22","kind":"source","name":"jwohlwend/boltz: docs/training.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/jwohlwend/boltz/blob/b1ebfc46ecf57f5414e0d1a6f9027bbb122c53bc/docs/training.md","artifact_url":"https://raw.githubusercontent.com/jwohlwend/boltz/b1ebfc46ecf57f5414e0d1a6f9027bbb122c53bc/docs/training.md","version":"b1ebfc46ecf57f5414e0d1a6f9027bbb122c53bc","retrieved_at":"2026-09-16T19:46:19.465940+00:00","artifact_sha256":"4574fa3cac9d086708dce59de68a144dc337722af3242db9adfd6798bb37e21c","locator":"docs/training.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-3aa749d3205c0a768b22","kind":"source","name":"dnabert2: Primary paper PDF","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://arxiv.org/pdf/2306.15006","artifact_url":"https://arxiv.org/pdf/2306.15006","version":"2306.15006v2","retrieved_at":"2026-09-16T20:04:04.777189+00:00","artifact_sha256":"49300acee3e4afd44bebc3de9893c3bc310d331bd4805374e0952fdfbf366f06","locator":"Primary paper PDF","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-3cec54ed6ded9cb0a092","kind":"source","name":"songlab-cal/tape: tape/models/modeling_resnet.py","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/songlab-cal/tape/blob/6d345c2b2bbf52cd32cf179325c222afd92aec7e/tape/models/modeling_resnet.py","artifact_url":"https://raw.githubusercontent.com/songlab-cal/tape/6d345c2b2bbf52cd32cf179325c222afd92aec7e/tape/models/modeling_resnet.py","version":"6d345c2b2bbf52cd32cf179325c222afd92aec7e","retrieved_at":"2026-09-16T19:46:20.570146+00:00","artifact_sha256":"fc97f083f059f5b587d568aa0c32f7cca227e247160a7e22e3a168aef1469ebe","locator":"tape/models/modeling_resnet.py","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-3e1890c3652bc5c150f7","kind":"source","name":"rhofold: Journal full-text XML","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11621015/","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11621015/fullTextXML","version":"Retrieved page snapshot; no immutable publisher revision supplied","retrieved_at":"2026-09-16T19:53:03.009162+00:00","artifact_sha256":"a74ba0e47c4b0cdc4481f10ffd10d323eaabc9906d55315f9b2009898d8ae803","locator":"Journal full-text XML","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-3fde3df73e79e455bd86","kind":"source","name":"ml4bio/RNA-FM: LICENSE","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/ml4bio/RNA-FM/blob/348951516e0963d22bbb33b3c9fc18c89081d38e/LICENSE","artifact_url":"https://raw.githubusercontent.com/ml4bio/RNA-FM/348951516e0963d22bbb33b3c9fc18c89081d38e/LICENSE","version":"348951516e0963d22bbb33b3c9fc18c89081d38e","retrieved_at":"2026-09-16T19:46:19.769679+00:00","artifact_sha256":"b0809e99b532fdf51660f5a3d2a9010ed09d15aef0d131ad80fe80c2291a4fba","locator":"LICENSE","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-40f46dc4b9fcc306773b","kind":"source","name":"DerrickWood/kraken2: README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/DerrickWood/kraken2/blob/8c190b1b668825935dbf6dee5f969227dc8269bb/README.md","artifact_url":"https://raw.githubusercontent.com/DerrickWood/kraken2/8c190b1b668825935dbf6dee5f969227dc8269bb/README.md","version":"8c190b1b668825935dbf6dee5f969227dc8269bb","retrieved_at":"2026-09-16T19:46:17.767980+00:00","artifact_sha256":"2ea33af266b4268a55fd750d0f3265cd3165d61f6be375c5ea3b3ff5c58c7c8c","locator":"README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-4269946a73db94a15289","kind":"source","name":"ViennaRNA/ViennaRNA: COPYING","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/ViennaRNA/ViennaRNA/blob/1ffec79f5e258896160f7362ced8263450f371dc/COPYING","artifact_url":"https://raw.githubusercontent.com/ViennaRNA/ViennaRNA/1ffec79f5e258896160f7362ced8263450f371dc/COPYING","version":"1ffec79f5e258896160f7362ced8263450f371dc","retrieved_at":"2026-09-16T19:46:18.166113+00:00","artifact_sha256":"7776cce4c6155eba82ea6bcb5c108f25b9fb0f604664421bfd6da94930028d8c","locator":"COPYING","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-43484d4de29aacd65ed7","kind":"source","name":"bowang-lab/scGPT: LICENSE","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/bowang-lab/scGPT/blob/cebd6fae655b9c585a4807daa3ac31bb764f06b4/LICENSE","artifact_url":"https://raw.githubusercontent.com/bowang-lab/scGPT/cebd6fae655b9c585a4807daa3ac31bb764f06b4/LICENSE","version":"cebd6fae655b9c585a4807daa3ac31bb764f06b4","retrieved_at":"2026-09-16T19:46:18.594429+00:00","artifact_sha256":"1ceeacbed51e2890187425547bc2efd16c1b7ad45189b7dfb21e83a45a2e9d9e","locator":"LICENSE","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-4565e66fdeb787b99afa","kind":"source","name":"polymathic-ai/MIMIC: config.json","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://huggingface.co/polymathic-ai/MIMIC/blob/72e63a1ece34928422fd46f89b6a6580ace99a97/config.json","artifact_url":"https://huggingface.co/polymathic-ai/MIMIC/blob/72e63a1ece34928422fd46f89b6a6580ace99a97/config.json","version":"72e63a1ece34928422fd46f89b6a6580ace99a97","retrieved_at":"2026-09-16T19:46:20.846683+00:00","artifact_sha256":"687e182e410fbfa235e24c337f51411c6fff583d3135fdbac202b08b0e63a8ef","locator":"config.json","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-45f05eebcf7a03ac99cf","kind":"source","name":"kundajelab/chrombpnet: LICENSE","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/kundajelab/chrombpnet/blob/09938fdb4397ec0006510e5251e48920a505d4de/LICENSE","artifact_url":"https://raw.githubusercontent.com/kundajelab/chrombpnet/09938fdb4397ec0006510e5251e48920a505d4de/LICENSE","version":"09938fdb4397ec0006510e5251e48920a505d4de","retrieved_at":"2026-09-16T19:46:19.514840+00:00","artifact_sha256":"eee7b4d55be619630ce91024410485d25226c387e72315b83712a8dbf89188ef","locator":"LICENSE","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-4603e2d255507e025c80","kind":"source","name":"InstaDeepAI/agro-nucleotide-transformer-1b: README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://huggingface.co/InstaDeepAI/agro-nucleotide-transformer-1b/blob/b0e1ea1f53a2bf5bb29f8eab7a7e553bf06c1ab1/README.md","artifact_url":"https://huggingface.co/InstaDeepAI/agro-nucleotide-transformer-1b/blob/b0e1ea1f53a2bf5bb29f8eab7a7e553bf06c1ab1/README.md","version":"b0e1ea1f53a2bf5bb29f8eab7a7e553bf06c1ab1","retrieved_at":"2026-09-16T20:12:20.913125+00:00","artifact_sha256":"552b83763443130a7a207a749d47026c5cc36be72268cb827afe251c30857807","locator":"README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-4920952f9b3c4b8909a0","kind":"source","name":"instadeepai/nucleotide-transformer: docs/segment_nt.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/instadeepai/nucleotide-transformer/blob/2dc37b86e16a6970fbc731751f7719d9f676f7f9/docs/segment_nt.md","artifact_url":"https://raw.githubusercontent.com/instadeepai/nucleotide-transformer/2dc37b86e16a6970fbc731751f7719d9f676f7f9/docs/segment_nt.md","version":"2dc37b86e16a6970fbc731751f7719d9f676f7f9","retrieved_at":"2026-09-16T19:46:19.364532+00:00","artifact_sha256":"8eec4580ba64ab944fb9b42674be70ffe793135f3503cce8640f5b08f8290f7a","locator":"docs/segment_nt.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-4abb9affb1a9e5438c91","kind":"source","name":"InstaDeepAI/nucleotide-transformer-v2-50m-multi-species: README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://huggingface.co/InstaDeepAI/nucleotide-transformer-v2-50m-multi-species/blob/81b29e5786726d891dbf929404ef20adca5b36f1/README.md","artifact_url":"https://huggingface.co/InstaDeepAI/nucleotide-transformer-v2-50m-multi-species/blob/81b29e5786726d891dbf929404ef20adca5b36f1/README.md","version":"81b29e5786726d891dbf929404ef20adca5b36f1","retrieved_at":"2026-09-16T19:46:20.607357+00:00","artifact_sha256":"e526d7b98f106bc2ca9ba73fa166ff5fd62853812e9757a6692eeedf25e42923","locator":"README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-4bf005200ebfb2586b78","kind":"source","name":"evolutionaryscale/esm: README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/evolutionaryscale/esm/blob/bf343ba264b650dff7a073643725f9aaa1fdbe8d/README.md","artifact_url":"https://raw.githubusercontent.com/evolutionaryscale/esm/bf343ba264b650dff7a073643725f9aaa1fdbe8d/README.md","version":"bf343ba264b650dff7a073643725f9aaa1fdbe8d","retrieved_at":"2026-09-16T19:46:18.983347+00:00","artifact_sha256":"74a897f0e97e3d4256f7cff11424dd264df4daaf8cabd9a6fb74b661421de038","locator":"README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-4e0280294c1215534077","kind":"source","name":"ml4bio/RhoFold: LICENSE","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/ml4bio/RhoFold/blob/6bdfbda720184409eb682ce08c05d258162ddc48/LICENSE","artifact_url":"https://raw.githubusercontent.com/ml4bio/RhoFold/6bdfbda720184409eb682ce08c05d258162ddc48/LICENSE","version":"6bdfbda720184409eb682ce08c05d258162ddc48","retrieved_at":"2026-09-16T19:46:19.891935+00:00","artifact_sha256":"cfc7749b96f63bd31c3c42b5c471bf756814053e847c10f3eb003417bc523d30","locator":"LICENSE","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-53934b67953ba33b2c95","kind":"source","name":"aqlaboratory/openfold: docs/source/original_readme.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/aqlaboratory/openfold/blob/be2ec1841f16c966c65ae0e7599ebbadc725757d/docs/source/original_readme.md","artifact_url":"https://raw.githubusercontent.com/aqlaboratory/openfold/be2ec1841f16c966c65ae0e7599ebbadc725757d/docs/source/original_readme.md","version":"be2ec1841f16c966c65ae0e7599ebbadc725757d","retrieved_at":"2026-09-16T19:46:18.314650+00:00","artifact_sha256":"aa0812608f4b4dca4b4d30b9e0eb6b7d08ca9d1596535419d6255c3ced48f36f","locator":"docs/source/original_readme.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-550f6719c2f22f4d57b3","kind":"source","name":"scverse/scvi-tools: README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/scverse/scvi-tools/blob/73b28e44223621470e582a81a102c107bb22678b/README.md","artifact_url":"https://raw.githubusercontent.com/scverse/scvi-tools/73b28e44223621470e582a81a102c107bb22678b/README.md","version":"73b28e44223621470e582a81a102c107bb22678b","retrieved_at":"2026-09-16T19:46:20.137176+00:00","artifact_sha256":"eb46b8a54e60643ca0cd8cb375ede05b01dcbd2380ca17ce8027a92ba13cebbb","locator":"README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-5631fb75aa8e5a7a89e1","kind":"source","name":"ODonnell-Lipidomics/LipidFinder: LICENSE","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/ODonnell-Lipidomics/LipidFinder/blob/8306ca014c2e6b34ce6ef3ec6cab01fa4c666a09/LICENSE","artifact_url":"https://raw.githubusercontent.com/ODonnell-Lipidomics/LipidFinder/8306ca014c2e6b34ce6ef3ec6cab01fa4c666a09/LICENSE","version":"8306ca014c2e6b34ce6ef3ec6cab01fa4c666a09","retrieved_at":"2026-09-16T19:57:35.576108+00:00","artifact_sha256":"86608efc82f1b2b37ff0ae607d15cd4cd5f779b9a1dd8fe0410c7634216bf9be","locator":"LICENSE","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-56e5abfb5f12f1cd3b20","kind":"source","name":"alphagenome: Journal full-text XML","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12851941/","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12851941/fullTextXML","version":"Retrieved page snapshot; no immutable publisher revision supplied","retrieved_at":"2026-09-16T19:53:03.009088+00:00","artifact_sha256":"d159b791fc6cb7b679c151727b7b05a6e5b6388b08a89b675014983d85b6df36","locator":"Journal full-text XML","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-56f02d45976d011d80aa","kind":"source","name":"instadeepai/nucleotide-transformer: docs/agro_nucleotide_transformer.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/instadeepai/nucleotide-transformer/blob/2dc37b86e16a6970fbc731751f7719d9f676f7f9/docs/agro_nucleotide_transformer.md","artifact_url":"https://raw.githubusercontent.com/instadeepai/nucleotide-transformer/2dc37b86e16a6970fbc731751f7719d9f676f7f9/docs/agro_nucleotide_transformer.md","version":"2dc37b86e16a6970fbc731751f7719d9f676f7f9","retrieved_at":"2026-09-16T19:46:19.364532+00:00","artifact_sha256":"23e27d40473bbabadac45c56e8e282349f0b053da85fda89ff2125b5fe381fc6","locator":"docs/agro_nucleotide_transformer.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-576c2ecba240ddfda8f2","kind":"source","name":"agront: Journal full-text XML","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11233511/","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11233511/fullTextXML","version":"Retrieved page snapshot; no immutable publisher revision supplied","retrieved_at":"2026-09-16T20:16:14.422658+00:00","artifact_sha256":"ac36567140994e07e029042e4ac088d5895eabc2cf7ee79b174aee7d7a074070","locator":"Journal full-text XML","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-578439da3f3a3a476b7f","kind":"source","name":"google-deepmind/alphagenome_research: README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/google-deepmind/alphagenome_research/blob/0db53bd4352c66d1e00a049a81da373a066e6670/README.md","artifact_url":"https://raw.githubusercontent.com/google-deepmind/alphagenome_research/0db53bd4352c66d1e00a049a81da373a066e6670/README.md","version":"0db53bd4352c66d1e00a049a81da373a066e6670","retrieved_at":"2026-09-16T19:46:19.361780+00:00","artifact_sha256":"c715b22c31bd4f69e9b76236de0ab6a26c96fa0f384ad56a2f9d2cb65e262681","locator":"README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-5a9e55abdf288d9dd8da","kind":"source","name":"jwohlwend/boltz: docs/prediction.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/jwohlwend/boltz/blob/b1ebfc46ecf57f5414e0d1a6f9027bbb122c53bc/docs/prediction.md","artifact_url":"https://raw.githubusercontent.com/jwohlwend/boltz/b1ebfc46ecf57f5414e0d1a6f9027bbb122c53bc/docs/prediction.md","version":"b1ebfc46ecf57f5414e0d1a6f9027bbb122c53bc","retrieved_at":"2026-09-16T19:46:19.465940+00:00","artifact_sha256":"b9cb2ff437389864bde02e7e9fd9fbcc7de189ca2e240a5f3fb6261ca595a795","locator":"docs/prediction.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-5e268f31f0347fc5564c","kind":"source","name":"prokbert: Journal full-text XML","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC10810988/","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC10810988/fullTextXML","version":"Retrieved page snapshot; no immutable publisher revision supplied","retrieved_at":"2026-09-16T19:53:03.847832+00:00","artifact_sha256":"8610e2a54aa877c8dc565a9cdb6e82099f284c5e0907a52cab18d994ea732436","locator":"Journal full-text XML","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-6189052c702a948da02d","kind":"source","name":"biomap-research/scFoundation: LICENSE","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/biomap-research/scFoundation/blob/397631c495eddf9ad6644fc00c6ea8139e651245/LICENSE","artifact_url":"https://raw.githubusercontent.com/biomap-research/scFoundation/397631c495eddf9ad6644fc00c6ea8139e651245/LICENSE","version":"397631c495eddf9ad6644fc00c6ea8139e651245","retrieved_at":"2026-09-16T19:46:18.581063+00:00","artifact_sha256":"c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4","locator":"LICENSE","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-640665f30e62eed9319b","kind":"source","name":"ccsb-scripps/AutoDock-Vina: LICENSE","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/ccsb-scripps/AutoDock-Vina/blob/3c65c0b3e6c2c1d183f6a175ecb65e3c5ba91645/LICENSE","artifact_url":"https://raw.githubusercontent.com/ccsb-scripps/AutoDock-Vina/3c65c0b3e6c2c1d183f6a175ecb65e3c5ba91645/LICENSE","version":"3c65c0b3e6c2c1d183f6a175ecb65e3c5ba91645","retrieved_at":"2026-09-16T19:46:18.718558+00:00","artifact_sha256":"cfc7749b96f63bd31c3c42b5c471bf756814053e847c10f3eb003417bc523d30","locator":"LICENSE","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-657e83427ab59f3aec83","kind":"source","name":"instadeepai/nucleotide-transformer: docs/nucleotide_transformer.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/instadeepai/nucleotide-transformer/blob/2dc37b86e16a6970fbc731751f7719d9f676f7f9/docs/nucleotide_transformer.md","artifact_url":"https://raw.githubusercontent.com/instadeepai/nucleotide-transformer/2dc37b86e16a6970fbc731751f7719d9f676f7f9/docs/nucleotide_transformer.md","version":"2dc37b86e16a6970fbc731751f7719d9f676f7f9","retrieved_at":"2026-09-16T19:46:19.364532+00:00","artifact_sha256":"ab16d582de98652526b5cebb120eec969328f9db29dc741826bcd81c397e0672","locator":"docs/nucleotide_transformer.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-67ee1cc31060ba8c9569","kind":"source","name":"ArcInstitute/evo2: LICENSE","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/ArcInstitute/evo2/blob/53f195997257c56c00e5ef8d33a54f5baad143a6/LICENSE","artifact_url":"https://raw.githubusercontent.com/ArcInstitute/evo2/53f195997257c56c00e5ef8d33a54f5baad143a6/LICENSE","version":"53f195997257c56c00e5ef8d33a54f5baad143a6","retrieved_at":"2026-09-16T19:46:17.765915+00:00","artifact_sha256":"5bb5812fc2bfb2d777fe5621767172f8ef30be62ad719c87e0f10832336a99e0","locator":"LICENSE","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-6851724e3bcb7e9d2781","kind":"source","name":"scverse/scvi-tools: LICENSE","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/scverse/scvi-tools/blob/73b28e44223621470e582a81a102c107bb22678b/LICENSE","artifact_url":"https://raw.githubusercontent.com/scverse/scvi-tools/73b28e44223621470e582a81a102c107bb22678b/LICENSE","version":"73b28e44223621470e582a81a102c107bb22678b","retrieved_at":"2026-09-16T19:46:20.137176+00:00","artifact_sha256":"66399db0284d2539790efb348886ab0c1f745bbe5fea6ac38a00465a14adc8f5","locator":"LICENSE","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-6a331ce74fa7d11e0bfe","kind":"source","name":"biohub/ESMFold2: config.json","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://huggingface.co/biohub/ESMFold2/blob/69869f737beffec5294845ede23db5fc0b4f509e/config.json","artifact_url":"https://huggingface.co/biohub/ESMFold2/blob/69869f737beffec5294845ede23db5fc0b4f509e/config.json","version":"69869f737beffec5294845ede23db5fc0b4f509e","retrieved_at":"2026-09-16T20:22:27.767500+00:00","artifact_sha256":"72054c1af92f432ac2d8e0628ba08b377d04c6a8679b5128b798dbc7f644ddc8","locator":"config.json","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-6b79ddfdd330693bf4fb","kind":"source","name":"InstaDeepAI/nucleotide-transformer-v2-50m-multi-species: config.json","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://huggingface.co/InstaDeepAI/nucleotide-transformer-v2-50m-multi-species/blob/81b29e5786726d891dbf929404ef20adca5b36f1/config.json","artifact_url":"https://huggingface.co/InstaDeepAI/nucleotide-transformer-v2-50m-multi-species/blob/81b29e5786726d891dbf929404ef20adca5b36f1/config.json","version":"81b29e5786726d891dbf929404ef20adca5b36f1","retrieved_at":"2026-09-16T19:46:20.607357+00:00","artifact_sha256":"e20f497248c7cb264c7cd4582dbcfd52dc4cbf74a97fc711559b8c8f71c635db","locator":"config.json","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-6bcab1b3e31b52c38ea5","kind":"source","name":"BojarLab/glycowork: LICENSE","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/BojarLab/glycowork/blob/3d63f1ec25c850da3cde4d25cb602d50b6b5732b/LICENSE","artifact_url":"https://raw.githubusercontent.com/BojarLab/glycowork/3d63f1ec25c850da3cde4d25cb602d50b6b5732b/LICENSE","version":"3d63f1ec25c850da3cde4d25cb602d50b6b5732b","retrieved_at":"2026-09-16T19:46:17.767851+00:00","artifact_sha256":"21d65f90ec86f746730748d9d53988b0936a1451642e8ec773c45380f3eb63d0","locator":"LICENSE","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-71e8ebc066164f0ce44e","kind":"source","name":"matchms/matchms: LICENSE","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/matchms/matchms/blob/066608589587c8d089afd2e8d55ceadb2766ea62/LICENSE","artifact_url":"https://raw.githubusercontent.com/matchms/matchms/066608589587c8d089afd2e8d55ceadb2766ea62/LICENSE","version":"066608589587c8d089afd2e8d55ceadb2766ea62","retrieved_at":"2026-09-16T19:46:19.674309+00:00","artifact_sha256":"38fcd5e9b1c63c25c5ca5ca2091dc66002be10ff2c78b0db71b80ef3b0d8335a","locator":"LICENSE","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-7267eabca7c4a7945878","kind":"source","name":"zhihan1996/DNABERT-2-117M: README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://huggingface.co/zhihan1996/DNABERT-2-117M/blob/7bce263b15377fc15361f52cfab88f8b586abda0/README.md","artifact_url":"https://huggingface.co/zhihan1996/DNABERT-2-117M/blob/7bce263b15377fc15361f52cfab88f8b586abda0/README.md","version":"7bce263b15377fc15361f52cfab88f8b586abda0","retrieved_at":"2026-09-16T19:46:20.874956+00:00","artifact_sha256":"48b18abd051eb4952e0c0a50a0740387a1cea145be06517b67ab73a29431a3bc","locator":"README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-72fdaed158e4ff851fbe","kind":"source","name":"biobakery/MetaPhlAn: README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/biobakery/MetaPhlAn/blob/424f3e6e30618266404353e1083c6405a9f02f48/README.md","artifact_url":"https://raw.githubusercontent.com/biobakery/MetaPhlAn/424f3e6e30618266404353e1083c6405a9f02f48/README.md","version":"424f3e6e30618266404353e1083c6405a9f02f48","retrieved_at":"2026-09-16T19:46:18.561509+00:00","artifact_sha256":"ce491bb2d686145e0773c685d0d02e8a5fabc7daaea60eef07cf54298561fba7","locator":"README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-734117ed8501eb26fefc","kind":"source","name":"jwohlwend/boltz: src/boltz/model/models/boltz2.py","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/jwohlwend/boltz/blob/b1ebfc46ecf57f5414e0d1a6f9027bbb122c53bc/src/boltz/model/models/boltz2.py","artifact_url":"https://raw.githubusercontent.com/jwohlwend/boltz/b1ebfc46ecf57f5414e0d1a6f9027bbb122c53bc/src/boltz/model/models/boltz2.py","version":"b1ebfc46ecf57f5414e0d1a6f9027bbb122c53bc","retrieved_at":"2026-09-16T19:46:19.465940+00:00","artifact_sha256":"f05169e66488910fc11c6a56b56d19a58e3f43649586218df75573eeff7a9945","locator":"src/boltz/model/models/boltz2.py","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-760ab2fa8c396aeb796c","kind":"source","name":"msalign: Primary paper PDF","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://arxiv.org/pdf/2605.19752","artifact_url":"https://arxiv.org/pdf/2605.19752","version":"2605.19752v1","retrieved_at":"2026-09-16T20:04:04.777728+00:00","artifact_sha256":"7395a55141ee7916741b7d9e6d4f42a1d3a03217a36ce6824ac3ff48afd4f26f","locator":"Primary paper PDF","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-790ab6283e5bb91474ed","kind":"source","name":"matsui-lab/GlycanGT: model/config_tokengt.py","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/matsui-lab/GlycanGT/blob/96611518c971deb89215ca163deaf9de3a59fa32/model/config_tokengt.py","artifact_url":"https://raw.githubusercontent.com/matsui-lab/GlycanGT/96611518c971deb89215ca163deaf9de3a59fa32/model/config_tokengt.py","version":"96611518c971deb89215ca163deaf9de3a59fa32","retrieved_at":"2026-09-16T19:46:19.737544+00:00","artifact_sha256":"355c055f44e4b09c33ed640606800014b411e96674a1cdb48157c6d57ea5d426","locator":"model/config_tokengt.py","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-7972ce2bdd6a5e40d8ad","kind":"source","name":"biomap-research/scFoundation: model/README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/biomap-research/scFoundation/blob/397631c495eddf9ad6644fc00c6ea8139e651245/model/README.md","artifact_url":"https://raw.githubusercontent.com/biomap-research/scFoundation/397631c495eddf9ad6644fc00c6ea8139e651245/model/README.md","version":"397631c495eddf9ad6644fc00c6ea8139e651245","retrieved_at":"2026-09-16T19:46:18.581063+00:00","artifact_sha256":"f62879233ecf5fac407cd516aca2ddb3df20370523cb111c9d057676b4721d5d","locator":"model/README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-7b6fdf915d9c6950ad0d","kind":"source","name":"dauparas/ProteinMPNN: training/README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/dauparas/ProteinMPNN/blob/8907e6671bfbfc92303b5f79c4b5e6ce47cdef57/training/README.md","artifact_url":"https://raw.githubusercontent.com/dauparas/ProteinMPNN/8907e6671bfbfc92303b5f79c4b5e6ce47cdef57/training/README.md","version":"8907e6671bfbfc92303b5f79c4b5e6ce47cdef57","retrieved_at":"2026-09-16T19:46:18.948060+00:00","artifact_sha256":"7cbe5f3f1cef53f1954b15710317d9755b7b7b3febace4758190850f12e23029","locator":"training/README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-7c43b700aa87c22843a6","kind":"source","name":"BojarLab/glycowork: README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/BojarLab/glycowork/blob/3d63f1ec25c850da3cde4d25cb602d50b6b5732b/README.md","artifact_url":"https://raw.githubusercontent.com/BojarLab/glycowork/3d63f1ec25c850da3cde4d25cb602d50b6b5732b/README.md","version":"3d63f1ec25c850da3cde4d25cb602d50b6b5732b","retrieved_at":"2026-09-16T19:46:17.767851+00:00","artifact_sha256":"4d907a6723f3f56b14b13e35eeb0833ddf9a2f3512a1fafd07d79259105b264c","locator":"README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-7cb8cb7f091536b92c11","kind":"source","name":"jwohlwend/boltz: scripts/train/configs/full.yaml","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/jwohlwend/boltz/blob/b1ebfc46ecf57f5414e0d1a6f9027bbb122c53bc/scripts/train/configs/full.yaml","artifact_url":"https://raw.githubusercontent.com/jwohlwend/boltz/b1ebfc46ecf57f5414e0d1a6f9027bbb122c53bc/scripts/train/configs/full.yaml","version":"b1ebfc46ecf57f5414e0d1a6f9027bbb122c53bc","retrieved_at":"2026-09-16T19:46:19.465940+00:00","artifact_sha256":"70aa005d0c29b1c28918e104b92fd43b5cfde6157993dfcd8fe832e9a2e107dc","locator":"scripts/train/configs/full.yaml","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-7e4b193e47ba209860a1","kind":"source","name":"instadeepai/nucleotide-transformer: LICENSE.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/instadeepai/nucleotide-transformer/blob/2dc37b86e16a6970fbc731751f7719d9f676f7f9/LICENSE.md","artifact_url":"https://raw.githubusercontent.com/instadeepai/nucleotide-transformer/2dc37b86e16a6970fbc731751f7719d9f676f7f9/LICENSE.md","version":"2dc37b86e16a6970fbc731751f7719d9f676f7f9","retrieved_at":"2026-09-16T19:46:19.364532+00:00","artifact_sha256":"1349a4b6148492b44f629e64eed676612e234fe9a839e4f3b277c1482c8849f1","locator":"LICENSE.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-7e59c59bfea79722fc33","kind":"source","name":"InstaDeepAI/segment_nt: README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://huggingface.co/InstaDeepAI/segment_nt/blob/1048ad869036bb8e50a15d82b323ebeb52e8ab92/README.md","artifact_url":"https://huggingface.co/InstaDeepAI/segment_nt/blob/1048ad869036bb8e50a15d82b323ebeb52e8ab92/README.md","version":"1048ad869036bb8e50a15d82b323ebeb52e8ab92","retrieved_at":"2026-09-16T20:12:20.913226+00:00","artifact_sha256":"6e19e9c9293b0c392c053f92360b674b86c86c4457608b00c2fcb962bbd1d1bc","locator":"README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-81e51077d7d3352a6de4","kind":"source","name":"soedinglab/MMseqs2: README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/soedinglab/MMseqs2/blob/d401e78c2d18a822cdb1527d7464a043f6035a15/README.md","artifact_url":"https://raw.githubusercontent.com/soedinglab/MMseqs2/d401e78c2d18a822cdb1527d7464a043f6035a15/README.md","version":"d401e78c2d18a822cdb1527d7464a043f6035a15","retrieved_at":"2026-09-16T19:46:20.231620+00:00","artifact_sha256":"b6c591a763bf99c027857385f0e87ce5aa96caeaa74d71afd1fcec449eadb3d7","locator":"README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-849b751ed4bf88622749","kind":"source","name":"ArcInstitute/evo2_7b: README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://huggingface.co/ArcInstitute/evo2_7b/blob/bda0089f92582d5baabf0f22d9fc85f3588f6b58/README.md","artifact_url":"https://huggingface.co/ArcInstitute/evo2_7b/blob/bda0089f92582d5baabf0f22d9fc85f3588f6b58/README.md","version":"bda0089f92582d5baabf0f22d9fc85f3588f6b58","retrieved_at":"2026-09-16T20:04:02.231112+00:00","artifact_sha256":"802cb1e030bc560414a9fabcddd8fe243c296ad7d0b523dde0a3a0e1a7eaf794","locator":"README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-863f6c2e6e18152c2f8b","kind":"source","name":"soedinglab/hh-suite: README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/soedinglab/hh-suite/blob/43095e46ada4ec2a8a47d47ef5ad7e38b1429f7b/README.md","artifact_url":"https://raw.githubusercontent.com/soedinglab/hh-suite/43095e46ada4ec2a8a47d47ef5ad7e38b1429f7b/README.md","version":"43095e46ada4ec2a8a47d47ef5ad7e38b1429f7b","retrieved_at":"2026-09-16T19:46:20.402052+00:00","artifact_sha256":"2f8690d6a4a9767973ba9ea2af019d43d6108555ef0e9a69786033a71547d133","locator":"README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-864a2c5ea6e61aa310f9","kind":"source","name":"Illumina/SpliceAI: README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/Illumina/SpliceAI/blob/03f42437aaf56dc5dfd822c4ccee5aec1a705079/README.md","artifact_url":"https://raw.githubusercontent.com/Illumina/SpliceAI/03f42437aaf56dc5dfd822c4ccee5aec1a705079/README.md","version":"03f42437aaf56dc5dfd822c4ccee5aec1a705079","retrieved_at":"2026-09-16T19:46:17.769160+00:00","artifact_sha256":"8e5203afe343100832391e6155c7112f15cfe60bf0c21681d64e3420f854ef4d","locator":"README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-867f2d1dcb4084248daa","kind":"source","name":"facebook/esmfold_v1: config.json","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://huggingface.co/facebook/esmfold_v1/blob/75a3841ee059df2bf4d56688166c8fb459ddd97a/config.json","artifact_url":"https://huggingface.co/facebook/esmfold_v1/blob/75a3841ee059df2bf4d56688166c8fb459ddd97a/config.json","version":"75a3841ee059df2bf4d56688166c8fb459ddd97a","retrieved_at":"2026-09-16T20:04:02.231029+00:00","artifact_sha256":"6b98125e2685fef2875499f6bd7c83968a077993ab53f99bf5581113665f7cc6","locator":"config.json","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-8717ffb993cc8f7eddec","kind":"source","name":"evo2: Journal full-text XML","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC13128491/","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC13128491/fullTextXML","version":"Retrieved page snapshot; no immutable publisher revision supplied","retrieved_at":"2026-09-16T20:16:14.422554+00:00","artifact_sha256":"d043dbda49e023ef6b70e36e7ca7832bda6b7af6c5934368e8778a7f23cda8bd","locator":"Journal full-text XML","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-87c70bddba97a72a5e5d","kind":"source","name":"glycangt: Journal full-text XML","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://europepmc.org/article/PMC/PMC13105845","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC13105845/fullTextXML","version":"Primary article XML snapshot","retrieved_at":"2026-09-16T20:20:56.439056+00:00","artifact_sha256":"53e89a636c868c0329ee7eb6ae92f1028ec891940bc61730a148981b647fbbe5","locator":"Journal full-text XML","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-89de5de6fecdda12060d","kind":"source","name":"pluskal-lab/DreaMS: README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/pluskal-lab/DreaMS/blob/dbec3a0b514a99e5056cfccde4559fda8cfe8129/README.md","artifact_url":"https://raw.githubusercontent.com/pluskal-lab/DreaMS/dbec3a0b514a99e5056cfccde4559fda8cfe8129/README.md","version":"dbec3a0b514a99e5056cfccde4559fda8cfe8129","retrieved_at":"2026-09-16T19:46:20.105311+00:00","artifact_sha256":"a6afe934f9894cc71fb9b561c7a122d043b377c27b3fc6f18505fe65b6226256","locator":"README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-8a985eabfd054f0dec4d","kind":"source","name":"jwohlwend/boltz: LICENSE","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/jwohlwend/boltz/blob/b1ebfc46ecf57f5414e0d1a6f9027bbb122c53bc/LICENSE","artifact_url":"https://raw.githubusercontent.com/jwohlwend/boltz/b1ebfc46ecf57f5414e0d1a6f9027bbb122c53bc/LICENSE","version":"b1ebfc46ecf57f5414e0d1a6f9027bbb122c53bc","retrieved_at":"2026-09-16T19:46:19.465940+00:00","artifact_sha256":"f0667fd5e66c51e1ba8ddaa0249c6d7225b30037e02c45782d8f2c2943ac2617","locator":"LICENSE","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-8b4e1bd487c21cda9862","kind":"source","name":"cuhkaih/rhofold: README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://huggingface.co/cuhkaih/rhofold/blob/4458c1c5484a3a10a7f3059b9f0e8ca0447b23ac/README.md","artifact_url":"https://huggingface.co/cuhkaih/rhofold/blob/4458c1c5484a3a10a7f3059b9f0e8ca0447b23ac/README.md","version":"4458c1c5484a3a10a7f3059b9f0e8ca0447b23ac","retrieved_at":"2026-09-16T20:04:02.231237+00:00","artifact_sha256":"f5672ffed6edaea59f16fe42dfa0bac09473801cd90cbcb6d8ca00995a2969c2","locator":"README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-8e894fc6d180746a1808","kind":"source","name":"metagene-ai/METAGENE-1: README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://huggingface.co/metagene-ai/METAGENE-1/blob/ad8a1e0ee62b85058bfc05d823d8e8d4759edc48/README.md","artifact_url":"https://huggingface.co/metagene-ai/METAGENE-1/blob/ad8a1e0ee62b85058bfc05d823d8e8d4759edc48/README.md","version":"ad8a1e0ee62b85058bfc05d823d8e8d4759edc48","retrieved_at":"2026-09-16T19:46:20.736359+00:00","artifact_sha256":"277517aa69527fccaeb85c124f3ec9772c5508fadb5c6ad423bee39a3071ad17","locator":"README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-92b0ae57a020cbc40fe7","kind":"source","name":"proteinmpnn: Journal full-text XML","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC9997061/","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC9997061/fullTextXML","version":"Retrieved page snapshot; no immutable publisher revision supplied","retrieved_at":"2026-09-16T19:53:03.193318+00:00","artifact_sha256":"3e9042dc0ac2837e07654a43e74abcbcdd7dc4cf017fa81e2d9869b1fdb3c52e","locator":"Journal full-text XML","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-94c61d963d28fa3a8993","kind":"source","name":"https://meme-suite.org/meme/doc/copyright.html: copyright.html","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://meme-suite.org/meme/doc/copyright.html","artifact_url":"https://meme-suite.org/meme/doc/copyright.html","version":"Copyright 1994–2025; source snapshot 2026-09-16","retrieved_at":"2026-09-16T20:41:37.711422+00:00","artifact_sha256":"f54d004ea7c8e38222b76ffa4c04e072a204ccd948d41c3d46f55e36516b7262","locator":"copyright.html","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-950c72bc9d4bc037f6e2","kind":"source","name":"soedinglab/MMseqs2: LICENSE.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/soedinglab/MMseqs2/blob/d401e78c2d18a822cdb1527d7464a043f6035a15/LICENSE.md","artifact_url":"https://raw.githubusercontent.com/soedinglab/MMseqs2/d401e78c2d18a822cdb1527d7464a043f6035a15/LICENSE.md","version":"d401e78c2d18a822cdb1527d7464a043f6035a15","retrieved_at":"2026-09-16T19:46:20.231620+00:00","artifact_sha256":"adc3ea1f2f5096d2464460495e12a65a94f466edb9ccca50d1f25844ca83792a","locator":"LICENSE.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-955922130ff85a74fa18","kind":"source","name":"ml4bio/RhoFold: README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/ml4bio/RhoFold/blob/6bdfbda720184409eb682ce08c05d258162ddc48/README.md","artifact_url":"https://raw.githubusercontent.com/ml4bio/RhoFold/6bdfbda720184409eb682ce08c05d258162ddc48/README.md","version":"6bdfbda720184409eb682ce08c05d258162ddc48","retrieved_at":"2026-09-16T19:46:19.891935+00:00","artifact_sha256":"530ee4a54cc39d30c7b66ddafe6f28a6162219812d977473a2d0ade7ad1fcba6","locator":"README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-9576e3936b5a16a999cf","kind":"source","name":"matsui-lab/GlycanGT: LICENSE","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/matsui-lab/GlycanGT/blob/96611518c971deb89215ca163deaf9de3a59fa32/LICENSE","artifact_url":"https://raw.githubusercontent.com/matsui-lab/GlycanGT/96611518c971deb89215ca163deaf9de3a59fa32/LICENSE","version":"96611518c971deb89215ca163deaf9de3a59fa32","retrieved_at":"2026-09-16T19:46:19.737544+00:00","artifact_sha256":"c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4","locator":"LICENSE","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-95a620349435ae6651c5","kind":"source","name":"https://www.lipidmaps.org/resources/tools/lipidfinder/: page.html","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://www.lipidmaps.org/resources/tools/lipidfinder/","artifact_url":"https://www.lipidmaps.org/resources/tools/lipidfinder/","version":"Retrieved page snapshot; no immutable publisher revision supplied","retrieved_at":"2026-09-16T19:46:20.922105+00:00","artifact_sha256":"ae877d2c58b02e412782873bee4fc6216e8cb74e4497aa67f324e7f38a72e8ff","locator":"page.html","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-965600234c25fbd104c8","kind":"source","name":"neuralbioinfo/prokbert-mini: README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://huggingface.co/neuralbioinfo/prokbert-mini/blob/feb2520a43cd9cdb5b3d8477e47209dbcb55d1dc/README.md","artifact_url":"https://huggingface.co/neuralbioinfo/prokbert-mini/blob/feb2520a43cd9cdb5b3d8477e47209dbcb55d1dc/README.md","version":"feb2520a43cd9cdb5b3d8477e47209dbcb55d1dc","retrieved_at":"2026-09-16T20:04:02.231162+00:00","artifact_sha256":"172ec5b600e342302df5a22d84b12e1ab61d05722adda549b6d82f65e5c5d658","locator":"README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-96b5c3a31a50f7c259d6","kind":"source","name":"MAGICS-LAB/DNABERT_2: README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/MAGICS-LAB/DNABERT_2/blob/f25bed9ee20db966dff39e5c1571249d04e36404/README.md","artifact_url":"https://raw.githubusercontent.com/MAGICS-LAB/DNABERT_2/f25bed9ee20db966dff39e5c1571249d04e36404/README.md","version":"f25bed9ee20db966dff39e5c1571249d04e36404","retrieved_at":"2026-09-16T19:46:17.892989+00:00","artifact_sha256":"734a8cec5f667d74d421bf3b273ad7e256216109636da45aa7ceba21cd34de16","locator":"README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-97071500fc3422c426d4","kind":"source","name":"InstaDeepAI/segment_nt: config.json","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://huggingface.co/InstaDeepAI/segment_nt/blob/1048ad869036bb8e50a15d82b323ebeb52e8ab92/config.json","artifact_url":"https://huggingface.co/InstaDeepAI/segment_nt/blob/1048ad869036bb8e50a15d82b323ebeb52e8ab92/config.json","version":"1048ad869036bb8e50a15d82b323ebeb52e8ab92","retrieved_at":"2026-09-16T20:12:20.913226+00:00","artifact_sha256":"1412a48b99cfe79e6f34f84131ed964fd3ad5c4f3b3ea7db7ba2b0ac3466eadc","locator":"config.json","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-9863509ad10b824dc96b","kind":"source","name":"scfoundation-preprint: Primary paper PDF","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://yiheng-zhu.github.io/Yiheng/papers/5/scFoundation_bioRxiv_2023.pdf","artifact_url":"https://yiheng-zhu.github.io/Yiheng/papers/5/scFoundation_bioRxiv_2023.pdf","version":"bioRxiv manuscript posted 2023-06-15","retrieved_at":"2026-09-16T20:35:33.635910+00:00","artifact_sha256":"e2ce9d623a09b53eb863ae6ca359fe2956b7b0a7c3ebe7de6911aca3a660f737","locator":"Primary paper PDF","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-9a47f0326b33870b3113","kind":"source","name":"illumina/SpliceAI: LICENSE","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/illumina/SpliceAI/blob/03f42437aaf56dc5dfd822c4ccee5aec1a705079/LICENSE","artifact_url":"https://raw.githubusercontent.com/illumina/SpliceAI/03f42437aaf56dc5dfd822c4ccee5aec1a705079/LICENSE","version":"03f42437aaf56dc5dfd822c4ccee5aec1a705079","retrieved_at":"2026-09-16T19:46:19.364217+00:00","artifact_sha256":"67a909a0a8f8f7f45152207b6bcf9c78dd8a4dd3c8eef5bd11cd80a72e15344e","locator":"LICENSE","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-9b820532ba8e3965f64e","kind":"source","name":"illumina/SpliceAI: README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/illumina/SpliceAI/blob/03f42437aaf56dc5dfd822c4ccee5aec1a705079/README.md","artifact_url":"https://raw.githubusercontent.com/illumina/SpliceAI/03f42437aaf56dc5dfd822c4ccee5aec1a705079/README.md","version":"03f42437aaf56dc5dfd822c4ccee5aec1a705079","retrieved_at":"2026-09-16T19:46:19.364217+00:00","artifact_sha256":"8e5203afe343100832391e6155c7112f15cfe60bf0c21681d64e3420f854ef4d","locator":"README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-9ba2b6dc494c7b6df951","kind":"source","name":"DerrickWood/kraken2: docs/MANUAL.html","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/DerrickWood/kraken2/blob/8c190b1b668825935dbf6dee5f969227dc8269bb/docs/MANUAL.html","artifact_url":"https://raw.githubusercontent.com/DerrickWood/kraken2/8c190b1b668825935dbf6dee5f969227dc8269bb/docs/MANUAL.html","version":"8c190b1b668825935dbf6dee5f969227dc8269bb","retrieved_at":"2026-09-16T19:46:17.767980+00:00","artifact_sha256":"ef0962a736771bfb8373905d612d64f334544d625236f89e54cf3a04c9737016","locator":"docs/MANUAL.html","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-a07254268b1d1de1f7f8","kind":"source","name":"yeqinglin/genie3: README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://huggingface.co/yeqinglin/genie3/blob/9ae31ebb8c56eebdc05ab282a8fd3f6a6d2a03a2/README.md","artifact_url":"https://huggingface.co/yeqinglin/genie3/blob/9ae31ebb8c56eebdc05ab282a8fd3f6a6d2a03a2/README.md","version":"9ae31ebb8c56eebdc05ab282a8fd3f6a6d2a03a2","retrieved_at":"2026-09-16T20:22:27.767036+00:00","artifact_sha256":"4bcf87ecfbbb8e07a01b21415a970c8b53a5283bf6872b657040d3f45c9241f7","locator":"README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-a248990b44d481814f47","kind":"source","name":"ArcInstitute/evo2: README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/ArcInstitute/evo2/blob/53f195997257c56c00e5ef8d33a54f5baad143a6/README.md","artifact_url":"https://raw.githubusercontent.com/ArcInstitute/evo2/53f195997257c56c00e5ef8d33a54f5baad143a6/README.md","version":"53f195997257c56c00e5ef8d33a54f5baad143a6","retrieved_at":"2026-09-16T19:46:17.765915+00:00","artifact_sha256":"58787c8ef5cb4fba4c04322a4ceb9f174e2233ec22d4193622fb6bc67d651d89","locator":"README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-a48ae27aa000bdab7442","kind":"source","name":"zhihan1996/DNABERT-2-117M: LICENSE","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://huggingface.co/zhihan1996/DNABERT-2-117M/blob/7bce263b15377fc15361f52cfab88f8b586abda0/LICENSE","artifact_url":"https://huggingface.co/zhihan1996/DNABERT-2-117M/blob/7bce263b15377fc15361f52cfab88f8b586abda0/LICENSE","version":"7bce263b15377fc15361f52cfab88f8b586abda0","retrieved_at":"2026-09-16T19:46:20.874956+00:00","artifact_sha256":"c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4","locator":"LICENSE","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-a6b46962aae5cfc348c6","kind":"source","name":"metagene-ai/metagene-pretrain: README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/metagene-ai/metagene-pretrain/blob/82b9e142db2c7e0a268346d53344d6c3bf223066/README.md","artifact_url":"https://raw.githubusercontent.com/metagene-ai/metagene-pretrain/82b9e142db2c7e0a268346d53344d6c3bf223066/README.md","version":"82b9e142db2c7e0a268346d53344d6c3bf223066","retrieved_at":"2026-09-16T20:04:02.447523+00:00","artifact_sha256":"f09b3bc7f21c19d31ad90205da47010dc73b75096c12af5b350a925b4cac5a7b","locator":"README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-a766fb9176a491dc1ec3","kind":"source","name":"kundajelab/chrombpnet: README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/kundajelab/chrombpnet/blob/09938fdb4397ec0006510e5251e48920a505d4de/README.md","artifact_url":"https://raw.githubusercontent.com/kundajelab/chrombpnet/09938fdb4397ec0006510e5251e48920a505d4de/README.md","version":"09938fdb4397ec0006510e5251e48920a505d4de","retrieved_at":"2026-09-16T19:46:19.514840+00:00","artifact_sha256":"058c4f98218ea2ed854681126b6f682c9f3beec91275781fb37e39c55d0c92ea","locator":"README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-aadbeb0f10ec55d34f5f","kind":"source","name":"nt: Journal full-text XML","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11810778/","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11810778/fullTextXML","version":"Retrieved page snapshot; no immutable publisher revision supplied","retrieved_at":"2026-09-16T20:16:14.422628+00:00","artifact_sha256":"7c3b78a4f38ef053a08e466222e5662dd9711d91c535512d0b35d459d1fc7249","locator":"Journal full-text XML","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-ad7f5c1eb802d9895413","kind":"source","name":"aertslab/GENIE3: DESCRIPTION","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/aertslab/GENIE3/blob/54bc15636322e8773357de6e0b6683c6bc802825/DESCRIPTION","artifact_url":"https://raw.githubusercontent.com/aertslab/GENIE3/54bc15636322e8773357de6e0b6683c6bc802825/DESCRIPTION","version":"54bc15636322e8773357de6e0b6683c6bc802825","retrieved_at":"2026-09-16T19:46:18.166390+00:00","artifact_sha256":"4a2d48a4198b2ea6cbe5fa7749e9aa7286b318ebe19021bd1be52f29435e3196","locator":"DESCRIPTION","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-af4cadc1f0b4c69528d4","kind":"source","name":"aertslab/GRNBoost: LICENSE.txt","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/aertslab/GRNBoost/blob/26c836b3dcbb85852d3c6f4b8340e8655434da02/LICENSE.txt","artifact_url":"https://raw.githubusercontent.com/aertslab/GRNBoost/26c836b3dcbb85852d3c6f4b8340e8655434da02/LICENSE.txt","version":"26c836b3dcbb85852d3c6f4b8340e8655434da02","retrieved_at":"2026-09-16T19:46:18.179520+00:00","artifact_sha256":"16019e3b76d2c09bebcfb1186f59d35e40763142471ad79da82c87a8e27afaef","locator":"LICENSE.txt","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-af60e327735834e48fd7","kind":"source","name":"gcorso/DiffDock: LICENSE","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/gcorso/DiffDock/blob/85c49b60d3e0b0182a59ee43a34a6d7036981284/LICENSE","artifact_url":"https://raw.githubusercontent.com/gcorso/DiffDock/85c49b60d3e0b0182a59ee43a34a6d7036981284/LICENSE","version":"85c49b60d3e0b0182a59ee43a34a6d7036981284","retrieved_at":"2026-09-16T19:46:19.329065+00:00","artifact_sha256":"8efe3b8dac5c278ba290a66887e46261056f71dec75c2815c7da75451856cf1f","locator":"LICENSE","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-af8da643e933eae36a68","kind":"source","name":"polymathic-ai/MIMIC: README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://huggingface.co/polymathic-ai/MIMIC/blob/72e63a1ece34928422fd46f89b6a6580ace99a97/README.md","artifact_url":"https://huggingface.co/polymathic-ai/MIMIC/blob/72e63a1ece34928422fd46f89b6a6580ace99a97/README.md","version":"72e63a1ece34928422fd46f89b6a6580ace99a97","retrieved_at":"2026-09-16T19:46:20.846683+00:00","artifact_sha256":"462acc643cba24d1d4c6f1419838402c5f3196e94260039baf61a59ef0a0a107","locator":"README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-affbe2d511ff2a80f457","kind":"source","name":"chaidiscovery/chai-lab: README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/chaidiscovery/chai-lab/blob/66c38d1fe5c6756a89ff8596b1dea87d305ec06f/README.md","artifact_url":"https://raw.githubusercontent.com/chaidiscovery/chai-lab/66c38d1fe5c6756a89ff8596b1dea87d305ec06f/README.md","version":"66c38d1fe5c6756a89ff8596b1dea87d305ec06f","retrieved_at":"2026-09-16T19:46:18.824295+00:00","artifact_sha256":"ea6f6e64f6fc73d0e3dcbe6755c2aab226fa13bad279d1213acc0217f3f5013f","locator":"README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-b554679ea1503ce3d9f6","kind":"source","name":"metagene-ai/metagene-pretrain: train/config_hub/pretrain/genomicsllama.yml","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/metagene-ai/metagene-pretrain/blob/82b9e142db2c7e0a268346d53344d6c3bf223066/train/config_hub/pretrain/genomicsllama.yml","artifact_url":"https://raw.githubusercontent.com/metagene-ai/metagene-pretrain/82b9e142db2c7e0a268346d53344d6c3bf223066/train/config_hub/pretrain/genomicsllama.yml","version":"82b9e142db2c7e0a268346d53344d6c3bf223066","retrieved_at":"2026-09-16T20:04:02.447523+00:00","artifact_sha256":"0e764482ebf7e68c4751adfcf6430acdaa67fb683f59a31b9e25f8412d8ff5ad","locator":"train/config_hub/pretrain/genomicsllama.yml","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-b5e81b2c497214b18f8a","kind":"source","name":"snap-stanford/GEARS: gears/model.py","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/snap-stanford/GEARS/blob/f374e43e197b295016d80395d7a54ddb81cc6769/gears/model.py","artifact_url":"https://raw.githubusercontent.com/snap-stanford/GEARS/f374e43e197b295016d80395d7a54ddb81cc6769/gears/model.py","version":"f374e43e197b295016d80395d7a54ddb81cc6769","retrieved_at":"2026-09-16T19:46:20.191253+00:00","artifact_sha256":"07fee864bc5020807c90ed5443ff17530116f47f257332466243c233ef2bc857","locator":"gears/model.py","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-b8df803e8f55324a4eb5","kind":"source","name":"ArcInstitute/evo2_7b: config.json","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://huggingface.co/ArcInstitute/evo2_7b/blob/bda0089f92582d5baabf0f22d9fc85f3588f6b58/config.json","artifact_url":"https://huggingface.co/ArcInstitute/evo2_7b/blob/bda0089f92582d5baabf0f22d9fc85f3588f6b58/config.json","version":"bda0089f92582d5baabf0f22d9fc85f3588f6b58","retrieved_at":"2026-09-16T20:04:02.231112+00:00","artifact_sha256":"7f2e195e156de678b6d7db090dca19b37e88671f72ea7c9e1866e103946b69b2","locator":"config.json","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-b91fc1ec601eec6598c3","kind":"source","name":"chai1-web: Browser-extracted primary-paper passages","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://chaiassets.com/chai-1/paper/technical_report_v1.pdf","artifact_url":"https://chaiassets.com/chai-1/paper/technical_report_v1.pdf?trk=public_post_comment-text","version":"Technical report v1, 9 September 2024; browser-extracted passages pp.1–2,7–10","retrieved_at":"2026-09-16T20:25:06.247632+00:00","artifact_sha256":"8b6fcde51e45e97254308fead77ad8dae47d9990e5d6052221ce185644798608","locator":"Browser-extracted primary-paper passages","artifact_format":"browser_extracted_text","hash_scope":"SHA-256 of browser-extracted text artifact, not original PDF bytes; direct HTTP download returned403.","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-ba08b650243c9212de3f","kind":"source","name":"https://arxiv.org/abs/2605.19752: page.html","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://arxiv.org/abs/2605.19752","artifact_url":"https://arxiv.org/abs/2605.19752","version":"Retrieved page snapshot; no immutable publisher revision supplied","retrieved_at":"2026-09-16T19:46:17.765548+00:00","artifact_sha256":"e503f496841fd1d3836ebdab573de645257fc92fe7b5479056592da512e0ef3f","locator":"page.html","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-ba12a2d759aba6800ae6","kind":"source","name":"glycangt-supp: Publisher supplementary archive","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC13105845/supplementaryFiles","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC13105845/supplementaryFiles","version":"Retrieved page snapshot; no immutable publisher revision supplied","retrieved_at":"2026-09-16T20:42:10.495586+00:00","artifact_sha256":"17987fb07979d88c8f073905247bdacd22e4607e16c3c399bb292685e711846f","locator":"Publisher supplementary archive","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-bbfa62298e3bb7c5494d","kind":"source","name":"aqlaboratory/genie3: LICENSE","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/aqlaboratory/genie3/blob/d77ae5ac04212ff1e8b29b585859a3244c614804/LICENSE","artifact_url":"https://raw.githubusercontent.com/aqlaboratory/genie3/d77ae5ac04212ff1e8b29b585859a3244c614804/LICENSE","version":"d77ae5ac04212ff1e8b29b585859a3244c614804","retrieved_at":"2026-09-16T19:46:18.306022+00:00","artifact_sha256":"c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4","locator":"LICENSE","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-bc0327ef2ea6107e1773","kind":"source","name":"alphagenome-supp: Publisher supplementary archive","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12851941/supplementaryFiles","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12851941/supplementaryFiles","version":"Retrieved page snapshot; no immutable publisher revision supplied","retrieved_at":"2026-09-16T20:43:11.978411+00:00","artifact_sha256":"a006a3373cfdf42c41f9be7f64b883b38ea3f16be8037ea6024b5d0463f811dc","locator":"Publisher supplementary archive","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-bc3f1b0fa5b0999bbfa0","kind":"source","name":"kundajelab/chrombpnet: chrombpnet/training/models/chrombpnet_with_bias_model.py","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/kundajelab/chrombpnet/blob/09938fdb4397ec0006510e5251e48920a505d4de/chrombpnet/training/models/chrombpnet_with_bias_model.py","artifact_url":"https://raw.githubusercontent.com/kundajelab/chrombpnet/09938fdb4397ec0006510e5251e48920a505d4de/chrombpnet/training/models/chrombpnet_with_bias_model.py","version":"09938fdb4397ec0006510e5251e48920a505d4de","retrieved_at":"2026-09-16T19:46:19.514840+00:00","artifact_sha256":"bedc63a36ee27bb25ab39a00f1d9e3d4d64809754bf46aebe10d17b3a331dd33","locator":"chrombpnet/training/models/chrombpnet_with_bias_model.py","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-bd6ec1f819e22f075c32","kind":"source","name":"boltz2: Journal full-text XML","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://europepmc.org/article/PPR/PPR1039145","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PPR1039145/fullTextXML","version":"preprint; 1","retrieved_at":"2026-09-16T20:20:56.438787+00:00","artifact_sha256":"4ba7533096d594725e7f6362807aee6cc19596571fd09d9ae82fc49c6daa16fd","locator":"Journal full-text XML","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-bf1a5e03cd84873f4b04","kind":"source","name":"DerrickWood/kraken2: LICENSE","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/DerrickWood/kraken2/blob/8c190b1b668825935dbf6dee5f969227dc8269bb/LICENSE","artifact_url":"https://raw.githubusercontent.com/DerrickWood/kraken2/8c190b1b668825935dbf6dee5f969227dc8269bb/LICENSE","version":"8c190b1b668825935dbf6dee5f969227dc8269bb","retrieved_at":"2026-09-16T19:46:17.767980+00:00","artifact_sha256":"ef3803fed10bf0eae6919db5e12204af4460e73234953373b22bc6de04ed840a","locator":"LICENSE","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-bfbc8babf3630ddb3ed7","kind":"source","name":"biomap-research/scFoundation: README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/biomap-research/scFoundation/blob/397631c495eddf9ad6644fc00c6ea8139e651245/README.md","artifact_url":"https://raw.githubusercontent.com/biomap-research/scFoundation/397631c495eddf9ad6644fc00c6ea8139e651245/README.md","version":"397631c495eddf9ad6644fc00c6ea8139e651245","retrieved_at":"2026-09-16T19:46:18.581063+00:00","artifact_sha256":"02a7ae44cf2cc9948b5c1520b18a261b45abf9f17e3149e37d9f48072184f376","locator":"README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-c0351b621280a9c081da","kind":"source","name":"rfdiffusion: Journal full-text XML","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC10468394/","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC10468394/fullTextXML","version":"Retrieved page snapshot; no immutable publisher revision supplied","retrieved_at":"2026-09-16T20:16:14.422733+00:00","artifact_sha256":"bd41c7070976ae82e317a10367c81492a2692482c302a25c7f2a28ffe6ef4110","locator":"Journal full-text XML","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-c037e3419c04936262a0","kind":"source","name":"snap-stanford/GEARS: README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/snap-stanford/GEARS/blob/f374e43e197b295016d80395d7a54ddb81cc6769/README.md","artifact_url":"https://raw.githubusercontent.com/snap-stanford/GEARS/f374e43e197b295016d80395d7a54ddb81cc6769/README.md","version":"f374e43e197b295016d80395d7a54ddb81cc6769","retrieved_at":"2026-09-16T19:46:20.191253+00:00","artifact_sha256":"fe78b7b1ede50c673b8f0d91fb2637707685c78b269ee157edb6cbc4d510d61b","locator":"README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-c0ac4d10941b082e5f86","kind":"source","name":"aqlaboratory/openfold: README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/aqlaboratory/openfold/blob/be2ec1841f16c966c65ae0e7599ebbadc725757d/README.md","artifact_url":"https://raw.githubusercontent.com/aqlaboratory/openfold/be2ec1841f16c966c65ae0e7599ebbadc725757d/README.md","version":"be2ec1841f16c966c65ae0e7599ebbadc725757d","retrieved_at":"2026-09-16T19:46:18.314650+00:00","artifact_sha256":"985b6dc09144ea14378dd2c543b288e8d2a05cb342a77cbc56b7a39c5170f638","locator":"README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-c14c0e6a0fdbe1479658","kind":"source","name":"PolymathicAI/MIMIC: README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/PolymathicAI/MIMIC/blob/9e652f16491e6c2e3881111e24b285c288554275/README.md","artifact_url":"https://raw.githubusercontent.com/PolymathicAI/MIMIC/9e652f16491e6c2e3881111e24b285c288554275/README.md","version":"9e652f16491e6c2e3881111e24b285c288554275","retrieved_at":"2026-09-16T20:04:02.231308+00:00","artifact_sha256":"694bed54356d27cf087890ed80649c08ef4520a0eb5c85374a7347138a6cbb2f","locator":"README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-c2dd029568d234cb0d16","kind":"source","name":"metagene-ai/metagene-pretrain: train/LICENSE","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/metagene-ai/metagene-pretrain/blob/82b9e142db2c7e0a268346d53344d6c3bf223066/train/LICENSE","artifact_url":"https://raw.githubusercontent.com/metagene-ai/metagene-pretrain/82b9e142db2c7e0a268346d53344d6c3bf223066/train/LICENSE","version":"82b9e142db2c7e0a268346d53344d6c3bf223066","retrieved_at":"2026-09-16T20:04:02.447523+00:00","artifact_sha256":"ab89ceca19ca531005699095ca542f982fac92efd82f954f49ac17aa99a37cda","locator":"train/LICENSE","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-c303652f3d79132a00c2","kind":"source","name":"songlab-cal/tape: tape/models/modeling_bert.py","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/songlab-cal/tape/blob/6d345c2b2bbf52cd32cf179325c222afd92aec7e/tape/models/modeling_bert.py","artifact_url":"https://raw.githubusercontent.com/songlab-cal/tape/6d345c2b2bbf52cd32cf179325c222afd92aec7e/tape/models/modeling_bert.py","version":"6d345c2b2bbf52cd32cf179325c222afd92aec7e","retrieved_at":"2026-09-16T19:46:20.570146+00:00","artifact_sha256":"eba47206598a0e93e2f662d9c87343222556f236ac37373236e3970c43fb521d","locator":"tape/models/modeling_bert.py","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-c3d7a64294dcad7c6405","kind":"source","name":"google-deepmind/alphagenome_research: LICENSE","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/google-deepmind/alphagenome_research/blob/0db53bd4352c66d1e00a049a81da373a066e6670/LICENSE","artifact_url":"https://raw.githubusercontent.com/google-deepmind/alphagenome_research/0db53bd4352c66d1e00a049a81da373a066e6670/LICENSE","version":"0db53bd4352c66d1e00a049a81da373a066e6670","retrieved_at":"2026-09-16T19:46:19.361780+00:00","artifact_sha256":"cfc7749b96f63bd31c3c42b5c471bf756814053e847c10f3eb003417bc523d30","locator":"LICENSE","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-c465f6d4fcc04f9afe4b","kind":"source","name":"Illumina/SpliceAI: LICENSE","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/Illumina/SpliceAI/blob/03f42437aaf56dc5dfd822c4ccee5aec1a705079/LICENSE","artifact_url":"https://raw.githubusercontent.com/Illumina/SpliceAI/03f42437aaf56dc5dfd822c4ccee5aec1a705079/LICENSE","version":"03f42437aaf56dc5dfd822c4ccee5aec1a705079","retrieved_at":"2026-09-16T19:46:17.769160+00:00","artifact_sha256":"67a909a0a8f8f7f45152207b6bcf9c78dd8a4dd3c8eef5bd11cd80a72e15344e","locator":"LICENSE","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-c7befbb0fbd8620d5c3c","kind":"source","name":"RosettaCommons/RFdiffusion: README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/RosettaCommons/RFdiffusion/blob/86507b6538f51fce57b5a72477165f03999ed7ae/README.md","artifact_url":"https://raw.githubusercontent.com/RosettaCommons/RFdiffusion/86507b6538f51fce57b5a72477165f03999ed7ae/README.md","version":"86507b6538f51fce57b5a72477165f03999ed7ae","retrieved_at":"2026-09-16T19:46:18.126662+00:00","artifact_sha256":"d89eb9790ce805ce23e1b1a6804f4d67c81056025e4852b875f8ddfd87a08125","locator":"README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-c7d214a82cfd1afc827b","kind":"source","name":"aqlaboratory/openfold: LICENSE","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/aqlaboratory/openfold/blob/be2ec1841f16c966c65ae0e7599ebbadc725757d/LICENSE","artifact_url":"https://raw.githubusercontent.com/aqlaboratory/openfold/be2ec1841f16c966c65ae0e7599ebbadc725757d/LICENSE","version":"be2ec1841f16c966c65ae0e7599ebbadc725757d","retrieved_at":"2026-09-16T19:46:18.314650+00:00","artifact_sha256":"7a77a0f9b49715d3b124b39607c53c5df8795d26f7aae08b12e6dbb9b1ec1a40","locator":"LICENSE","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-d3d9afa5c6172982ed01","kind":"source","name":"ctheodoris/Geneformer: README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://huggingface.co/ctheodoris/Geneformer/blob/1f7fbae4e469a5f4f1af8c111a529cfe1b3829f5/README.md","artifact_url":"https://huggingface.co/ctheodoris/Geneformer/blob/1f7fbae4e469a5f4f1af8c111a529cfe1b3829f5/README.md","version":"1f7fbae4e469a5f4f1af8c111a529cfe1b3829f5","retrieved_at":"2026-09-16T19:46:20.640731+00:00","artifact_sha256":"56d6e570b349cbedae9a54634421c94e7af8ea467ce0fdd79193372ae3cbdbd8","locator":"README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-d65bf8ba587b6841e0d9","kind":"source","name":"biohub/ESMFold2: README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://huggingface.co/biohub/ESMFold2/blob/69869f737beffec5294845ede23db5fc0b4f509e/README.md","artifact_url":"https://huggingface.co/biohub/ESMFold2/blob/69869f737beffec5294845ede23db5fc0b4f509e/README.md","version":"69869f737beffec5294845ede23db5fc0b4f509e","retrieved_at":"2026-09-16T20:22:27.767500+00:00","artifact_sha256":"7f20294b352b7e490689229c0fe26760115e3ffc66869d96dd89347729fe941d","locator":"README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-d7e5c63feb0e0e627625","kind":"source","name":"facebook/esm2_t33_650M_UR50D: README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://huggingface.co/facebook/esm2_t33_650M_UR50D/blob/08e4846e537177426273712802403f7ba8261b6c/README.md","artifact_url":"https://huggingface.co/facebook/esm2_t33_650M_UR50D/blob/08e4846e537177426273712802403f7ba8261b6c/README.md","version":"08e4846e537177426273712802403f7ba8261b6c","retrieved_at":"2026-09-16T20:04:02.230771+00:00","artifact_sha256":"462a2f24724e19c6be0efab926315c294a863c9a9770e2c8b3d859b2d81a07de","locator":"README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-d8e2d4f2e752370b6565","kind":"source","name":"bowang-lab/scGPT: scgpt/model/model.py","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/bowang-lab/scGPT/blob/cebd6fae655b9c585a4807daa3ac31bb764f06b4/scgpt/model/model.py","artifact_url":"https://raw.githubusercontent.com/bowang-lab/scGPT/cebd6fae655b9c585a4807daa3ac31bb764f06b4/scgpt/model/model.py","version":"cebd6fae655b9c585a4807daa3ac31bb764f06b4","retrieved_at":"2026-09-16T19:46:18.594429+00:00","artifact_sha256":"4ae77618cc6f12a7d1f6d946fee3f7bd5c56f3e9e103073394a590ceba5a35ed","locator":"scgpt/model/model.py","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-da4566a88ed9fa42fdcb","kind":"source","name":"segmentnt: Journal full-text XML","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12615259/","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12615259/fullTextXML","version":"Retrieved page snapshot; no immutable publisher revision supplied","retrieved_at":"2026-09-16T20:16:14.422701+00:00","artifact_sha256":"b9bddff89c5f8898fc677008205a47b960ea569f3887cc15845e2547b92aaa70","locator":"Journal full-text XML","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-dacb012b22c0a7de4880","kind":"source","name":"Akikitani295/GlycanGT: README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://huggingface.co/Akikitani295/GlycanGT/blob/7c9f4a19b366ca48ed03995a0ad47fa1e745f24c/README.md","artifact_url":"https://huggingface.co/Akikitani295/GlycanGT/blob/7c9f4a19b366ca48ed03995a0ad47fa1e745f24c/README.md","version":"7c9f4a19b366ca48ed03995a0ad47fa1e745f24c","retrieved_at":"2026-09-16T20:04:02.231272+00:00","artifact_sha256":"16a2967195f6be2d4478faf7151ceb7b96de4fa97ff168c7c975dc01d1dd69d3","locator":"README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-dc53e8df56bb7352c014","kind":"source","name":"calico/basenji: README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/calico/basenji/blob/06ce5d387e20b47184d05433b3983163c5f923cd/README.md","artifact_url":"https://raw.githubusercontent.com/calico/basenji/06ce5d387e20b47184d05433b3983163c5f923cd/README.md","version":"06ce5d387e20b47184d05433b3983163c5f923cd","retrieved_at":"2026-09-16T19:46:18.649671+00:00","artifact_sha256":"b77d93b4ff352874ecd21e074357b0bb7f4f09e7bfd53d8f7f152d05063c30c1","locator":"README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-dd29c6cbb629171059a6","kind":"source","name":"tkzeng/Pangolin: LICENSE","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/tkzeng/Pangolin/blob/5cf94b8db938c658391b4305cd7ce33297d44ff7/LICENSE","artifact_url":"https://raw.githubusercontent.com/tkzeng/Pangolin/5cf94b8db938c658391b4305cd7ce33297d44ff7/LICENSE","version":"5cf94b8db938c658391b4305cd7ce33297d44ff7","retrieved_at":"2026-09-16T19:46:20.582143+00:00","artifact_sha256":"3972dc9744f6499f0f9b2dbf76696f2ae7ad8af9b23dde66d6af86c9dfb36986","locator":"LICENSE","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-de1ee43bdd6f04de9850","kind":"source","name":"pangolin: Journal full-text XML","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC9022248/","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC9022248/fullTextXML","version":"Retrieved page snapshot; no immutable publisher revision supplied","retrieved_at":"2026-09-16T19:53:03.008861+00:00","artifact_sha256":"c51d34f0bc17ffd34ee5c7caf4bbf627edd9f75f57497c8d6b133f405ec4c307","locator":"Journal full-text XML","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-ded281404bd1a0f3fdb7","kind":"source","name":"spliceai: Primary paper PDF","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://doi.org/10.1016/j.cell.2018.12.015","artifact_url":"https://www.marcottelab.org/users/BCH394P_364C_2021/SplicingAI-jaganathan2019.pdf","version":"Cell 2019 published article PDF, university-hosted copy","retrieved_at":"2026-09-16T20:04:38.697838+00:00","artifact_sha256":"e0d62bfd97c26907e184a6a262eb277e3ad0737322373df282c9832b56697dda","locator":"Primary paper PDF","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-e17e3864c464d981afc7","kind":"source","name":"ml4bio/RNA-FM: fm/pretrained.py","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/ml4bio/RNA-FM/blob/348951516e0963d22bbb33b3c9fc18c89081d38e/fm/pretrained.py","artifact_url":"https://raw.githubusercontent.com/ml4bio/RNA-FM/348951516e0963d22bbb33b3c9fc18c89081d38e/fm/pretrained.py","version":"348951516e0963d22bbb33b3c9fc18c89081d38e","retrieved_at":"2026-09-16T19:46:19.769679+00:00","artifact_sha256":"ab7271f8dbb876dbc5e2010e2d88b086088aada096f04149436d315a4f256aac","locator":"fm/pretrained.py","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-e508b40498c3a3ba53d5","kind":"source","name":"matchms/matchms: README.rst","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/matchms/matchms/blob/066608589587c8d089afd2e8d55ceadb2766ea62/README.rst","artifact_url":"https://raw.githubusercontent.com/matchms/matchms/066608589587c8d089afd2e8d55ceadb2766ea62/README.rst","version":"066608589587c8d089afd2e8d55ceadb2766ea62","retrieved_at":"2026-09-16T19:46:19.674309+00:00","artifact_sha256":"06a0c6b7adf444d7af53d20ce94c3cef483bd7be8db02fd8960c0704680d61ee","locator":"README.rst","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-e604b144dfbd97bf8917","kind":"source","name":"CAMI-challenge/CAMISIM: README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/CAMI-challenge/CAMISIM/blob/7ce6013c6d5a0fac8ba8a80e52a03560d3546fa6/README.md","artifact_url":"https://raw.githubusercontent.com/CAMI-challenge/CAMISIM/7ce6013c6d5a0fac8ba8a80e52a03560d3546fa6/README.md","version":"7ce6013c6d5a0fac8ba8a80e52a03560d3546fa6","retrieved_at":"2026-09-16T19:46:17.767933+00:00","artifact_sha256":"ec5309bf16daab6c9a0adb393b191a3e716b9c627834cf3e08655c250a63da61","locator":"README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-e69877a2a54345e162b3","kind":"source","name":"biohub/ESMC-6B: config.json","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://huggingface.co/biohub/ESMC-6B/blob/af1602ba7406f521b11bf8f81d52af378cde09e4/config.json","artifact_url":"https://huggingface.co/biohub/ESMC-6B/blob/af1602ba7406f521b11bf8f81d52af378cde09e4/config.json","version":"af1602ba7406f521b11bf8f81d52af378cde09e4","retrieved_at":"2026-09-16T20:22:27.767420+00:00","artifact_sha256":"80ccf981fbbd4f9f0a64830a0b64f1c79ed0f629d3f3b95f4424fd73174ac540","locator":"config.json","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-e847949b10d83e3b0efe","kind":"source","name":"songlab-cal/tape: LICENSE","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/songlab-cal/tape/blob/6d345c2b2bbf52cd32cf179325c222afd92aec7e/LICENSE","artifact_url":"https://raw.githubusercontent.com/songlab-cal/tape/6d345c2b2bbf52cd32cf179325c222afd92aec7e/LICENSE","version":"6d345c2b2bbf52cd32cf179325c222afd92aec7e","retrieved_at":"2026-09-16T19:46:20.570146+00:00","artifact_sha256":"b03177f56a7d17c782dfd8400ca4a29c272f46fd2ed48569d9194682f27e7001","locator":"LICENSE","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-ea568a52ea3ab476db6a","kind":"source","name":"dauparas/ProteinMPNN: README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/dauparas/ProteinMPNN/blob/8907e6671bfbfc92303b5f79c4b5e6ce47cdef57/README.md","artifact_url":"https://raw.githubusercontent.com/dauparas/ProteinMPNN/8907e6671bfbfc92303b5f79c4b5e6ce47cdef57/README.md","version":"8907e6671bfbfc92303b5f79c4b5e6ce47cdef57","retrieved_at":"2026-09-16T19:46:18.948060+00:00","artifact_sha256":"772ebe52d2ba5100a28a888910c6f0c9fd4ded1d1372e3d89f6f1c48707e0365","locator":"README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-eaf3efab8850672f44e4","kind":"source","name":"tkzeng/Pangolin: README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/tkzeng/Pangolin/blob/5cf94b8db938c658391b4305cd7ce33297d44ff7/README.md","artifact_url":"https://raw.githubusercontent.com/tkzeng/Pangolin/5cf94b8db938c658391b4305cd7ce33297d44ff7/README.md","version":"5cf94b8db938c658391b4305cd7ce33297d44ff7","retrieved_at":"2026-09-16T19:46:20.582143+00:00","artifact_sha256":"9117f9d255d6b6e810d224a600d381192bccccd3f0cd417fed9559d49ca8fffd","locator":"README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-eb328c1926fbe500f86b","kind":"source","name":"opencobra/cobrapy: README.rst","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/opencobra/cobrapy/blob/5aa19300fbf5dd632a9ac5c39ca28c1c621b943d/README.rst","artifact_url":"https://raw.githubusercontent.com/opencobra/cobrapy/5aa19300fbf5dd632a9ac5c39ca28c1c621b943d/README.rst","version":"5aa19300fbf5dd632a9ac5c39ca28c1c621b943d","retrieved_at":"2026-09-16T19:46:19.916278+00:00","artifact_sha256":"ae2da4ad3ae1d822e0010b6c023b26ef1d02b97007e17611dd4b1e22c70d4998","locator":"README.rst","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-ed0382dc026ea7080abd","kind":"source","name":"ArcInstitute/state: LICENSE","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/ArcInstitute/state/blob/9bbfe78a434a55205e4de834e1ea99f85f7a3add/LICENSE","artifact_url":"https://raw.githubusercontent.com/ArcInstitute/state/9bbfe78a434a55205e4de834e1ea99f85f7a3add/LICENSE","version":"9bbfe78a434a55205e4de834e1ea99f85f7a3add","retrieved_at":"2026-09-16T19:46:17.767683+00:00","artifact_sha256":"e66c269d4819aaab34b49ef5220c4ddab6756f21bb5180761a4eb8561f2b7bbd","locator":"LICENSE","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-edce0c06db05117706cb","kind":"source","name":"calico/basenji: docs/train.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/calico/basenji/blob/06ce5d387e20b47184d05433b3983163c5f923cd/docs/train.md","artifact_url":"https://raw.githubusercontent.com/calico/basenji/06ce5d387e20b47184d05433b3983163c5f923cd/docs/train.md","version":"06ce5d387e20b47184d05433b3983163c5f923cd","retrieved_at":"2026-09-16T19:46:18.649671+00:00","artifact_sha256":"2e42f0173b07e3493a570b901639a182b421c66c60741141278336b7757b76e6","locator":"docs/train.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-ee6e270af125ddf16eb3","kind":"source","name":"neuralbioinfo/prokbert-mini: config.json","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://huggingface.co/neuralbioinfo/prokbert-mini/blob/feb2520a43cd9cdb5b3d8477e47209dbcb55d1dc/config.json","artifact_url":"https://huggingface.co/neuralbioinfo/prokbert-mini/blob/feb2520a43cd9cdb5b3d8477e47209dbcb55d1dc/config.json","version":"feb2520a43cd9cdb5b3d8477e47209dbcb55d1dc","retrieved_at":"2026-09-16T20:04:02.231162+00:00","artifact_sha256":"6d68c922f919e0de3f87de60c07ebc1823927df806645937f94911c5b8245de6","locator":"config.json","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-eebe7c91156963e6ddc0","kind":"source","name":"dauparas/ProteinMPNN: LICENSE","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/dauparas/ProteinMPNN/blob/8907e6671bfbfc92303b5f79c4b5e6ce47cdef57/LICENSE","artifact_url":"https://raw.githubusercontent.com/dauparas/ProteinMPNN/8907e6671bfbfc92303b5f79c4b5e6ce47cdef57/LICENSE","version":"8907e6671bfbfc92303b5f79c4b5e6ce47cdef57","retrieved_at":"2026-09-16T19:46:18.948060+00:00","artifact_sha256":"82009d25ce585631f452b2b24589bdb29c559ccfefa2f200ef312ed5b501a586","locator":"LICENSE","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-efa6ec9d886a21534a20","kind":"source","name":"scgpt-may2023: Primary paper PDF","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://www.biorxiv.org/content/10.1101/2023.04.30.538439v1","artifact_url":"https://www.biorxiv.org/content/10.1101/2023.04.30.538439v1.full.pdf","version":"bioRxiv version posted 1 May 2023; re-inspected existing local research artifact","retrieved_at":"2026-09-16T20:21:23.046376+00:00","artifact_sha256":"4ef64d2b3431812df07c4b3e428950644ff264ec5728267c2e4692588f557419","locator":"Primary paper PDF","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Reused local archived primary PDF; no successful new publisher download claimed.","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-efc93d867cbf62e80ebb","kind":"source","name":"sokrypton/ColabFold: README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/sokrypton/ColabFold/blob/84c27d9cc500489fd9b97545d2325b9d00f251d5/README.md","artifact_url":"https://raw.githubusercontent.com/sokrypton/ColabFold/84c27d9cc500489fd9b97545d2325b9d00f251d5/README.md","version":"84c27d9cc500489fd9b97545d2325b9d00f251d5","retrieved_at":"2026-09-16T19:46:20.488416+00:00","artifact_sha256":"a384ec62c86d568aeb0fe3e8a3b111f071fbcfacb937b407eb6da24fc70c823f","locator":"README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-f06a8695b2915f86a45a","kind":"source","name":"snap-stanford/GEARS: LICENSE","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/snap-stanford/GEARS/blob/f374e43e197b295016d80395d7a54ddb81cc6769/LICENSE","artifact_url":"https://raw.githubusercontent.com/snap-stanford/GEARS/f374e43e197b295016d80395d7a54ddb81cc6769/LICENSE","version":"f374e43e197b295016d80395d7a54ddb81cc6769","retrieved_at":"2026-09-16T19:46:20.191253+00:00","artifact_sha256":"5c608f2c9421da729963328261ebffd23127dc55184fa8acc96acc52d8fd8b90","locator":"LICENSE","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-f078ec10fa45b5d77ad7","kind":"source","name":"aertslab/GENIE3: README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/aertslab/GENIE3/blob/54bc15636322e8773357de6e0b6683c6bc802825/README.md","artifact_url":"https://raw.githubusercontent.com/aertslab/GENIE3/54bc15636322e8773357de6e0b6683c6bc802825/README.md","version":"54bc15636322e8773357de6e0b6683c6bc802825","retrieved_at":"2026-09-16T19:46:18.166390+00:00","artifact_sha256":"045aa63715acf615327d45ff42a897e77fa6d4b88084d2ad4e2ea836eb5fb48a","locator":"README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-f0b17e30e69338b7762f","kind":"source","name":"https://meme-suite.org/meme/tools/fimo: page.html","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://meme-suite.org/meme/tools/fimo","artifact_url":"https://meme-suite.org/meme/tools/fimo","version":"Retrieved page snapshot; no immutable publisher revision supplied","retrieved_at":"2026-09-16T19:46:20.892034+00:00","artifact_sha256":"c0acb8f22d51b718d5c3a253e81b9882d804fa2aa153b6fcf3e826e22f085dd3","locator":"page.html","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-f29710879cf7a4438b12","kind":"source","name":"biobakery/humann: LICENSE","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/biobakery/humann/blob/e07b3a34d0b94c09a8ac5d28ff95009611178be2/LICENSE","artifact_url":"https://raw.githubusercontent.com/biobakery/humann/e07b3a34d0b94c09a8ac5d28ff95009611178be2/LICENSE","version":"e07b3a34d0b94c09a8ac5d28ff95009611178be2","retrieved_at":"2026-09-16T19:46:18.580457+00:00","artifact_sha256":"8b5d1a6cf9972029766b8909ad6d8a9459b2e6bab38d356b6a7953d850d74709","locator":"LICENSE","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-f3fd16387ba8459ecf34","kind":"source","name":"sokrypton/ColabFold: LICENSE","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/sokrypton/ColabFold/blob/84c27d9cc500489fd9b97545d2325b9d00f251d5/LICENSE","artifact_url":"https://raw.githubusercontent.com/sokrypton/ColabFold/84c27d9cc500489fd9b97545d2325b9d00f251d5/LICENSE","version":"84c27d9cc500489fd9b97545d2325b9d00f251d5","retrieved_at":"2026-09-16T19:46:20.488416+00:00","artifact_sha256":"f7a0429ad8a5ebc35994ba24d65a2962fc8c106ff28d2c7e43125ba4d2208108","locator":"LICENSE","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-f464cd407a6a9d031720","kind":"source","name":"gears: Journal full-text XML","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11180609/","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11180609/fullTextXML","version":"Retrieved page snapshot; no immutable publisher revision supplied","retrieved_at":"2026-09-16T20:16:14.422594+00:00","artifact_sha256":"6e9a8f4a2b8ccbc9aa46c43a34701b687f71b797048adf27aae2a99d9e2dcad9","locator":"Journal full-text XML","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-f60303c5eaa60ed9c77a","kind":"source","name":"ctheodoris/Geneformer: config.json","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://huggingface.co/ctheodoris/Geneformer/blob/1f7fbae4e469a5f4f1af8c111a529cfe1b3829f5/config.json","artifact_url":"https://huggingface.co/ctheodoris/Geneformer/blob/1f7fbae4e469a5f4f1af8c111a529cfe1b3829f5/config.json","version":"1f7fbae4e469a5f4f1af8c111a529cfe1b3829f5","retrieved_at":"2026-09-16T19:46:20.640731+00:00","artifact_sha256":"2cc4af3442644e84af71814a607b61c958d7b4e5ebde6791ab77b5f534ac6f6e","locator":"config.json","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-f6c94a04c51af984a008","kind":"source","name":"pluskal-lab/DreaMS: LICENSE","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/pluskal-lab/DreaMS/blob/dbec3a0b514a99e5056cfccde4559fda8cfe8129/LICENSE","artifact_url":"https://raw.githubusercontent.com/pluskal-lab/DreaMS/dbec3a0b514a99e5056cfccde4559fda8cfe8129/LICENSE","version":"dbec3a0b514a99e5056cfccde4559fda8cfe8129","retrieved_at":"2026-09-16T19:46:20.105311+00:00","artifact_sha256":"747bb9267b9d0e006a0486165ac2c9d5bd6481dfa7aefe808832703224af005d","locator":"LICENSE","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-f71e6c488025a4d82997","kind":"source","name":"metagene-ai/METAGENE-1: config.json","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://huggingface.co/metagene-ai/METAGENE-1/blob/ad8a1e0ee62b85058bfc05d823d8e8d4759edc48/config.json","artifact_url":"https://huggingface.co/metagene-ai/METAGENE-1/blob/ad8a1e0ee62b85058bfc05d823d8e8d4759edc48/config.json","version":"ad8a1e0ee62b85058bfc05d823d8e8d4759edc48","retrieved_at":"2026-09-16T19:46:20.736359+00:00","artifact_sha256":"dc3751f8648b4ab24a3c0c42026267e1a2561b93249f54b460c0374e62d33f98","locator":"config.json","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-f780521ccebbb0691f9e","kind":"source","name":"jwohlwend/boltz: README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/jwohlwend/boltz/blob/b1ebfc46ecf57f5414e0d1a6f9027bbb122c53bc/README.md","artifact_url":"https://raw.githubusercontent.com/jwohlwend/boltz/b1ebfc46ecf57f5414e0d1a6f9027bbb122c53bc/README.md","version":"b1ebfc46ecf57f5414e0d1a6f9027bbb122c53bc","retrieved_at":"2026-09-16T19:46:19.465940+00:00","artifact_sha256":"79149435313841e6f9cc0c8b581259f3952a2469cbe6f9a733f7143b557025c2","locator":"README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-fa53ab621979fb5f37c8","kind":"source","name":"facebook/esmfold_v1: README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://huggingface.co/facebook/esmfold_v1/blob/75a3841ee059df2bf4d56688166c8fb459ddd97a/README.md","artifact_url":"https://huggingface.co/facebook/esmfold_v1/blob/75a3841ee059df2bf4d56688166c8fb459ddd97a/README.md","version":"75a3841ee059df2bf4d56688166c8fb459ddd97a","retrieved_at":"2026-09-16T20:04:02.231029+00:00","artifact_sha256":"12c535c3711211dcd3947ff8164d2e1428ca3461ef8a631ea7b6419673322fe2","locator":"README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-fa7a6c37148223466dee","kind":"source","name":"gcorso/DiffDock: README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/gcorso/DiffDock/blob/85c49b60d3e0b0182a59ee43a34a6d7036981284/README.md","artifact_url":"https://raw.githubusercontent.com/gcorso/DiffDock/85c49b60d3e0b0182a59ee43a34a6d7036981284/README.md","version":"85c49b60d3e0b0182a59ee43a34a6d7036981284","retrieved_at":"2026-09-16T19:46:19.329065+00:00","artifact_sha256":"6f63088d85b5f05d58416ede319387c1b7f3661b27741a36314ada861f2056de","locator":"README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-fa91833592f376f6fb51","kind":"source","name":"RosettaCommons/RFdiffusion: LICENSE","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/RosettaCommons/RFdiffusion/blob/86507b6538f51fce57b5a72477165f03999ed7ae/LICENSE","artifact_url":"https://raw.githubusercontent.com/RosettaCommons/RFdiffusion/86507b6538f51fce57b5a72477165f03999ed7ae/LICENSE","version":"86507b6538f51fce57b5a72477165f03999ed7ae","retrieved_at":"2026-09-16T19:46:18.126662+00:00","artifact_sha256":"eefc2ae77cb92b1414a6ac76b246642fae2189747b1b07747e92e5ddc696ec24","locator":"LICENSE","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-fadad0bf451a225696d1","kind":"source","name":"https://fiehnlab.ucdavis.edu/projects/LipidBlast/: page.html","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://fiehnlab.ucdavis.edu/projects/LipidBlast/","artifact_url":"https://fiehnlab.ucdavis.edu/projects/LipidBlast/","version":"Retrieved page snapshot; no immutable publisher revision supplied","retrieved_at":"2026-09-16T19:46:17.765823+00:00","artifact_sha256":"584a7adeeb0f7ac33ca6f7434aaa16c945d2ababb53ea38b06e6cadde559880d","locator":"page.html","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-fd8e332abdf04a75195b","kind":"source","name":"ml4bio/RNA-FM: README.md","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/ml4bio/RNA-FM/blob/348951516e0963d22bbb33b3c9fc18c89081d38e/README.md","artifact_url":"https://raw.githubusercontent.com/ml4bio/RNA-FM/348951516e0963d22bbb33b3c9fc18c89081d38e/README.md","version":"348951516e0963d22bbb33b3c9fc18c89081d38e","retrieved_at":"2026-09-16T19:46:19.769679+00:00","artifact_sha256":"f9f1c1d62adc471661ca98b30c0250e9f3ce0cff7433830f149f5f48ea41c3da","locator":"README.md","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-fe7d7d8a007d344f589c","kind":"source","name":"calico/basenji: LICENSE","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/calico/basenji/blob/06ce5d387e20b47184d05433b3983163c5f923cd/LICENSE","artifact_url":"https://raw.githubusercontent.com/calico/basenji/06ce5d387e20b47184d05433b3983163c5f923cd/LICENSE","version":"06ce5d387e20b47184d05433b3983163c5f923cd","retrieved_at":"2026-09-16T19:46:18.649671+00:00","artifact_sha256":"3760a0dba24568a5b4a4e41e8aabe5e5e23d81422f26e7623c8ba5cf13ef6c05","locator":"LICENSE","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-official-fffd4026cb8f4018fe2e","kind":"source","name":"diffdockl: Primary paper PDF","description":"Primary-source artifact retrieved for field-level model-profile review. Source inspection is not independent model reproduction.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://arxiv.org/pdf/2402.18396","artifact_url":"https://arxiv.org/pdf/2402.18396","version":"2402.18396v1","retrieved_at":"2026-09-16T20:04:04.777388+00:00","artifact_sha256":"c8022ecbf5a2e628e42a03302a5595390020d40d65cc630490c5803ff6c14170","locator":"Primary paper PDF","artifact_format":"original_artifact","hash_scope":"SHA-256 of retrieved original artifact bytes","retrieval_note":"Direct source retrieval","review_method":"automated_source_review","review_scope":"Only cited profile claims; no numerical performance result added or independently reproduced."}}
{"id":"evidence-reported-2ome-lm-2025-readme-md","kind":"source","name":"CSUBioGroup/2OMe-LM README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"fc90c421ffa9766c1981d6ea15eb73a9a2049f99858e1e3bef99f11263dee2bd","artifact_url":"https://raw.githubusercontent.com/CSUBioGroup/2OMe-LM/2e22439723777b5bacdce72cdcd7cfbde9e88cd1/README.md","retrieved_at":"2026-09-16T19:54:12.011441+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/CSUBioGroup/2OMe-LM/blob/2e22439723777b5bacdce72cdcd7cfbde9e88cd1/README.md","version":"2e22439723777b5bacdce72cdcd7cfbde9e88cd1"}}
{"id":"evidence-reported-adar-gpt-editing-2026-license","kind":"source","name":"Scientific-Computing-Lab/ADAR-GPT LICENSE","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"02f64bae2cb1b1025702d2224fc7c0c762bdcfe58c9f83ec2d86f150f2f0fc3d","artifact_url":"https://raw.githubusercontent.com/Scientific-Computing-Lab/ADAR-GPT/c0fd23679922d91a45520455d4ca0202a5ca609f/LICENSE","retrieved_at":"2026-09-16T19:54:12.011620+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/Scientific-Computing-Lab/ADAR-GPT/blob/c0fd23679922d91a45520455d4ca0202a5ca609f/LICENSE","version":"c0fd23679922d91a45520455d4ca0202a5ca609f"}}
{"id":"evidence-reported-adar-gpt-editing-2026-readme-md","kind":"source","name":"Scientific-Computing-Lab/ADAR-GPT README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"ae16bae98ce731633c412a0279285112ebe40db0f867907bd0caa91fce66ded8","artifact_url":"https://raw.githubusercontent.com/Scientific-Computing-Lab/ADAR-GPT/c0fd23679922d91a45520455d4ca0202a5ca609f/README.md","retrieved_at":"2026-09-16T19:54:12.011620+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/Scientific-Computing-Lab/ADAR-GPT/blob/c0fd23679922d91a45520455d4ca0202a5ca609f/README.md","version":"c0fd23679922d91a45520455d4ca0202a5ca609f"}}
{"id":"evidence-reported-arsenal-regulatory-dna-2026-readme-md","kind":"source","name":"kundajelab/regulatory_lm README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"cae8dd0c9c9ba367844158a6474a961e7184633c6376c4fae81610054827218f","artifact_url":"https://raw.githubusercontent.com/kundajelab/regulatory_lm/2c4594f5b7649df3f1372238916f6d95bd31ba1d/README.md","retrieved_at":"2026-09-16T19:54:12.011707+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/kundajelab/regulatory_lm/blob/2c4594f5b7649df3f1372238916f6d95bd31ba1d/README.md","version":"2c4594f5b7649df3f1372238916f6d95bd31ba1d"}}
{"id":"evidence-reported-barcodebert-2026-license","kind":"source","name":"bioscan-ml/BarcodeBERT LICENSE","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"6d98a262a5a27bc0671d802e51b9bd088a1020f64350a449e899533d153240a3","artifact_url":"https://raw.githubusercontent.com/bioscan-ml/BarcodeBERT/00e492374eb748ed0f034a3a5981ab4eeffd92cc/LICENSE","retrieved_at":"2026-09-16T19:54:12.012721+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/bioscan-ml/BarcodeBERT/blob/00e492374eb748ed0f034a3a5981ab4eeffd92cc/LICENSE","version":"00e492374eb748ed0f034a3a5981ab4eeffd92cc"}}
{"id":"evidence-reported-barcodebert-2026-readme-md","kind":"source","name":"bioscan-ml/BarcodeBERT README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"540be344aa41f5b2740bed1819ccebdba6ce9681b7434c7fca02c6c1ce5ad360","artifact_url":"https://raw.githubusercontent.com/bioscan-ml/BarcodeBERT/00e492374eb748ed0f034a3a5981ab4eeffd92cc/README.md","retrieved_at":"2026-09-16T19:54:12.012721+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/bioscan-ml/BarcodeBERT/blob/00e492374eb748ed0f034a3a5981ab4eeffd92cc/README.md","version":"00e492374eb748ed0f034a3a5981ab4eeffd92cc"}}
{"id":"evidence-reported-base-boltz-license","kind":"source","name":"jwohlwend/boltz LICENSE","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"f0667fd5e66c51e1ba8ddaa0249c6d7225b30037e02c45782d8f2c2943ac2617","artifact_url":"https://raw.githubusercontent.com/jwohlwend/boltz/b1ebfc46ecf57f5414e0d1a6f9027bbb122c53bc/LICENSE","retrieved_at":"2026-09-16T20:30:16.575839+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/jwohlwend/boltz/blob/b1ebfc46ecf57f5414e0d1a6f9027bbb122c53bc/LICENSE","version":"b1ebfc46ecf57f5414e0d1a6f9027bbb122c53bc"}}
{"id":"evidence-reported-base-boltz-readme-md","kind":"source","name":"jwohlwend/boltz README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"79149435313841e6f9cc0c8b581259f3952a2469cbe6f9a733f7143b557025c2","artifact_url":"https://raw.githubusercontent.com/jwohlwend/boltz/b1ebfc46ecf57f5414e0d1a6f9027bbb122c53bc/README.md","retrieved_at":"2026-09-16T20:30:16.575839+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/jwohlwend/boltz/blob/b1ebfc46ecf57f5414e0d1a6f9027bbb122c53bc/README.md","version":"b1ebfc46ecf57f5414e0d1a6f9027bbb122c53bc"}}
{"id":"evidence-reported-base-caduceus-license","kind":"source","name":"kuleshov-group/caduceus LICENSE","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"53c3bce42b068bd4c9a7831e18d4d7e7eab1b9cd00b8a3faac0aa96793c99bc5","artifact_url":"https://raw.githubusercontent.com/kuleshov-group/caduceus/0060a6d8079b6a040fc55d505e15972a327b70a6/LICENSE","retrieved_at":"2026-09-16T20:00:02.715892+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/kuleshov-group/caduceus/blob/0060a6d8079b6a040fc55d505e15972a327b70a6/LICENSE","version":"0060a6d8079b6a040fc55d505e15972a327b70a6"}}
{"id":"evidence-reported-base-caduceus-readme-md","kind":"source","name":"kuleshov-group/caduceus README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"e508e5199d0cfb9c36dbd503cdccf50d734f419fa8cf1f890dec0db7e741ad42","artifact_url":"https://raw.githubusercontent.com/kuleshov-group/caduceus/0060a6d8079b6a040fc55d505e15972a327b70a6/README.md","retrieved_at":"2026-09-16T20:00:02.715892+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/kuleshov-group/caduceus/blob/0060a6d8079b6a040fc55d505e15972a327b70a6/README.md","version":"0060a6d8079b6a040fc55d505e15972a327b70a6"}}
{"id":"evidence-reported-base-chai-license","kind":"source","name":"github.com/chaidiscovery/chai-lab LICENSE","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"511edf51c5c6f47bae9ae19c59d98666a682e0ccc98e90c2de2a9a897c44c003","artifact_url":"https://raw.githubusercontent.com/chaidiscovery/chai-lab/66c38d1fe5c6756a89ff8596b1dea87d305ec06f/LICENSE","retrieved_at":"2026-09-16T19:46:18.824295+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/chaidiscovery/chai-lab/blob/66c38d1fe5c6756a89ff8596b1dea87d305ec06f/LICENSE","version":"66c38d1fe5c6756a89ff8596b1dea87d305ec06f"}}
{"id":"evidence-reported-base-chai-readme-md","kind":"source","name":"github.com/chaidiscovery/chai-lab README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"ea6f6e64f6fc73d0e3dcbe6755c2aab226fa13bad279d1213acc0217f3f5013f","artifact_url":"https://raw.githubusercontent.com/chaidiscovery/chai-lab/66c38d1fe5c6756a89ff8596b1dea87d305ec06f/README.md","retrieved_at":"2026-09-16T19:46:18.824295+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/chaidiscovery/chai-lab/blob/66c38d1fe5c6756a89ff8596b1dea87d305ec06f/README.md","version":"66c38d1fe5c6756a89ff8596b1dea87d305ec06f"}}
{"id":"evidence-reported-base-diffdock-license","kind":"source","name":"gcorso/DiffDock LICENSE","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"8efe3b8dac5c278ba290a66887e46261056f71dec75c2815c7da75451856cf1f","artifact_url":"https://raw.githubusercontent.com/gcorso/DiffDock/85c49b60d3e0b0182a59ee43a34a6d7036981284/LICENSE","retrieved_at":"2026-09-16T20:00:00.818010+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/gcorso/DiffDock/blob/85c49b60d3e0b0182a59ee43a34a6d7036981284/LICENSE","version":"85c49b60d3e0b0182a59ee43a34a6d7036981284"}}
{"id":"evidence-reported-base-diffdock-readme-md","kind":"source","name":"gcorso/DiffDock README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"6f63088d85b5f05d58416ede319387c1b7f3661b27741a36314ada861f2056de","artifact_url":"https://raw.githubusercontent.com/gcorso/DiffDock/85c49b60d3e0b0182a59ee43a34a6d7036981284/README.md","retrieved_at":"2026-09-16T20:00:00.818010+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/gcorso/DiffDock/blob/85c49b60d3e0b0182a59ee43a34a6d7036981284/README.md","version":"85c49b60d3e0b0182a59ee43a34a6d7036981284"}}
{"id":"evidence-reported-base-dnabert2-card-license","kind":"source","name":"huggingface.co/zhihan1996/DNABERT-2-117M LICENSE","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4","artifact_url":"https://huggingface.co/zhihan1996/DNABERT-2-117M/blob/7bce263b15377fc15361f52cfab88f8b586abda0/LICENSE","retrieved_at":"2026-09-16T19:46:20.874956+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://huggingface.co/zhihan1996/DNABERT-2-117M/blob/7bce263b15377fc15361f52cfab88f8b586abda0/LICENSE","version":"7bce263b15377fc15361f52cfab88f8b586abda0"}}
{"id":"evidence-reported-base-dnabert2-license","kind":"source","name":"MAGICS-LAB/DNABERT_2 LICENSE","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4","artifact_url":"https://raw.githubusercontent.com/MAGICS-LAB/DNABERT_2/f25bed9ee20db966dff39e5c1571249d04e36404/LICENSE","retrieved_at":"2026-09-16T20:00:02.624261+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/MAGICS-LAB/DNABERT_2/blob/f25bed9ee20db966dff39e5c1571249d04e36404/LICENSE","version":"f25bed9ee20db966dff39e5c1571249d04e36404"}}
{"id":"evidence-reported-base-dnabert2-readme-md","kind":"source","name":"MAGICS-LAB/DNABERT_2 README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"734a8cec5f667d74d421bf3b273ad7e256216109636da45aa7ceba21cd34de16","artifact_url":"https://raw.githubusercontent.com/MAGICS-LAB/DNABERT_2/f25bed9ee20db966dff39e5c1571249d04e36404/README.md","retrieved_at":"2026-09-16T20:00:02.624261+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/MAGICS-LAB/DNABERT_2/blob/f25bed9ee20db966dff39e5c1571249d04e36404/README.md","version":"f25bed9ee20db966dff39e5c1571249d04e36404"}}
{"id":"evidence-reported-base-esm-license","kind":"source","name":"facebookresearch/esm LICENSE","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"da6d3703ed11cbe42bd212c725957c98da23cbff1998c05fa4b3d976d1a58e93","artifact_url":"https://raw.githubusercontent.com/facebookresearch/esm/2b369911bb5b4b0dda914521b9475cad1656b2ac/LICENSE","retrieved_at":"2026-09-16T20:00:00.816433+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/facebookresearch/esm/blob/2b369911bb5b4b0dda914521b9475cad1656b2ac/LICENSE","version":"2b369911bb5b4b0dda914521b9475cad1656b2ac"}}
{"id":"evidence-reported-base-esm-readme-md","kind":"source","name":"facebookresearch/esm README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"8b273c21a322fc9473d1b68d0dd40c8166ab2f89e4a190aa26ca87251b97cba9","artifact_url":"https://raw.githubusercontent.com/facebookresearch/esm/2b369911bb5b4b0dda914521b9475cad1656b2ac/README.md","retrieved_at":"2026-09-16T20:00:00.816433+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/facebookresearch/esm/blob/2b369911bb5b4b0dda914521b9475cad1656b2ac/README.md","version":"2b369911bb5b4b0dda914521b9475cad1656b2ac"}}
{"id":"evidence-reported-base-evo2-license","kind":"source","name":"ArcInstitute/evo2 LICENSE","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"5bb5812fc2bfb2d777fe5621767172f8ef30be62ad719c87e0f10832336a99e0","artifact_url":"https://raw.githubusercontent.com/ArcInstitute/evo2/53f195997257c56c00e5ef8d33a54f5baad143a6/LICENSE","retrieved_at":"2026-09-16T20:00:02.431871+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/ArcInstitute/evo2/blob/53f195997257c56c00e5ef8d33a54f5baad143a6/LICENSE","version":"53f195997257c56c00e5ef8d33a54f5baad143a6"}}
{"id":"evidence-reported-base-evo2-readme-md","kind":"source","name":"ArcInstitute/evo2 README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"58787c8ef5cb4fba4c04322a4ceb9f174e2233ec22d4193622fb6bc67d651d89","artifact_url":"https://raw.githubusercontent.com/ArcInstitute/evo2/53f195997257c56c00e5ef8d33a54f5baad143a6/README.md","retrieved_at":"2026-09-16T20:00:02.431871+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/ArcInstitute/evo2/blob/53f195997257c56c00e5ef8d33a54f5baad143a6/README.md","version":"53f195997257c56c00e5ef8d33a54f5baad143a6"}}
{"id":"evidence-reported-base-geneformer-readme-md","kind":"source","name":"huggingface.co/ctheodoris/Geneformer README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"56d6e570b349cbedae9a54634421c94e7af8ea467ce0fdd79193372ae3cbdbd8","artifact_url":"https://huggingface.co/ctheodoris/Geneformer/blob/1f7fbae4e469a5f4f1af8c111a529cfe1b3829f5/README.md","retrieved_at":"2026-09-16T19:46:20.640731+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://huggingface.co/ctheodoris/Geneformer/blob/1f7fbae4e469a5f4f1af8c111a529cfe1b3829f5/README.md","version":"1f7fbae4e469a5f4f1af8c111a529cfe1b3829f5"}}
{"id":"evidence-reported-base-genomad-license","kind":"source","name":"apcamargo/genomad LICENSE","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"2878b1cb109d8dde17ea2607f6def6e3225b003caccf1dba98ad7233e331db87","artifact_url":"https://raw.githubusercontent.com/apcamargo/genomad/8c5fd0d1722d458a3e8ff50278cdc00ed4a514fc/LICENSE","retrieved_at":"2026-09-16T20:00:02.618496+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/apcamargo/genomad/blob/8c5fd0d1722d458a3e8ff50278cdc00ed4a514fc/LICENSE","version":"8c5fd0d1722d458a3e8ff50278cdc00ed4a514fc"}}
{"id":"evidence-reported-base-genomad-readme-md","kind":"source","name":"apcamargo/genomad README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"0d7ab49689a8b402b2e081cecf38e6fe2fb44bccbbd2b8bd61bb08aaa53720af","artifact_url":"https://raw.githubusercontent.com/apcamargo/genomad/8c5fd0d1722d458a3e8ff50278cdc00ed4a514fc/README.md","retrieved_at":"2026-09-16T20:00:02.618496+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/apcamargo/genomad/blob/8c5fd0d1722d458a3e8ff50278cdc00ed4a514fc/README.md","version":"8c5fd0d1722d458a3e8ff50278cdc00ed4a514fc"}}
{"id":"evidence-reported-base-gtdbtk-license","kind":"source","name":"Ecogenomics/GTDBTk LICENSE","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"12ac5047f2af0522f06798b1589ffc4599bc29c91f954d7874e0320634e777c0","artifact_url":"https://raw.githubusercontent.com/Ecogenomics/GTDBTk/f17decef1f9d9cf5b4d31fd21f5c9d32d813abdc/LICENSE","retrieved_at":"2026-09-16T20:00:03.977720+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/Ecogenomics/GTDBTk/blob/f17decef1f9d9cf5b4d31fd21f5c9d32d813abdc/LICENSE","version":"f17decef1f9d9cf5b4d31fd21f5c9d32d813abdc"}}
{"id":"evidence-reported-base-gtdbtk-readme-md","kind":"source","name":"Ecogenomics/GTDBTk README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"72124a5fb40dc02379ca775daae2bf894bc9e1b240a0a31d65e6c6c5c99e268c","artifact_url":"https://raw.githubusercontent.com/Ecogenomics/GTDBTk/f17decef1f9d9cf5b4d31fd21f5c9d32d813abdc/README.md","retrieved_at":"2026-09-16T20:00:03.977720+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/Ecogenomics/GTDBTk/blob/f17decef1f9d9cf5b4d31fd21f5c9d32d813abdc/README.md","version":"f17decef1f9d9cf5b4d31fd21f5c9d32d813abdc"}}
{"id":"evidence-reported-base-hyenadna-license","kind":"source","name":"HazyResearch/hyena-dna LICENSE","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4","artifact_url":"https://raw.githubusercontent.com/HazyResearch/hyena-dna/d553021b483b82980aa4b868b37ec2d4332e198a/LICENSE","retrieved_at":"2026-09-16T20:00:02.742030+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/HazyResearch/hyena-dna/blob/d553021b483b82980aa4b868b37ec2d4332e198a/LICENSE","version":"d553021b483b82980aa4b868b37ec2d4332e198a"}}
{"id":"evidence-reported-base-hyenadna-readme-md","kind":"source","name":"HazyResearch/hyena-dna README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"2b551789c76b8a552c1d989b194ef72a4cd83790d06bc2a8ac2af6ebee92e8db","artifact_url":"https://raw.githubusercontent.com/HazyResearch/hyena-dna/d553021b483b82980aa4b868b37ec2d4332e198a/README.md","retrieved_at":"2026-09-16T20:00:02.742030+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/HazyResearch/hyena-dna/blob/d553021b483b82980aa4b868b37ec2d4332e198a/README.md","version":"d553021b483b82980aa4b868b37ec2d4332e198a"}}
{"id":"evidence-reported-base-kraken2-license","kind":"source","name":"DerrickWood/kraken2 LICENSE","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"ef3803fed10bf0eae6919db5e12204af4460e73234953373b22bc6de04ed840a","artifact_url":"https://raw.githubusercontent.com/DerrickWood/kraken2/8c190b1b668825935dbf6dee5f969227dc8269bb/LICENSE","retrieved_at":"2026-09-16T20:00:00.821204+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/DerrickWood/kraken2/blob/8c190b1b668825935dbf6dee5f969227dc8269bb/LICENSE","version":"8c190b1b668825935dbf6dee5f969227dc8269bb"}}
{"id":"evidence-reported-base-kraken2-readme-md","kind":"source","name":"DerrickWood/kraken2 README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"2ea33af266b4268a55fd750d0f3265cd3165d61f6be375c5ea3b3ff5c58c7c8c","artifact_url":"https://raw.githubusercontent.com/DerrickWood/kraken2/8c190b1b668825935dbf6dee5f969227dc8269bb/README.md","retrieved_at":"2026-09-16T20:00:00.821204+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/DerrickWood/kraken2/blob/8c190b1b668825935dbf6dee5f969227dc8269bb/README.md","version":"8c190b1b668825935dbf6dee5f969227dc8269bb"}}
{"id":"evidence-reported-base-metaphlan-license-txt","kind":"source","name":"biobakery/MetaPhlAn license.txt","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"ecf18c2928e49997ba1f098f0f0da0958d257d69fffb48875162dbd24f3d7762","artifact_url":"https://raw.githubusercontent.com/biobakery/MetaPhlAn/424f3e6e30618266404353e1083c6405a9f02f48/license.txt","retrieved_at":"2026-09-16T20:00:02.332125+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/biobakery/MetaPhlAn/blob/424f3e6e30618266404353e1083c6405a9f02f48/license.txt","version":"424f3e6e30618266404353e1083c6405a9f02f48"}}
{"id":"evidence-reported-base-metaphlan-readme-md","kind":"source","name":"biobakery/MetaPhlAn README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"ce491bb2d686145e0773c685d0d02e8a5fabc7daaea60eef07cf54298561fba7","artifact_url":"https://raw.githubusercontent.com/biobakery/MetaPhlAn/424f3e6e30618266404353e1083c6405a9f02f48/README.md","retrieved_at":"2026-09-16T20:00:02.332125+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/biobakery/MetaPhlAn/blob/424f3e6e30618266404353e1083c6405a9f02f48/README.md","version":"424f3e6e30618266404353e1083c6405a9f02f48"}}
{"id":"evidence-reported-base-pangolin-license","kind":"source","name":"github.com/tkzeng/Pangolin LICENSE","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"3972dc9744f6499f0f9b2dbf76696f2ae7ad8af9b23dde66d6af86c9dfb36986","artifact_url":"https://raw.githubusercontent.com/tkzeng/Pangolin/5cf94b8db938c658391b4305cd7ce33297d44ff7/LICENSE","retrieved_at":"2026-09-16T19:46:20.582143+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/tkzeng/Pangolin/blob/5cf94b8db938c658391b4305cd7ce33297d44ff7/LICENSE","version":"5cf94b8db938c658391b4305cd7ce33297d44ff7"}}
{"id":"evidence-reported-base-pangolin-readme-md","kind":"source","name":"github.com/tkzeng/Pangolin README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"9117f9d255d6b6e810d224a600d381192bccccd3f0cd417fed9559d49ca8fffd","artifact_url":"https://raw.githubusercontent.com/tkzeng/Pangolin/5cf94b8db938c658391b4305cd7ce33297d44ff7/README.md","retrieved_at":"2026-09-16T19:46:20.582143+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/tkzeng/Pangolin/blob/5cf94b8db938c658391b4305cd7ce33297d44ff7/README.md","version":"5cf94b8db938c658391b4305cd7ce33297d44ff7"}}
{"id":"evidence-reported-base-promotech-license","kind":"source","name":"BioinformaticsLabAtMUN/Promotech LICENSE","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"3972dc9744f6499f0f9b2dbf76696f2ae7ad8af9b23dde66d6af86c9dfb36986","artifact_url":"https://raw.githubusercontent.com/BioinformaticsLabAtMUN/Promotech/56251ad9b883ef831b4753fc623d5ec970fe65e0/LICENSE","retrieved_at":"2026-09-16T20:40:28.842295+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/BioinformaticsLabAtMUN/Promotech/blob/56251ad9b883ef831b4753fc623d5ec970fe65e0/LICENSE","version":"56251ad9b883ef831b4753fc623d5ec970fe65e0"}}
{"id":"evidence-reported-base-promotech-readme-md","kind":"source","name":"BioinformaticsLabAtMUN/Promotech README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"51050c0bf5982be985dd87e450ea5c2408a2d87cd32c679fc639246d64e5ca80","artifact_url":"https://raw.githubusercontent.com/BioinformaticsLabAtMUN/Promotech/56251ad9b883ef831b4753fc623d5ec970fe65e0/README.md","retrieved_at":"2026-09-16T20:40:28.842295+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/BioinformaticsLabAtMUN/Promotech/blob/56251ad9b883ef831b4753fc623d5ec970fe65e0/README.md","version":"56251ad9b883ef831b4753fc623d5ec970fe65e0"}}
{"id":"evidence-reported-base-proteinmpnn-license","kind":"source","name":"dauparas/ProteinMPNN LICENSE","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"82009d25ce585631f452b2b24589bdb29c559ccfefa2f200ef312ed5b501a586","artifact_url":"https://raw.githubusercontent.com/dauparas/ProteinMPNN/8907e6671bfbfc92303b5f79c4b5e6ce47cdef57/LICENSE","retrieved_at":"2026-09-16T20:00:02.940818+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/dauparas/ProteinMPNN/blob/8907e6671bfbfc92303b5f79c4b5e6ce47cdef57/LICENSE","version":"8907e6671bfbfc92303b5f79c4b5e6ce47cdef57"}}
{"id":"evidence-reported-base-proteinmpnn-readme-md","kind":"source","name":"dauparas/ProteinMPNN README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"772ebe52d2ba5100a28a888910c6f0c9fd4ded1d1372e3d89f6f1c48707e0365","artifact_url":"https://raw.githubusercontent.com/dauparas/ProteinMPNN/8907e6671bfbfc92303b5f79c4b5e6ce47cdef57/README.md","retrieved_at":"2026-09-16T20:00:02.940818+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/dauparas/ProteinMPNN/blob/8907e6671bfbfc92303b5f79c4b5e6ce47cdef57/README.md","version":"8907e6671bfbfc92303b5f79c4b5e6ce47cdef57"}}
{"id":"evidence-reported-base-rinalmo-license","kind":"source","name":"lbcb-sci/RiNALMo LICENSE","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4","artifact_url":"https://raw.githubusercontent.com/lbcb-sci/RiNALMo/2c2c5c14a5ae609d8c560a5d9ca32e51e0288955/LICENSE","retrieved_at":"2026-09-16T20:00:00.817876+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/lbcb-sci/RiNALMo/blob/2c2c5c14a5ae609d8c560a5d9ca32e51e0288955/LICENSE","version":"2c2c5c14a5ae609d8c560a5d9ca32e51e0288955"}}
{"id":"evidence-reported-base-rinalmo-readme-md","kind":"source","name":"lbcb-sci/RiNALMo README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"de3d60dd5aedc63c4f401a92876c2d1e053011e008773ce1be38511360463b63","artifact_url":"https://raw.githubusercontent.com/lbcb-sci/RiNALMo/2c2c5c14a5ae609d8c560a5d9ca32e51e0288955/README.md","retrieved_at":"2026-09-16T20:00:00.817876+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/lbcb-sci/RiNALMo/blob/2c2c5c14a5ae609d8c560a5d9ca32e51e0288955/README.md","version":"2c2c5c14a5ae609d8c560a5d9ca32e51e0288955"}}
{"id":"evidence-reported-base-rnafm-license","kind":"source","name":"ml4bio/RNA-FM LICENSE","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"b0809e99b532fdf51660f5a3d2a9010ed09d15aef0d131ad80fe80c2291a4fba","artifact_url":"https://raw.githubusercontent.com/ml4bio/RNA-FM/348951516e0963d22bbb33b3c9fc18c89081d38e/LICENSE","retrieved_at":"2026-09-16T20:00:00.817676+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/ml4bio/RNA-FM/blob/348951516e0963d22bbb33b3c9fc18c89081d38e/LICENSE","version":"348951516e0963d22bbb33b3c9fc18c89081d38e"}}
{"id":"evidence-reported-base-rnafm-readme-md","kind":"source","name":"ml4bio/RNA-FM README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"f9f1c1d62adc471661ca98b30c0250e9f3ce0cff7433830f149f5f48ea41c3da","artifact_url":"https://raw.githubusercontent.com/ml4bio/RNA-FM/348951516e0963d22bbb33b3c9fc18c89081d38e/README.md","retrieved_at":"2026-09-16T20:00:00.817676+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/ml4bio/RNA-FM/blob/348951516e0963d22bbb33b3c9fc18c89081d38e/README.md","version":"348951516e0963d22bbb33b3c9fc18c89081d38e"}}
{"id":"evidence-reported-base-rnafold-license-txt","kind":"source","name":"ViennaRNA/ViennaRNA license.txt","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"04ca785cab24944ad8ea6a2ddfea47b91246806c937ee65b5cc33f32c9dd897d","artifact_url":"https://raw.githubusercontent.com/ViennaRNA/ViennaRNA/1ffec79f5e258896160f7362ced8263450f371dc/license.txt","retrieved_at":"2026-09-16T20:00:00.820987+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/ViennaRNA/ViennaRNA/blob/1ffec79f5e258896160f7362ced8263450f371dc/license.txt","version":"1ffec79f5e258896160f7362ced8263450f371dc"}}
{"id":"evidence-reported-base-rnafold-readme-md","kind":"source","name":"ViennaRNA/ViennaRNA README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"d37146b01e4273062a5230c496a9af8414ee7ef14fcd905cf415959256d59c4e","artifact_url":"https://raw.githubusercontent.com/ViennaRNA/ViennaRNA/1ffec79f5e258896160f7362ced8263450f371dc/README.md","retrieved_at":"2026-09-16T20:00:00.820987+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/ViennaRNA/ViennaRNA/blob/1ffec79f5e258896160f7362ced8263450f371dc/README.md","version":"1ffec79f5e258896160f7362ced8263450f371dc"}}
{"id":"evidence-reported-base-scgpt-license","kind":"source","name":"bowang-lab/scGPT LICENSE","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"1ceeacbed51e2890187425547bc2efd16c1b7ad45189b7dfb21e83a45a2e9d9e","artifact_url":"https://raw.githubusercontent.com/bowang-lab/scGPT/cebd6fae655b9c585a4807daa3ac31bb764f06b4/LICENSE","retrieved_at":"2026-09-16T20:00:00.816587+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/bowang-lab/scGPT/blob/cebd6fae655b9c585a4807daa3ac31bb764f06b4/LICENSE","version":"cebd6fae655b9c585a4807daa3ac31bb764f06b4"}}
{"id":"evidence-reported-base-scgpt-readme-md","kind":"source","name":"bowang-lab/scGPT README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"b0503e8ca789f19f1fc2350c5aaf57b1b323bbae43b354655231b5f4a1586c83","artifact_url":"https://raw.githubusercontent.com/bowang-lab/scGPT/cebd6fae655b9c585a4807daa3ac31bb764f06b4/README.md","retrieved_at":"2026-09-16T20:00:00.816587+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/bowang-lab/scGPT/blob/cebd6fae655b9c585a4807daa3ac31bb764f06b4/README.md","version":"cebd6fae655b9c585a4807daa3ac31bb764f06b4"}}
{"id":"evidence-reported-base-scvi-license","kind":"source","name":"scverse/scvi-tools LICENSE","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"66399db0284d2539790efb348886ab0c1f745bbe5fea6ac38a00465a14adc8f5","artifact_url":"https://raw.githubusercontent.com/scverse/scvi-tools/73b28e44223621470e582a81a102c107bb22678b/LICENSE","retrieved_at":"2026-09-16T20:00:02.370441+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/scverse/scvi-tools/blob/73b28e44223621470e582a81a102c107bb22678b/LICENSE","version":"73b28e44223621470e582a81a102c107bb22678b"}}
{"id":"evidence-reported-base-scvi-readme-md","kind":"source","name":"scverse/scvi-tools README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"eb46b8a54e60643ca0cd8cb375ede05b01dcbd2380ca17ce8027a92ba13cebbb","artifact_url":"https://raw.githubusercontent.com/scverse/scvi-tools/73b28e44223621470e582a81a102c107bb22678b/README.md","retrieved_at":"2026-09-16T20:00:02.370441+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/scverse/scvi-tools/blob/73b28e44223621470e582a81a102c107bb22678b/README.md","version":"73b28e44223621470e582a81a102c107bb22678b"}}
{"id":"evidence-reported-base-spliceai-license","kind":"source","name":"github.com/Illumina/SpliceAI LICENSE","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"67a909a0a8f8f7f45152207b6bcf9c78dd8a4dd3c8eef5bd11cd80a72e15344e","artifact_url":"https://raw.githubusercontent.com/Illumina/SpliceAI/03f42437aaf56dc5dfd822c4ccee5aec1a705079/LICENSE","retrieved_at":"2026-09-16T19:46:17.769160+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/Illumina/SpliceAI/blob/03f42437aaf56dc5dfd822c4ccee5aec1a705079/LICENSE","version":"03f42437aaf56dc5dfd822c4ccee5aec1a705079"}}
{"id":"evidence-reported-base-spliceai-readme-md","kind":"source","name":"github.com/Illumina/SpliceAI README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"8e5203afe343100832391e6155c7112f15cfe60bf0c21681d64e3420f854ef4d","artifact_url":"https://raw.githubusercontent.com/Illumina/SpliceAI/03f42437aaf56dc5dfd822c4ccee5aec1a705079/README.md","retrieved_at":"2026-09-16T19:46:17.769160+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/Illumina/SpliceAI/blob/03f42437aaf56dc5dfd822c4ccee5aec1a705079/README.md","version":"03f42437aaf56dc5dfd822c4ccee5aec1a705079"}}
{"id":"evidence-reported-base-ufold-license","kind":"source","name":"uci-cbcl/UFold LICENSE","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"5c93b3a2d31b6e7b964633ff5e81bd9e31ff9ee845996ab3b0d3d71b0d368f2e","artifact_url":"https://raw.githubusercontent.com/uci-cbcl/UFold/75bd9acc83826059682dfca9d3659df66b132cd1/LICENSE","retrieved_at":"2026-09-16T20:00:03.913031+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/uci-cbcl/UFold/blob/75bd9acc83826059682dfca9d3659df66b132cd1/LICENSE","version":"75bd9acc83826059682dfca9d3659df66b132cd1"}}
{"id":"evidence-reported-base-ufold-readme-md","kind":"source","name":"uci-cbcl/UFold README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"9ef66e737e1047f7028f8f636516f5d0144f9395fe5aceb1a387d1ba457c81f2","artifact_url":"https://raw.githubusercontent.com/uci-cbcl/UFold/75bd9acc83826059682dfca9d3659df66b132cd1/README.md","retrieved_at":"2026-09-16T20:00:03.913031+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/uci-cbcl/UFold/blob/75bd9acc83826059682dfca9d3659df66b132cd1/README.md","version":"75bd9acc83826059682dfca9d3659df66b132cd1"}}
{"id":"evidence-reported-base-vina-license","kind":"source","name":"ccsb-scripps/AutoDock-Vina LICENSE","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"cfc7749b96f63bd31c3c42b5c471bf756814053e847c10f3eb003417bc523d30","artifact_url":"https://raw.githubusercontent.com/ccsb-scripps/AutoDock-Vina/3c65c0b3e6c2c1d183f6a175ecb65e3c5ba91645/LICENSE","retrieved_at":"2026-09-16T20:00:01.500953+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/ccsb-scripps/AutoDock-Vina/blob/3c65c0b3e6c2c1d183f6a175ecb65e3c5ba91645/LICENSE","version":"3c65c0b3e6c2c1d183f6a175ecb65e3c5ba91645"}}
{"id":"evidence-reported-base-vina-readme-md","kind":"source","name":"ccsb-scripps/AutoDock-Vina README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"4f1728521ab79de1c33e1cf8605b31037effed5de2a2fbbccba58d7b0a005ae7","artifact_url":"https://raw.githubusercontent.com/ccsb-scripps/AutoDock-Vina/3c65c0b3e6c2c1d183f6a175ecb65e3c5ba91645/README.md","retrieved_at":"2026-09-16T20:00:01.500953+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/ccsb-scripps/AutoDock-Vina/blob/3c65c0b3e6c2c1d183f6a175ecb65e3c5ba91645/README.md","version":"3c65c0b3e6c2c1d183f6a175ecb65e3c5ba91645"}}
{"id":"evidence-reported-birna-bert-2025-readme-md","kind":"source","name":"buetnlpbio/BiRNA-BERT README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"00f3835cae38a2307d8095fc5240496e01a9a981c8450230f9ae5722153ac2cc","artifact_url":"https://raw.githubusercontent.com/buetnlpbio/BiRNA-BERT/14dc86b1b44c266f01025fc425103f2878646b39/README.md","retrieved_at":"2026-09-16T19:54:12.014900+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/buetnlpbio/BiRNA-BERT/blob/14dc86b1b44c266f01025fc425103f2878646b39/README.md","version":"14dc86b1b44c266f01025fc425103f2878646b39"}}
{"id":"evidence-reported-bpfold-2025-license","kind":"source","name":"heqin-zhu/BPfold LICENSE","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"dd82fc4d63d005354cb0f1eab0d3f80248586dcfdaac4286c700787ca28cf141","artifact_url":"https://raw.githubusercontent.com/heqin-zhu/BPfold/d37d6aa10cbca13e590ff83917fc4d63fec2ddbc/LICENSE","retrieved_at":"2026-09-16T19:54:12.015212+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/heqin-zhu/BPfold/blob/d37d6aa10cbca13e590ff83917fc4d63fec2ddbc/LICENSE","version":"d37d6aa10cbca13e590ff83917fc4d63fec2ddbc"}}
{"id":"evidence-reported-bpfold-2025-readme-md","kind":"source","name":"heqin-zhu/BPfold README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"e3d23006a2a2e4524208bdc4561d3091aeaad1f08c1e79ce2b23549b0ffbdd51","artifact_url":"https://raw.githubusercontent.com/heqin-zhu/BPfold/d37d6aa10cbca13e590ff83917fc4d63fec2ddbc/README.md","retrieved_at":"2026-09-16T19:54:12.015212+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/heqin-zhu/BPfold/blob/d37d6aa10cbca13e590ff83917fc4d63fec2ddbc/README.md","version":"d37d6aa10cbca13e590ff83917fc4d63fec2ddbc"}}
{"id":"evidence-reported-bpfold-supplement","kind":"source","name":"BPfold supplementary information","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_member":"41467_2025_60048_MOESM1_ESM.pdf","artifact_sha256":"254d4842e4690d21a565833aa0e27c7d45cc1ef926dd52527596cc171be91e64","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12216785/supplementaryFiles","retrieved_at":"2026-09-16T20:39:46.798664+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12216785/supplementaryFiles","version":"s41467-025-60048-1 published supplement"}}
{"id":"evidence-reported-cammiq-2022-license","kind":"source","name":"algo-cancer/CAMMiQ LICENSE","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"92dd25a63adf6299964d45946d39a1d6e47a12957b8c60570b94df58510aa535","artifact_url":"https://raw.githubusercontent.com/algo-cancer/CAMMiQ/6142150d427a74cc21a5ee4d8b37a3b78884f163/LICENSE","retrieved_at":"2026-09-16T19:54:12.016679+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/algo-cancer/CAMMiQ/blob/6142150d427a74cc21a5ee4d8b37a3b78884f163/LICENSE","version":"6142150d427a74cc21a5ee4d8b37a3b78884f163"}}
{"id":"evidence-reported-cammiq-2022-readme-md","kind":"source","name":"algo-cancer/CAMMiQ README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"94b01aa6c1d8657fe6d6c77367aeb70991557d6dcd33e8229acc931bf5ce6b3c","artifact_url":"https://raw.githubusercontent.com/algo-cancer/CAMMiQ/6142150d427a74cc21a5ee4d8b37a3b78884f163/README.md","retrieved_at":"2026-09-16T19:54:12.016679+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/algo-cancer/CAMMiQ/blob/6142150d427a74cc21a5ee4d8b37a3b78884f163/README.md","version":"6142150d427a74cc21a5ee4d8b37a3b78884f163"}}
{"id":"evidence-reported-cathe2-2025-license","kind":"source","name":"Mouret-Orfeu/CATHe2 LICENSE","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"71ecefd54a74530222dbfa19f602ca441c4042183cec725a4c4db44b556eaef0","artifact_url":"https://raw.githubusercontent.com/Mouret-Orfeu/CATHe2/cdaf5f4c8d9dfda86f78fd0efc83ab22a30e9ffd/LICENSE","retrieved_at":"2026-09-16T20:30:16.571174+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/Mouret-Orfeu/CATHe2/blob/cdaf5f4c8d9dfda86f78fd0efc83ab22a30e9ffd/LICENSE","version":"cdaf5f4c8d9dfda86f78fd0efc83ab22a30e9ffd"}}
{"id":"evidence-reported-cathe2-2025-readme-md","kind":"source","name":"Mouret-Orfeu/CATHe2 README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"8e7bd5bd27851c9b96bb7edadeee6dac9ab5cc1913e831f2234055ccb0b5f779","artifact_url":"https://raw.githubusercontent.com/Mouret-Orfeu/CATHe2/cdaf5f4c8d9dfda86f78fd0efc83ab22a30e9ffd/README.md","retrieved_at":"2026-09-16T20:30:16.571174+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/Mouret-Orfeu/CATHe2/blob/cdaf5f4c8d9dfda86f78fd0efc83ab22a30e9ffd/README.md","version":"cdaf5f4c8d9dfda86f78fd0efc83ab22a30e9ffd"}}
{"id":"evidence-reported-cell-dino-2025-docs-readme-cell-dino-md","kind":"source","name":"facebookresearch/dinov2 docs/README_CELL_DINO.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"b9ec5771aaa41d571f5c2b54300a7f8bf7ab5d75f7679d2a80f260247a71d42f","artifact_url":"https://raw.githubusercontent.com/facebookresearch/dinov2/7764ea0f912e53c92e82eb78a2a1631e92725fc8/docs/README_CELL_DINO.md","retrieved_at":"2026-09-16T20:30:16.572226+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/facebookresearch/dinov2/blob/7764ea0f912e53c92e82eb78a2a1631e92725fc8/docs/README_CELL_DINO.md","version":"7764ea0f912e53c92e82eb78a2a1631e92725fc8"}}
{"id":"evidence-reported-cell-dino-2025-license-cell-dino-code","kind":"source","name":"facebookresearch/dinov2 LICENSE_CELL_DINO_CODE","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"fe7b4ce83b8381cc5b216bbb4af73c570688d1b819c73bbaed8ca401f4677cd6","artifact_url":"https://raw.githubusercontent.com/facebookresearch/dinov2/7764ea0f912e53c92e82eb78a2a1631e92725fc8/LICENSE_CELL_DINO_CODE","retrieved_at":"2026-09-16T20:30:16.572226+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/facebookresearch/dinov2/blob/7764ea0f912e53c92e82eb78a2a1631e92725fc8/LICENSE_CELL_DINO_CODE","version":"7764ea0f912e53c92e82eb78a2a1631e92725fc8"}}
{"id":"evidence-reported-cell-dino-2025-license-cell-dino-models","kind":"source","name":"facebookresearch/dinov2 LICENSE_CELL_DINO_MODELS","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"d779b47a4ef8bbfc6c90d14768fe6a0a2c1c08bbc3712084496058bfd83cef4f","artifact_url":"https://raw.githubusercontent.com/facebookresearch/dinov2/7764ea0f912e53c92e82eb78a2a1631e92725fc8/LICENSE_CELL_DINO_MODELS","retrieved_at":"2026-09-16T20:30:16.572226+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/facebookresearch/dinov2/blob/7764ea0f912e53c92e82eb78a2a1631e92725fc8/LICENSE_CELL_DINO_MODELS","version":"7764ea0f912e53c92e82eb78a2a1631e92725fc8"}}
{"id":"evidence-reported-cell2sentence-2024-license","kind":"source","name":"vandijklab/cell2sentence LICENSE","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"cbb25571ab70c6ebeebb0068243283a887a9353808543d535fedcd882be30fd6","artifact_url":"https://raw.githubusercontent.com/vandijklab/cell2sentence/a6efaf079f98491d4723ced44b929936b94368aa/LICENSE","retrieved_at":"2026-09-16T20:42:56.481821+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/vandijklab/cell2sentence/blob/a6efaf079f98491d4723ced44b929936b94368aa/LICENSE","version":"a6efaf079f98491d4723ced44b929936b94368aa"}}
{"id":"evidence-reported-cell2sentence-2024-readme-md","kind":"source","name":"vandijklab/cell2sentence README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"60b183e5311a46eff139d728d860875874ff255242bcd7a237789b893ffc262c","artifact_url":"https://raw.githubusercontent.com/vandijklab/cell2sentence/a6efaf079f98491d4723ced44b929936b94368aa/README.md","retrieved_at":"2026-09-16T20:42:56.481821+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/vandijklab/cell2sentence/blob/a6efaf079f98491d4723ced44b929936b94368aa/README.md","version":"a6efaf079f98491d4723ced44b929936b94368aa"}}
{"id":"evidence-reported-cobra-rna-binding-2026-license","kind":"source","name":"kucm-lsbi/CoBRA LICENSE","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"58dbf5d5eba9fda0cb32f823379e9a5ed549a31308f36bf72fa33c747e0fbcb9","artifact_url":"https://raw.githubusercontent.com/kucm-lsbi/CoBRA/415fd05cabf990f28a46cc2ba651531a28f7d249/LICENSE","retrieved_at":"2026-09-16T19:54:13.479891+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/kucm-lsbi/CoBRA/blob/415fd05cabf990f28a46cc2ba651531a28f7d249/LICENSE","version":"415fd05cabf990f28a46cc2ba651531a28f7d249"}}
{"id":"evidence-reported-cobra-rna-binding-2026-readme-md","kind":"source","name":"kucm-lsbi/CoBRA README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"9ad9fea6b42049ca4ca3f71dd906e6d1289ef842aee14bb4241e880866003bc9","artifact_url":"https://raw.githubusercontent.com/kucm-lsbi/CoBRA/415fd05cabf990f28a46cc2ba651531a28f7d249/README.md","retrieved_at":"2026-09-16T19:54:13.479891+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/kucm-lsbi/CoBRA/blob/415fd05cabf990f28a46cc2ba651531a28f7d249/README.md","version":"415fd05cabf990f28a46cc2ba651531a28f7d249"}}
{"id":"evidence-reported-codonbert-vaccines-2024-readme-md","kind":"source","name":"Sanofi-Public/CodonBert README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"9efbe98650ca68f2d4797776df178dce59489fdac6beb6fa7d8390f74dfe9eb1","artifact_url":"https://raw.githubusercontent.com/Sanofi-Public/CodonBert/451a1b167c06028dfbf2ff7aa2cfdea46fbcc4f4/README.md","retrieved_at":"2026-09-16T19:54:13.645687+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/Sanofi-Public/CodonBert/blob/451a1b167c06028dfbf2ff7aa2cfdea46fbcc4f4/README.md","version":"451a1b167c06028dfbf2ff7aa2cfdea46fbcc4f4"}}
{"id":"evidence-reported-cupid-rna-interactions-2026-license-txt","kind":"source","name":"AnacletoLAB/ncRNA-CUPID LICENSE.txt","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"3376f2209385e0f99839898bebb8cb39edb37ce7c23121c0b5501b99dab6ca40","artifact_url":"https://raw.githubusercontent.com/AnacletoLAB/ncRNA-CUPID/f663c10d2f6c33f8513614badbbc673e654a64d8/LICENSE.txt","retrieved_at":"2026-09-16T19:54:13.686121+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/AnacletoLAB/ncRNA-CUPID/blob/f663c10d2f6c33f8513614badbbc673e654a64d8/LICENSE.txt","version":"f663c10d2f6c33f8513614badbbc673e654a64d8"}}
{"id":"evidence-reported-cupid-rna-interactions-2026-readme-md","kind":"source","name":"AnacletoLAB/ncRNA-CUPID README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"17e3dbece27b0cd7b6031ecc817f941ad351c95ee0416e0bd8e4020ecaed7f5b","artifact_url":"https://raw.githubusercontent.com/AnacletoLAB/ncRNA-CUPID/f663c10d2f6c33f8513614badbbc673e654a64d8/README.md","retrieved_at":"2026-09-16T19:54:13.686121+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/AnacletoLAB/ncRNA-CUPID/blob/f663c10d2f6c33f8513614badbbc673e654a64d8/README.md","version":"f663c10d2f6c33f8513614badbbc673e654a64d8"}}
{"id":"evidence-reported-cyaprombert-2022-readme-md","kind":"source","name":"hanepira/TSSnote-CyaPromBert README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"c916c6c99b07fa3a440264eeefeac8a6b1508867da31c7233bc4d24367fd8bbd","artifact_url":"https://raw.githubusercontent.com/hanepira/TSSnote-CyaPromBert/e86f5449e2e2af3fead1b418ba721f38feb61318/README.md","retrieved_at":"2026-09-16T19:54:13.729708+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/hanepira/TSSnote-CyaPromBert/blob/e86f5449e2e2af3fead1b418ba721f38feb61318/README.md","version":"e86f5449e2e2af3fead1b418ba721f38feb61318"}}
{"id":"evidence-reported-deelig-2021-readme-md","kind":"source","name":"asadahmedtech/DEELIG README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"a6c75971df142bd46ea22287f97aeafd861b035548cddef2bd8ad75f75f3aafd","artifact_url":"https://raw.githubusercontent.com/asadahmedtech/DEELIG/3a3993fc903c40f1ce904111c8e085c79fb45df6/README.md","retrieved_at":"2026-09-16T19:54:13.759450+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/asadahmedtech/DEELIG/blob/3a3993fc903c40f1ce904111c8e085c79fb45df6/README.md","version":"3a3993fc903c40f1ce904111c8e085c79fb45df6"}}
{"id":"evidence-reported-deepinteraware-2025-readme-md","kind":"source","name":"BioMedicalBigDataMiningLab/DeepInterAware README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"9f1cef47f48b4b3b0084e6de73d271166f5c85344603b6aac5a64521eac10a9a","artifact_url":"https://raw.githubusercontent.com/BioMedicalBigDataMiningLab/DeepInterAware/11a283264fb7177f842d56d5c6b49f9bb7f10abd/README.md","retrieved_at":"2026-09-16T20:42:56.583634+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/BioMedicalBigDataMiningLab/DeepInterAware/blob/11a283264fb7177f842d56d5c6b49f9bb7f10abd/README.md","version":"11a283264fb7177f842d56d5c6b49f9bb7f10abd"}}
{"id":"evidence-reported-detire-viral-metagenomes-2023-readme-md","kind":"source","name":"crazyinter/DETIRE README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"cb1ac5f1df284b248f02fd1ac8f6e43a825c7c918e64cba381df3792e2403e9f","artifact_url":"https://raw.githubusercontent.com/crazyinter/DETIRE/6b48c5bcb1303abe593173633d1f13da1d8d5869/README.md","retrieved_at":"2026-09-16T19:54:13.776207+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/crazyinter/DETIRE/blob/6b48c5bcb1303abe593173633d1f13da1d8d5869/README.md","version":"6b48c5bcb1303abe593173633d1f13da1d8d5869"}}
{"id":"evidence-reported-dna-foundation-models-2025-readme-md","kind":"source","name":"ChongWuLab/dna_foundation_benchmark README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"e678c69a91fd0b709864d2c0b83f33e273b82e389dd07b6f380b76236631fac2","artifact_url":"https://raw.githubusercontent.com/ChongWuLab/dna_foundation_benchmark/3f4c81ce066f3c47422a83466b085aac1a6be902/README.md","retrieved_at":"2026-09-16T19:54:14.870570+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/ChongWuLab/dna_foundation_benchmark/blob/3f4c81ce066f3c47422a83466b085aac1a6be902/README.md","version":"3f4c81ce066f3c47422a83466b085aac1a6be902"}}
{"id":"evidence-reported-enhancer-position-encoding-2024-readme-md","kind":"source","name":"xing1999/PDCNN README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"77382e4a76f669866bb8e81551aba59b5993a1ff816c9b530a823940e6db1992","artifact_url":"https://raw.githubusercontent.com/xing1999/PDCNN/ff302344eb1a03bca4802408d03651cb4d2abc99/README.md","retrieved_at":"2026-09-16T19:54:14.982961+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/xing1999/PDCNN/blob/ff302344eb1a03bca4802408d03651cb4d2abc99/README.md","version":"ff302344eb1a03bca4802408d03651cb4d2abc99"}}
{"id":"evidence-reported-ernie-rna-2025-license","kind":"source","name":"Bruce-ywj/ERNIE-RNA LICENSE","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"b7aa7b89bff7b3c4cca2f78dc0900f038a7253bcbea9b780eca30b2bc1610ffb","artifact_url":"https://raw.githubusercontent.com/Bruce-ywj/ERNIE-RNA/43bc06de1088ed03ffd7de918ad4b2c2a3346a43/LICENSE","retrieved_at":"2026-09-16T19:54:15.083678+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/Bruce-ywj/ERNIE-RNA/blob/43bc06de1088ed03ffd7de918ad4b2c2a3346a43/LICENSE","version":"43bc06de1088ed03ffd7de918ad4b2c2a3346a43"}}
{"id":"evidence-reported-ernie-rna-2025-readme-md","kind":"source","name":"Bruce-ywj/ERNIE-RNA README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"d2b2ca2f33c4d4d1dac63659cff494631732b7ec87ac07ad99a0826e55bfa603","artifact_url":"https://raw.githubusercontent.com/Bruce-ywj/ERNIE-RNA/43bc06de1088ed03ffd7de918ad4b2c2a3346a43/README.md","retrieved_at":"2026-09-16T19:54:15.083678+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/Bruce-ywj/ERNIE-RNA/blob/43bc06de1088ed03ffd7de918ad4b2c2a3346a43/README.md","version":"43bc06de1088ed03ffd7de918ad4b2c2a3346a43"}}
{"id":"evidence-reported-fujisan-2024-readme-md","kind":"source","name":"sfujita0601/FUJISAN README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"cb023829651bce1023dd11f603db6731efbff534e04e863821b8677fcec5a19e","artifact_url":"https://raw.githubusercontent.com/sfujita0601/FUJISAN/588daf65810c49b021af9e47b11ec752986dcb94/README.md","retrieved_at":"2026-09-16T19:54:15.262300+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/sfujita0601/FUJISAN/blob/588daf65810c49b021af9e47b11ec752986dcb94/README.md","version":"588daf65810c49b021af9e47b11ec752986dcb94"}}
{"id":"evidence-reported-fusion-breakpoint-foundation-models-2026-license","kind":"source","name":"kbi-fbmi/articles--2026fusionEmbBenchmark LICENSE","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"fbc0aec4a1e90f07e8867eda6bc64940755f69d49c11ce284c6b6075df98eabf","artifact_url":"https://raw.githubusercontent.com/kbi-fbmi/articles--2026fusionEmbBenchmark/085a6d7d2f899b0f62d764f35d1248b2eda567da/LICENSE","retrieved_at":"2026-09-16T19:54:16.293121+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/kbi-fbmi/articles--2026fusionEmbBenchmark/blob/085a6d7d2f899b0f62d764f35d1248b2eda567da/LICENSE","version":"085a6d7d2f899b0f62d764f35d1248b2eda567da"}}
{"id":"evidence-reported-fusion-breakpoint-foundation-models-2026-readme-md","kind":"source","name":"kbi-fbmi/articles--2026fusionEmbBenchmark README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"895b3a5d9f861be27e7f19b77b5f875f72abcea680ed444d8794d0d6604b4dc3","artifact_url":"https://raw.githubusercontent.com/kbi-fbmi/articles--2026fusionEmbBenchmark/085a6d7d2f899b0f62d764f35d1248b2eda567da/README.md","retrieved_at":"2026-09-16T19:54:16.293121+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/kbi-fbmi/articles--2026fusionEmbBenchmark/blob/085a6d7d2f899b0f62d764f35d1248b2eda567da/README.md","version":"085a6d7d2f899b0f62d764f35d1248b2eda567da"}}
{"id":"evidence-reported-genept-2024-readme-md","kind":"source","name":"yiqunchen/GenePT README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"8740cf98377f964131399de818db5de938db4d9e6abe20fac3cd0354887a99c2","artifact_url":"https://raw.githubusercontent.com/yiqunchen/GenePT/3602699e7425a7be577771f8f07e218db6c79b9f/README.md","retrieved_at":"2026-09-16T19:54:16.422790+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/yiqunchen/GenePT/blob/3602699e7425a7be577771f8f07e218db6c79b9f/README.md","version":"3602699e7425a7be577771f8f07e218db6c79b9f"}}
{"id":"evidence-reported-genomeocean-2025-license","kind":"source","name":"jgi-genomeocean/genomeocean LICENSE","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"1700c41a252e2fa9dd284a70387de610f2b47e663b8954c0538f1a9b9227e7a9","artifact_url":"https://raw.githubusercontent.com/jgi-genomeocean/genomeocean/06fa433169539a3c84d7366a663933b888a5386b/LICENSE","retrieved_at":"2026-09-16T19:54:16.470820+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/jgi-genomeocean/genomeocean/blob/06fa433169539a3c84d7366a663933b888a5386b/LICENSE","version":"06fa433169539a3c84d7366a663933b888a5386b"}}
{"id":"evidence-reported-genomeocean-2025-readme-md","kind":"source","name":"jgi-genomeocean/genomeocean README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"57bfdecac0eee10307ebc396fed2f9398492ac9d493eccb86e3607599f8ae907","artifact_url":"https://raw.githubusercontent.com/jgi-genomeocean/genomeocean/06fa433169539a3c84d7366a663933b888a5386b/README.md","retrieved_at":"2026-09-16T19:54:16.470820+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/jgi-genomeocean/genomeocean/blob/06fa433169539a3c84d7366a663933b888a5386b/README.md","version":"06fa433169539a3c84d7366a663933b888a5386b"}}
{"id":"evidence-reported-gremln-2026-license-md","kind":"source","name":"czi-ai/GREmLN LICENSE.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"4538ca25a86f413b4f0856752768517f93a98d76f53ee09989d9b162144a8f6d","artifact_url":"https://raw.githubusercontent.com/czi-ai/GREmLN/e1c5d8edbe2fe96568ff5451f15bc691722bb6a1/LICENSE.md","retrieved_at":"2026-09-16T20:42:56.518821+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/czi-ai/GREmLN/blob/e1c5d8edbe2fe96568ff5451f15bc691722bb6a1/LICENSE.md","version":"e1c5d8edbe2fe96568ff5451f15bc691722bb6a1"}}
{"id":"evidence-reported-gremln-2026-readme-md","kind":"source","name":"czi-ai/GREmLN README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"e6365c6df93ad707c57e04d57b6e3aa4121a1d36672e9a4f75db4a8ac664a858","artifact_url":"https://raw.githubusercontent.com/czi-ai/GREmLN/e1c5d8edbe2fe96568ff5451f15bc691722bb6a1/README.md","retrieved_at":"2026-09-16T20:42:56.518821+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/czi-ai/GREmLN/blob/e1c5d8edbe2fe96568ff5451f15bc691722bb6a1/README.md","version":"e1c5d8edbe2fe96568ff5451f15bc691722bb6a1"}}
{"id":"evidence-reported-gsmformer-ppi-2026-readme-md","kind":"source","name":"ChervovNikita/gsmformer-ppi README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"45003a18fbc137225141cc473b1399c47952183bf784cef8904d8ec3e298cf6b","artifact_url":"https://raw.githubusercontent.com/ChervovNikita/gsmformer-ppi/db9886e8b295f35b544a5703659e2a9115ce9e22/README.md","retrieved_at":"2026-09-16T19:54:16.546691+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/ChervovNikita/gsmformer-ppi/blob/db9886e8b295f35b544a5703659e2a9115ce9e22/README.md","version":"db9886e8b295f35b544a5703659e2a9115ce9e22"}}
{"id":"evidence-reported-hi-enhancer-2025-readme-txt","kind":"source","name":"emanlee/Hi-Enhancer README.txt","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"d2880c9ae82730700607d25c7347641dba5484604e841a696be874589d744942","artifact_url":"https://raw.githubusercontent.com/emanlee/Hi-Enhancer/435bb1cc9ec2909d6ee551bb53c56f6a7bdde8fd/README.txt","retrieved_at":"2026-09-16T19:54:16.659616+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/emanlee/Hi-Enhancer/blob/435bb1cc9ec2909d6ee551bb53c56f6a7bdde8fd/README.txt","version":"435bb1cc9ec2909d6ee551bb53c56f6a7bdde8fd"}}
{"id":"evidence-reported-ibex-2025-license","kind":"source","name":"prescient-design/ibex LICENSE","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4","artifact_url":"https://raw.githubusercontent.com/prescient-design/ibex/2e785563806a3b63600eeaa2107d1254ddf5d196/LICENSE","retrieved_at":"2026-09-16T19:54:16.662235+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/prescient-design/ibex/blob/2e785563806a3b63600eeaa2107d1254ddf5d196/LICENSE","version":"2e785563806a3b63600eeaa2107d1254ddf5d196"}}
{"id":"evidence-reported-ibex-2025-readme-md","kind":"source","name":"prescient-design/ibex README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"e69723730602ff47d4f33e2eb37140bcfeff8c8a491ea15580f973fa51d1b4a9","artifact_url":"https://raw.githubusercontent.com/prescient-design/ibex/2e785563806a3b63600eeaa2107d1254ddf5d196/README.md","retrieved_at":"2026-09-16T19:54:16.662235+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/prescient-design/ibex/blob/2e785563806a3b63600eeaa2107d1254ddf5d196/README.md","version":"2e785563806a3b63600eeaa2107d1254ddf5d196"}}
{"id":"evidence-reported-icctax-2025-readme-md","kind":"source","name":"Ying-Lab/ICCTax README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"749dcacc87c288d7a248a9bd5f58d3ebe39f229ff8853b3fff694957541e0668","artifact_url":"https://raw.githubusercontent.com/Ying-Lab/ICCTax/6b7381c7111bde6d40324cd033501204bf3ac3bc/README.md","retrieved_at":"2026-09-16T19:54:16.993236+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/Ying-Lab/ICCTax/blob/6b7381c7111bde6d40324cd033501204bf3ac3bc/README.md","version":"6b7381c7111bde6d40324cd033501204bf3ac3bc"}}
{"id":"evidence-reported-insilico-perturbation-auprc-2025-readme-md","kind":"source","name":"hxzhu491/Cell-Perturbation-evaluation-Metric README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"f72281cb1777ca70ba37e974ad72ed7fcfd83639e7e008b0e83824a02208c567","artifact_url":"https://raw.githubusercontent.com/hxzhu491/Cell-Perturbation-evaluation-Metric/3b5f8a2ed001c074936287ece478747376c8f5bf/README.md","retrieved_at":"2026-09-16T19:54:17.543702+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/hxzhu491/Cell-Perturbation-evaluation-Metric/blob/3b5f8a2ed001c074936287ece478747376c8f5bf/README.md","version":"3b5f8a2ed001c074936287ece478747376c8f5bf"}}
{"id":"evidence-reported-ipro70-original","kind":"source","name":"iPro70-FMWin original paper","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"e4fa53c5574961acf4a93bf251b71c248025c64b1b79e74da40ff73091d8b205","artifact_url":"https://rafsanjani.pythonanywhere.com/static/Papers/iPro70.pdf","retrieved_at":"2026-09-16T20:40:28.842295+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://rafsanjani.pythonanywhere.com/static/Papers/iPro70.pdf","version":"10.1007/s00438-018-1487-5"}}
{"id":"evidence-reported-ipromp-2025-readme-md","kind":"source","name":"Jackie-Suv/iPro-MP README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"8245ed3d5199578dc5c47af893eb96b28b100ac5fc88dfc2585074096d7f8191","artifact_url":"https://raw.githubusercontent.com/Jackie-Suv/iPro-MP/4266b521bc6617db939c5871cb1b6850dff63fdb/README.md","retrieved_at":"2026-09-16T19:54:17.673506+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/Jackie-Suv/iPro-MP/blob/4266b521bc6617db939c5871cb1b6850dff63fdb/README.md","version":"4266b521bc6617db939c5871cb1b6850dff63fdb"}}
{"id":"evidence-reported-kmetashot-2025-license","kind":"source","name":"gdefazio/kMetaShot LICENSE","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"9e54fb30a82e1879c56cb0206a8107659190dd76396029df8b7c5363f7757cfc","artifact_url":"https://raw.githubusercontent.com/gdefazio/kMetaShot/95dac648aba94d119d20929478f6c3955206f9d4/LICENSE","retrieved_at":"2026-09-16T19:54:17.812170+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/gdefazio/kMetaShot/blob/95dac648aba94d119d20929478f6c3955206f9d4/LICENSE","version":"95dac648aba94d119d20929478f6c3955206f9d4"}}
{"id":"evidence-reported-kmetashot-2025-readme-md","kind":"source","name":"gdefazio/kMetaShot README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"0b1a56d4d6106e664ffdbfa0771cb6d9be65b55b1ada5c6b5299188dc00fc7dc","artifact_url":"https://raw.githubusercontent.com/gdefazio/kMetaShot/95dac648aba94d119d20929478f6c3955206f9d4/README.md","retrieved_at":"2026-09-16T19:54:17.812170+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/gdefazio/kMetaShot/blob/95dac648aba94d119d20929478f6c3955206f9d4/README.md","version":"95dac648aba94d119d20929478f6c3955206f9d4"}}
{"id":"evidence-reported-lemur-magnet-2024-license","kind":"source","name":"treangenlab/lemur LICENSE","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"493e198a2a2b2ddf102c371bdd51bf2be1e974e30117321d1a8a50e531b81787","artifact_url":"https://raw.githubusercontent.com/treangenlab/lemur/eda2cb57727b72fc5b1fb28be1fe45a4826100f9/LICENSE","retrieved_at":"2026-09-16T19:54:18.142852+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/treangenlab/lemur/blob/eda2cb57727b72fc5b1fb28be1fe45a4826100f9/LICENSE","version":"eda2cb57727b72fc5b1fb28be1fe45a4826100f9"}}
{"id":"evidence-reported-lemur-magnet-2024-readme-md","kind":"source","name":"treangenlab/lemur README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"cbe1079a582be626c015d00a80dffb199090761ed5c7d621def0aef28d032aeb","artifact_url":"https://raw.githubusercontent.com/treangenlab/lemur/eda2cb57727b72fc5b1fb28be1fe45a4826100f9/README.md","retrieved_at":"2026-09-16T19:54:18.142852+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/treangenlab/lemur/blob/eda2cb57727b72fc5b1fb28be1fe45a4826100f9/README.md","version":"eda2cb57727b72fc5b1fb28be1fe45a4826100f9"}}
{"id":"evidence-reported-ligand-affinity-meta-model-2024-license","kind":"source","name":"Lee1701/Lee2023a LICENSE","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"3972dc9744f6499f0f9b2dbf76696f2ae7ad8af9b23dde66d6af86c9dfb36986","artifact_url":"https://raw.githubusercontent.com/Lee1701/Lee2023a/92def517a0edcb0470826353afd81899bcfdb4b1/LICENSE","retrieved_at":"2026-09-16T20:30:16.574955+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/Lee1701/Lee2023a/blob/92def517a0edcb0470826353afd81899bcfdb4b1/LICENSE","version":"92def517a0edcb0470826353afd81899bcfdb4b1"}}
{"id":"evidence-reported-ligand-affinity-meta-model-2024-readme-md","kind":"source","name":"Lee1701/Lee2023a README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"a428cb0e912f6608de6a17ec162193eaefa9f96074278ac9d9b2ecf051cdbedd","artifact_url":"https://raw.githubusercontent.com/Lee1701/Lee2023a/92def517a0edcb0470826353afd81899bcfdb4b1/README.md","retrieved_at":"2026-09-16T20:30:16.574955+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/Lee1701/Lee2023a/blob/92def517a0edcb0470826353afd81899bcfdb4b1/README.md","version":"92def517a0edcb0470826353afd81899bcfdb4b1"}}
{"id":"evidence-reported-mdl4microbiome-2022-readme-md","kind":"source","name":"DMnBI/MDL4Microbiome README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"8737382048b17436607ae12302dfcfed021efcc39fa849d8936c6b9ac9f52d49","artifact_url":"https://raw.githubusercontent.com/DMnBI/MDL4Microbiome/0b2076cbd31af62224bd7a90ef0130e4cac50017/README.md","retrieved_at":"2026-09-16T19:54:18.201417+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/DMnBI/MDL4Microbiome/blob/0b2076cbd31af62224bd7a90ef0130e4cac50017/README.md","version":"0b2076cbd31af62224bd7a90ef0130e4cac50017"}}
{"id":"evidence-reported-megsite-2025-readme-md","kind":"source","name":"pengsl-lab/MegSite README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"226fee71219189aded7aa165f7de40c85d2f96f4106089a8cd047793b524d18c","artifact_url":"https://raw.githubusercontent.com/pengsl-lab/MegSite/4d1f5441f15bd20eb3e7c9e5be2d4ca5a857f46b/README.md","retrieved_at":"2026-09-16T19:54:18.210404+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/pengsl-lab/MegSite/blob/4d1f5441f15bd20eb3e7c9e5be2d4ca5a857f46b/README.md","version":"4d1f5441f15bd20eb3e7c9e5be2d4ca5a857f46b"}}
{"id":"evidence-reported-molas-2026-license","kind":"source","name":"BradWangW/MolAS LICENSE","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"33f037265e2bad90af34079cbee68538db96193a25db41d0e227b085fcb6177c","artifact_url":"https://raw.githubusercontent.com/BradWangW/MolAS/a6c417216ddb992f7dc513d511d0429aece4bd61/LICENSE","retrieved_at":"2026-09-16T20:36:11.670546+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/BradWangW/MolAS/blob/a6c417216ddb992f7dc513d511d0429aece4bd61/LICENSE","version":"a6c417216ddb992f7dc513d511d0429aece4bd61"}}
{"id":"evidence-reported-molas-2026-readme-md","kind":"source","name":"BradWangW/MolAS README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"b7a057042375a8c796278242b1b89d8ff7a1cd99203ff2fe31e7d1df9226256e","artifact_url":"https://raw.githubusercontent.com/BradWangW/MolAS/a6c417216ddb992f7dc513d511d0429aece4bd61/README.md","retrieved_at":"2026-09-16T20:36:11.670546+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/BradWangW/MolAS/blob/a6c417216ddb992f7dc513d511d0429aece4bd61/README.md","version":"a6c417216ddb992f7dc513d511d0429aece4bd61"}}
{"id":"evidence-reported-mouse-geneformer-2025-readme-md","kind":"source","name":"machine-perception-robotics-group/Mouse-Geneformer README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"1eefd8677ff2644f55272a1412cf612e1264f92456ae3b701598d9a5b558bc4f","artifact_url":"https://raw.githubusercontent.com/machine-perception-robotics-group/Mouse-Geneformer/ed17d455193ed4c4d93230a8f5f2de349cf20c81/README.md","retrieved_at":"2026-09-16T19:54:18.217286+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/machine-perception-robotics-group/Mouse-Geneformer/blob/ed17d455193ed4c4d93230a8f5f2de349cf20c81/README.md","version":"ed17d455193ed4c4d93230a8f5f2de349cf20c81"}}
{"id":"evidence-reported-mrna-lm-2025-license-txt","kind":"source","name":"Sanofi-Public/mRNA-LM LICENSE.txt","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"6503f6c09579df890e003db39c9b7c299592d25eee97066241b0e7ac9212a1ea","artifact_url":"https://raw.githubusercontent.com/Sanofi-Public/mRNA-LM/d7538c9aadbceb59a8832904292b279d0a4c2d12/LICENSE.txt","retrieved_at":"2026-09-16T19:54:18.716122+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/Sanofi-Public/mRNA-LM/blob/d7538c9aadbceb59a8832904292b279d0a4c2d12/LICENSE.txt","version":"d7538c9aadbceb59a8832904292b279d0a4c2d12"}}
{"id":"evidence-reported-mrna-lm-2025-readme-md","kind":"source","name":"Sanofi-Public/mRNA-LM README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"896d2f64f8a236c9f99cad115feb97dab369bbf4c42bb6025df01bbae0bd740b","artifact_url":"https://raw.githubusercontent.com/Sanofi-Public/mRNA-LM/d7538c9aadbceb59a8832904292b279d0a4c2d12/README.md","retrieved_at":"2026-09-16T19:54:18.716122+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/Sanofi-Public/mRNA-LM/blob/d7538c9aadbceb59a8832904292b279d0a4c2d12/README.md","version":"d7538c9aadbceb59a8832904292b279d0a4c2d12"}}
{"id":"evidence-reported-mrna-protein-diversity-2026-readme-md","kind":"source","name":"cobisLab/mRPI-issue README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"10b6607fc40d285e20b681a7584b3e980265d15dcc79c3ac073aa4f030cfc141","artifact_url":"https://raw.githubusercontent.com/cobisLab/mRPI-issue/0f2d27666c4876f00d4c4e6cb3bec0e1629d214c/README.md","retrieved_at":"2026-09-16T19:54:19.069409+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/cobisLab/mRPI-issue/blob/0f2d27666c4876f00d4c4e6cb3bec0e1629d214c/README.md","version":"0f2d27666c4876f00d4c4e6cb3bec0e1629d214c"}}
{"id":"evidence-reported-mrnabert-2025-license","kind":"source","name":"yyly6/mRNABERT LICENSE","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4","artifact_url":"https://raw.githubusercontent.com/yyly6/mRNABERT/893ccc920bb9be02a4677d14d96b03126da17689/LICENSE","retrieved_at":"2026-09-16T19:54:19.395631+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/yyly6/mRNABERT/blob/893ccc920bb9be02a4677d14d96b03126da17689/LICENSE","version":"893ccc920bb9be02a4677d14d96b03126da17689"}}
{"id":"evidence-reported-mrnabert-2025-readme-md","kind":"source","name":"yyly6/mRNABERT README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"bd50f1d7a71b1fd265b8b4c5590358d9bdabc4b6d897cd3b2ec6ba7ad5f7d2ff","artifact_url":"https://raw.githubusercontent.com/yyly6/mRNABERT/893ccc920bb9be02a4677d14d96b03126da17689/README.md","retrieved_at":"2026-09-16T19:54:19.395631+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/yyly6/mRNABERT/blob/893ccc920bb9be02a4677d14d96b03126da17689/README.md","version":"893ccc920bb9be02a4677d14d96b03126da17689"}}
{"id":"evidence-reported-nabas-plus-2025-license","kind":"source","name":"TakacsBertalan/NABAS_paper_scripts LICENSE","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"c84442114f4da9b2593efa430845e11a0605b0c16259feb906661bc0bd504fb8","artifact_url":"https://raw.githubusercontent.com/TakacsBertalan/NABAS_paper_scripts/7cab4d317a2c362988e7b96fb33f92a9c79a9fdc/LICENSE","retrieved_at":"2026-09-16T19:54:19.540591+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/TakacsBertalan/NABAS_paper_scripts/blob/7cab4d317a2c362988e7b96fb33f92a9c79a9fdc/LICENSE","version":"7cab4d317a2c362988e7b96fb33f92a9c79a9fdc"}}
{"id":"evidence-reported-nabas-plus-2025-readme-md","kind":"source","name":"TakacsBertalan/NABAS_paper_scripts README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"4dd306069e898af754cdf810adbcf7542a24918563d610933f50f48d41aeecac","artifact_url":"https://raw.githubusercontent.com/TakacsBertalan/NABAS_paper_scripts/7cab4d317a2c362988e7b96fb33f92a9c79a9fdc/README.md","retrieved_at":"2026-09-16T19:54:19.540591+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/TakacsBertalan/NABAS_paper_scripts/blob/7cab4d317a2c362988e7b96fb33f92a9c79a9fdc/README.md","version":"7cab4d317a2c362988e7b96fb33f92a9c79a9fdc"}}
{"id":"evidence-reported-ncd-metagenomics-2026-license","kind":"source","name":"ghproducts/genomics-ncd-gzip LICENSE","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"a2010f343487d3f7618affe54f789f5487602331c0a8d03f49e9a7c547cf0499","artifact_url":"https://raw.githubusercontent.com/ghproducts/genomics-ncd-gzip/d5bb37be194a2716bc65543c65a26a65eebc1849/LICENSE","retrieved_at":"2026-09-16T19:54:19.565893+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/ghproducts/genomics-ncd-gzip/blob/d5bb37be194a2716bc65543c65a26a65eebc1849/LICENSE","version":"d5bb37be194a2716bc65543c65a26a65eebc1849"}}
{"id":"evidence-reported-ncd-metagenomics-2026-readme-md","kind":"source","name":"ghproducts/genomics-ncd-gzip README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"79b4134061f21aa3baed67fa656f055686db77adb08ac0ebe2dd6c8a5d4fb92b","artifact_url":"https://raw.githubusercontent.com/ghproducts/genomics-ncd-gzip/d5bb37be194a2716bc65543c65a26a65eebc1849/README.md","retrieved_at":"2026-09-16T19:54:19.565893+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/ghproducts/genomics-ncd-gzip/blob/d5bb37be194a2716bc65543c65a26a65eebc1849/README.md","version":"d5bb37be194a2716bc65543c65a26a65eebc1849"}}
{"id":"evidence-reported-pc-mer-2024-readme-md","kind":"source","name":"SAkbari93/PC-mer_Metagenomics README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"07d86adec2ed17ed488c5af3d6348e9da967c6b6804d57093f2a8576abf132ed","artifact_url":"https://raw.githubusercontent.com/SAkbari93/PC-mer_Metagenomics/5c5f89dcaec5098372ad1fe82d4215186fe417c1/README.md","retrieved_at":"2026-09-16T19:54:20.219994+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/SAkbari93/PC-mer_Metagenomics/blob/5c5f89dcaec5098372ad1fe82d4215186fe417c1/README.md","version":"5c5f89dcaec5098372ad1fe82d4215186fe417c1"}}
{"id":"evidence-reported-phylogpn-2025-license","kind":"source","name":"songlab-cal/gpn LICENSE","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"cef312cb27bc4aae45e0ca7130643ca36f9aa4165ef09bcddf5ba5eb6fb76e90","artifact_url":"https://raw.githubusercontent.com/songlab-cal/gpn/6f28c81bcbfe7d65cb6d8ece9ce88f87ca583791/LICENSE","retrieved_at":"2026-09-16T19:54:20.313775+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/songlab-cal/gpn/blob/6f28c81bcbfe7d65cb6d8ece9ce88f87ca583791/LICENSE","version":"6f28c81bcbfe7d65cb6d8ece9ce88f87ca583791"}}
{"id":"evidence-reported-phylogpn-2025-readme-md","kind":"source","name":"songlab-cal/gpn README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"0312fb9646aef630a30cd3039337b31a48f4345f4dd1cdd71edcc483afd8b9d0","artifact_url":"https://raw.githubusercontent.com/songlab-cal/gpn/6f28c81bcbfe7d65cb6d8ece9ce88f87ca583791/README.md","retrieved_at":"2026-09-16T19:54:20.313775+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/songlab-cal/gpn/blob/6f28c81bcbfe7d65cb6d8ece9ce88f87ca583791/README.md","version":"6f28c81bcbfe7d65cb6d8ece9ce88f87ca583791"}}
{"id":"evidence-reported-plantcad2-2025-license","kind":"source","name":"plantcad/plantcad LICENSE","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4","artifact_url":"https://raw.githubusercontent.com/plantcad/plantcad/7240f0238f869b3ac25e4b5ad0996fad96ede9db/LICENSE","retrieved_at":"2026-09-16T19:54:20.832487+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/plantcad/plantcad/blob/7240f0238f869b3ac25e4b5ad0996fad96ede9db/LICENSE","version":"7240f0238f869b3ac25e4b5ad0996fad96ede9db"}}
{"id":"evidence-reported-plantcad2-2025-readme-md","kind":"source","name":"plantcad/plantcad README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"74ef9550db6752e6d99441c48751919e172ce5e2ce83cf57deaa08f24df1df8e","artifact_url":"https://raw.githubusercontent.com/plantcad/plantcad/7240f0238f869b3ac25e4b5ad0996fad96ede9db/README.md","retrieved_at":"2026-09-16T19:54:20.832487+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/plantcad/plantcad/blob/7240f0238f869b3ac25e4b5ad0996fad96ede9db/README.md","version":"7240f0238f869b3ac25e4b5ad0996fad96ede9db"}}
{"id":"evidence-reported-prime-2026-license-md","kind":"source","name":"lanl/prime LICENSE.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"1a852d32b1c472babd61e142e4c3674ac971112bd438a8f3eb006af752390c9a","artifact_url":"https://raw.githubusercontent.com/lanl/prime/d940c51aa0f475b8945789e89761fab0687d5b74/LICENSE.md","retrieved_at":"2026-09-16T19:54:21.047073+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/lanl/prime/blob/d940c51aa0f475b8945789e89761fab0687d5b74/LICENSE.md","version":"d940c51aa0f475b8945789e89761fab0687d5b74"}}
{"id":"evidence-reported-prime-2026-readme-md","kind":"source","name":"lanl/prime README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"35a8c35aad87c03f8a5afcb66dd1b6477ff7ab187e77d5951503773d7dfbc896","artifact_url":"https://raw.githubusercontent.com/lanl/prime/d940c51aa0f475b8945789e89761fab0687d5b74/README.md","retrieved_at":"2026-09-16T19:54:21.047073+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/lanl/prime/blob/d940c51aa0f475b8945789e89761fab0687d5b74/README.md","version":"d940c51aa0f475b8945789e89761fab0687d5b74"}}
{"id":"evidence-reported-prokbert-2024-license","kind":"source","name":"nbrg-ppcu/prokbert LICENSE","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"66e17f3a57ace034ab422667c02d75134bb531619d9491f3abc7ea42bb8a6643","artifact_url":"https://raw.githubusercontent.com/nbrg-ppcu/prokbert/8670ae92b816cff158a0b85647a8dea122e251eb/LICENSE","retrieved_at":"2026-09-16T19:54:21.078483+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/nbrg-ppcu/prokbert/blob/8670ae92b816cff158a0b85647a8dea122e251eb/LICENSE","version":"8670ae92b816cff158a0b85647a8dea122e251eb"}}
{"id":"evidence-reported-prokbert-2024-readme-md","kind":"source","name":"nbrg-ppcu/prokbert README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"29de39c6ad006ce704ab14240cfd97af93da411fb63ec89ebe499c2646928cfc","artifact_url":"https://raw.githubusercontent.com/nbrg-ppcu/prokbert/8670ae92b816cff158a0b85647a8dea122e251eb/README.md","retrieved_at":"2026-09-16T19:54:21.078483+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/nbrg-ppcu/prokbert/blob/8670ae92b816cff158a0b85647a8dea122e251eb/README.md","version":"8670ae92b816cff158a0b85647a8dea122e251eb"}}
{"id":"evidence-reported-protein-binding-sites-2023-readme-md","kind":"source","name":"houzl3416/EDLMPPI README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"95a7d8240f3e685689d2f978e7081a0ff7dd8d058fe850f54662f87530632e83","artifact_url":"https://raw.githubusercontent.com/houzl3416/EDLMPPI/78e4a7b36bb83ccf4274786b859125178804f434/README.md","retrieved_at":"2026-09-16T19:54:21.260563+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/houzl3416/EDLMPPI/blob/78e4a7b36bb83ccf4274786b859125178804f434/README.md","version":"78e4a7b36bb83ccf4274786b859125178804f434"}}
{"id":"evidence-reported-pst-2025-license","kind":"source","name":"BorgwardtLab/PST LICENSE","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"f2ce3382be97305fc31c8b061f1f08969e6f95c677db5b8a7c3bc81352237dfb","artifact_url":"https://raw.githubusercontent.com/BorgwardtLab/PST/57d9dcd8200900504a19de459450e137867262d7/LICENSE","retrieved_at":"2026-09-16T19:54:21.541747+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/BorgwardtLab/PST/blob/57d9dcd8200900504a19de459450e137867262d7/LICENSE","version":"57d9dcd8200900504a19de459450e137867262d7"}}
{"id":"evidence-reported-pst-2025-readme-md","kind":"source","name":"BorgwardtLab/PST README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"e3422ce3da04d004549e37cfcbe9a509d53a8cd9217b39598ac10b6b444a8121","artifact_url":"https://raw.githubusercontent.com/BorgwardtLab/PST/57d9dcd8200900504a19de459450e137867262d7/README.md","retrieved_at":"2026-09-16T19:54:21.541747+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/BorgwardtLab/PST/blob/57d9dcd8200900504a19de459450e137867262d7/README.md","version":"57d9dcd8200900504a19de459450e137867262d7"}}
{"id":"evidence-reported-r3design-2025-readme-md","kind":"source","name":"A4Bio/R3Design readme.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"132adf2bb9ae5cc090a33a01f2534bf37a5a7daee3b4a4f19b28803d0655da79","artifact_url":"https://raw.githubusercontent.com/A4Bio/R3Design/c05dc4b35100949011c77c08f61a9e80280987ce/readme.md","retrieved_at":"2026-09-16T19:54:22.378999+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/A4Bio/R3Design/blob/c05dc4b35100949011c77c08f61a9e80280987ce/readme.md","version":"c05dc4b35100949011c77c08f61a9e80280987ce"}}
{"id":"evidence-reported-rewire-license","kind":"source","name":"Rewire benchmark runner licence","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"4fff574dd02d01f7cba42c9a4472bce39bb96f9b650a65f51958ab79bc829ea4","artifact_url":"https://github.com/rewire-bio/rewire-benchmarks/blob/bee9133b83f3aedaf2bbb9013f1875515845607e/LICENSE","retrieved_at":"2026-09-16T20:47:15.959417+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/rewire-bio/rewire-benchmarks/blob/bee9133b83f3aedaf2bbb9013f1875515845607e/LICENSE","version":"bee9133b83f3aedaf2bbb9013f1875515845607e"}}
{"id":"evidence-reported-rewire-run-baseline","kind":"source","name":"MFASS v2 run_baseline.py","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"7066ae357aac960ddd6e95ee55a98dcd876b619edfadae1e1d9d03eb3f5b0acc","artifact_url":"https://github.com/rewire-bio/rewire-benchmarks/blob/bee9133b83f3aedaf2bbb9013f1875515845607e/benchmarks/mfass/src/mfass/run_baseline.py","retrieved_at":"2026-09-16T20:47:15.959417+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/rewire-bio/rewire-benchmarks/blob/bee9133b83f3aedaf2bbb9013f1875515845607e/benchmarks/mfass/src/mfass/run_baseline.py","version":"bee9133b83f3aedaf2bbb9013f1875515845607e"}}
{"id":"evidence-reported-rewire-run-dnabert2","kind":"source","name":"MFASS v2 run_dnabert2.py","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"b7ebb4324e22cb421482181c52722c64cac5fec7c43161c082fc34e9920c4a85","artifact_url":"https://github.com/rewire-bio/rewire-benchmarks/blob/bee9133b83f3aedaf2bbb9013f1875515845607e/benchmarks/mfass/src/mfass/run_dnabert2.py","retrieved_at":"2026-09-16T20:47:15.959417+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/rewire-bio/rewire-benchmarks/blob/bee9133b83f3aedaf2bbb9013f1875515845607e/benchmarks/mfass/src/mfass/run_dnabert2.py","version":"bee9133b83f3aedaf2bbb9013f1875515845607e"}}
{"id":"evidence-reported-rewire-run-pangolin","kind":"source","name":"MFASS v2 run_pangolin.py","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"83e5b71d7265c63fb0874d7b493ac2c8d205f645c079b09f350d69b44e5c921b","artifact_url":"https://github.com/rewire-bio/rewire-benchmarks/blob/bee9133b83f3aedaf2bbb9013f1875515845607e/benchmarks/mfass/src/mfass/run_pangolin.py","retrieved_at":"2026-09-16T20:47:15.959417+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/rewire-bio/rewire-benchmarks/blob/bee9133b83f3aedaf2bbb9013f1875515845607e/benchmarks/mfass/src/mfass/run_pangolin.py","version":"bee9133b83f3aedaf2bbb9013f1875515845607e"}}
{"id":"evidence-reported-rewire-run-spliceai","kind":"source","name":"MFASS v2 run_spliceai.py","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"3709bc12997744b77ce2f9ea4de79cdcafa6ffc136b61bb020f57e43acbe8685","artifact_url":"https://github.com/rewire-bio/rewire-benchmarks/blob/bee9133b83f3aedaf2bbb9013f1875515845607e/benchmarks/mfass/src/mfass/run_spliceai.py","retrieved_at":"2026-09-16T20:47:15.959417+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/rewire-bio/rewire-benchmarks/blob/bee9133b83f3aedaf2bbb9013f1875515845607e/benchmarks/mfass/src/mfass/run_spliceai.py","version":"bee9133b83f3aedaf2bbb9013f1875515845607e"}}
{"id":"evidence-reported-rlsite-rna-binding-2025-readme-md","kind":"source","name":"SaisaiSun/RLsite README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"56a8c16f47bec013c53fb2ce7ecc1c6f87db0b724176677b00d44279b4b3ec09","artifact_url":"https://raw.githubusercontent.com/SaisaiSun/RLsite/3d4a294aab295cbc1afd56b9ea6ced96e0717304/README.md","retrieved_at":"2026-09-16T19:54:22.406853+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/SaisaiSun/RLsite/blob/3d4a294aab295cbc1afd56b9ea6ced96e0717304/README.md","version":"3d4a294aab295cbc1afd56b9ea6ced96e0717304"}}
{"id":"evidence-reported-rnaret-2026-license","kind":"source","name":"DrBlackZJU/RNAret LICENSE","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"a2010f343487d3f7618affe54f789f5487602331c0a8d03f49e9a7c547cf0499","artifact_url":"https://raw.githubusercontent.com/DrBlackZJU/RNAret/40ddab25fc038ba2b96bc9b9f88216abe38b2f64/LICENSE","retrieved_at":"2026-09-16T19:54:22.459346+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/DrBlackZJU/RNAret/blob/40ddab25fc038ba2b96bc9b9f88216abe38b2f64/LICENSE","version":"40ddab25fc038ba2b96bc9b9f88216abe38b2f64"}}
{"id":"evidence-reported-rnaret-2026-readme-md","kind":"source","name":"DrBlackZJU/RNAret README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"b04da58b8774a9db3c6d873af7d49142c039afee958ab6dcd2d2ff3b29d7ddae","artifact_url":"https://raw.githubusercontent.com/DrBlackZJU/RNAret/40ddab25fc038ba2b96bc9b9f88216abe38b2f64/README.md","retrieved_at":"2026-09-16T19:54:22.459346+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/DrBlackZJU/RNAret/blob/40ddab25fc038ba2b96bc9b9f88216abe38b2f64/README.md","version":"40ddab25fc038ba2b96bc9b9f88216abe38b2f64"}}
{"id":"evidence-reported-scalr-2025-license","kind":"source","name":"infocusp/scaLR LICENSE","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"55860dd9f0c93456c17596a049cb34cded65afa3219888cb7d0a3eb18e86ff68","artifact_url":"https://raw.githubusercontent.com/infocusp/scaLR/b5f72ce8f9bd25cb90f4b6b3112a03288c788f2e/LICENSE","retrieved_at":"2026-09-16T19:54:22.478513+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/infocusp/scaLR/blob/b5f72ce8f9bd25cb90f4b6b3112a03288c788f2e/LICENSE","version":"b5f72ce8f9bd25cb90f4b6b3112a03288c788f2e"}}
{"id":"evidence-reported-scalr-2025-readme-md","kind":"source","name":"infocusp/scaLR README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"ba8ca22790d5265355298d4ca2adc7b5a01290d5bd07e45fac332b7dc57f31f5","artifact_url":"https://raw.githubusercontent.com/infocusp/scaLR/b5f72ce8f9bd25cb90f4b6b3112a03288c788f2e/README.md","retrieved_at":"2026-09-16T19:54:22.478513+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/infocusp/scaLR/blob/b5f72ce8f9bd25cb90f4b6b3112a03288c788f2e/README.md","version":"b5f72ce8f9bd25cb90f4b6b3112a03288c788f2e"}}
{"id":"evidence-reported-scatac-llmda-2026-readme-md","kind":"source","name":"sheng-guan-2001/scLLMDA README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"b8327f7bbd867d6e5b7ab369873b59aeba64fe4ca79c879dd54c3b036b127d3b","artifact_url":"https://raw.githubusercontent.com/sheng-guan-2001/scLLMDA/5e24025710bb068312d50a5749ef6bb838ef5a32/README.md","retrieved_at":"2026-09-16T19:54:22.541360+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/sheng-guan-2001/scLLMDA/blob/5e24025710bb068312d50a5749ef6bb838ef5a32/README.md","version":"5e24025710bb068312d50a5749ef6bb838ef5a32"}}
{"id":"evidence-reported-scregnet-2025-readme-md","kind":"source","name":"sindhura-cs/scRegNet README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"fa82e7aab8e650e488e267fbed583735e40f9d7e005e72d252c28860dd970d94","artifact_url":"https://raw.githubusercontent.com/sindhura-cs/scRegNet/30d0215efd99c40ceaceb37d161fc0a86a236e0b/README.md","retrieved_at":"2026-09-16T19:54:23.045099+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/sindhura-cs/scRegNet/blob/30d0215efd99c40ceaceb37d161fc0a86a236e0b/README.md","version":"30d0215efd99c40ceaceb37d161fc0a86a236e0b"}}
{"id":"evidence-reported-scxdr-2026-readme-md","kind":"source","name":"QiGuan1920/scXDR2025 README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"e61ed3f85619ffed72cde43a6b3f13491f0788c37edf03266ffcbb77a0c3894c","artifact_url":"https://raw.githubusercontent.com/QiGuan1920/scXDR2025/5b39f37ba4df186eeea0881d59366458d2535db9/README.md","retrieved_at":"2026-09-16T19:54:23.123989+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/QiGuan1920/scXDR2025/blob/5b39f37ba4df186eeea0881d59366458d2535db9/README.md","version":"5b39f37ba4df186eeea0881d59366458d2535db9"}}
{"id":"evidence-reported-single-cell-aging-probes-2026-readme-md","kind":"source","name":"Biodyn-AI/longevity-mechinterp README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"d858613c397dcfd9904d090eea810941babb7a98d6458275d71f7b8bc9689e71","artifact_url":"https://raw.githubusercontent.com/Biodyn-AI/longevity-mechinterp/5a61464632a3c3e8bebd396eb1ab17bce1dc2493/README.md","retrieved_at":"2026-09-16T19:54:23.526549+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/Biodyn-AI/longevity-mechinterp/blob/5a61464632a3c3e8bebd396eb1ab17bce1dc2493/README.md","version":"5a61464632a3c3e8bebd396eb1ab17bce1dc2493"}}
{"id":"evidence-reported-structure-informed-current-html","kind":"source","name":"Structure-Informed Protein Language Models are Robust Predictors for Variant Effects (reviewed HTML snapshot)","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"6686d2646e7b1b1203a7ef6d48bacf966b4cd6b0895fd2855551937c6bb87111","artifact_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12068927/","retrieved_at":"2026-09-16T19:58:11.209175+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12068927/","version":"Human Genetics 2025 journal article (online 2024)"}}
{"id":"evidence-reported-structure-informed-plm-2025-license","kind":"source","name":"Shen-Lab/Structure-informed_PLM LICENSE","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"4a8d4e3f2049f4ab2372435182b3cf913e9ee1c0c6efa56202f637017b94af50","artifact_url":"https://raw.githubusercontent.com/Shen-Lab/Structure-informed_PLM/2307b101f9bf08223729a68f52b8a6fb21f18991/LICENSE","retrieved_at":"2026-09-16T20:43:16.171470+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/Shen-Lab/Structure-informed_PLM/blob/2307b101f9bf08223729a68f52b8a6fb21f18991/LICENSE","version":"2307b101f9bf08223729a68f52b8a6fb21f18991"}}
{"id":"evidence-reported-structure-informed-plm-2025-readme-md","kind":"source","name":"Shen-Lab/Structure-informed_PLM readMe.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"06787a14ac0e9a094fe524472bb2b06b5025aa8af122d913f3ed9c352938b161","artifact_url":"https://raw.githubusercontent.com/Shen-Lab/Structure-informed_PLM/2307b101f9bf08223729a68f52b8a6fb21f18991/readMe.md","retrieved_at":"2026-09-16T20:43:16.171470+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/Shen-Lab/Structure-informed_PLM/blob/2307b101f9bf08223729a68f52b8a6fb21f18991/readMe.md","version":"2307b101f9bf08223729a68f52b8a6fb21f18991"}}
{"id":"evidence-reported-transbind-2026-license","kind":"source","name":"jianlin-cheng/TransBind LICENSE","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"3972dc9744f6499f0f9b2dbf76696f2ae7ad8af9b23dde66d6af86c9dfb36986","artifact_url":"https://raw.githubusercontent.com/jianlin-cheng/TransBind/7537f264c5ad94958bcad05bb57edd8028c323ff/LICENSE","retrieved_at":"2026-09-16T19:54:23.967114+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/jianlin-cheng/TransBind/blob/7537f264c5ad94958bcad05bb57edd8028c323ff/LICENSE","version":"7537f264c5ad94958bcad05bb57edd8028c323ff"}}
{"id":"evidence-reported-transbind-2026-readme-md","kind":"source","name":"jianlin-cheng/TransBind README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"2cdf3d45d6b98e44131007ede55fd1bd5a599a4d1963879eac4f2a9a6e598af7","artifact_url":"https://raw.githubusercontent.com/jianlin-cheng/TransBind/7537f264c5ad94958bcad05bb57edd8028c323ff/README.md","retrieved_at":"2026-09-16T19:54:23.967114+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/jianlin-cheng/TransBind/blob/7537f264c5ad94958bcad05bb57edd8028c323ff/README.md","version":"7537f264c5ad94958bcad05bb57edd8028c323ff"}}
{"id":"evidence-reported-tu-fold-2025-license","kind":"source","name":"ygjiyn/tu_fold LICENSE","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"69bd0ef33acdd9d4340c3925f27ef35141c26b4653b39deac3d78e5f5d5930c9","artifact_url":"https://raw.githubusercontent.com/ygjiyn/tu_fold/f0532b6bf38b2f57baf0ba6afce7766bfc64899b/LICENSE","retrieved_at":"2026-09-16T19:54:24.127143+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/ygjiyn/tu_fold/blob/f0532b6bf38b2f57baf0ba6afce7766bfc64899b/LICENSE","version":"f0532b6bf38b2f57baf0ba6afce7766bfc64899b"}}
{"id":"evidence-reported-tu-fold-2025-readme-md","kind":"source","name":"ygjiyn/tu_fold README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"84d9bbc42e28b17e10a672d9a9539bf26988d92fee8287d9b186e11678c4824b","artifact_url":"https://raw.githubusercontent.com/ygjiyn/tu_fold/f0532b6bf38b2f57baf0ba6afce7766bfc64899b/README.md","retrieved_at":"2026-09-16T19:54:24.127143+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/ygjiyn/tu_fold/blob/f0532b6bf38b2f57baf0ba6afce7766bfc64899b/README.md","version":"f0532b6bf38b2f57baf0ba6afce7766bfc64899b"}}
{"id":"evidence-reported-viral-contig-simulation-2021-license","kind":"source","name":"Strong-Lab/Viral_Classification_in_Metagenomics LICENSE","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"1b9591430c0d9b6e1beeb740ad2128a99f850a222ce85827a8070fdd54c7b286","artifact_url":"https://raw.githubusercontent.com/Strong-Lab/Viral_Classification_in_Metagenomics/f583cbff6b022ce3a7e3870003e22e14769566fa/LICENSE","retrieved_at":"2026-09-16T20:30:16.571343+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/Strong-Lab/Viral_Classification_in_Metagenomics/blob/f583cbff6b022ce3a7e3870003e22e14769566fa/LICENSE","version":"f583cbff6b022ce3a7e3870003e22e14769566fa"}}
{"id":"evidence-reported-viral-contig-simulation-2021-readme-md","kind":"source","name":"Strong-Lab/Viral_Classification_in_Metagenomics README.md","description":"Primary-source artifact inspected during the configuration-profile evidence review. Its licence and scope are not inherited by unrelated models or data.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"eb2f64d6845419c0e643cbe03e9c15b28d2860bf4fe49a0a640b328d50b09f19","artifact_url":"https://raw.githubusercontent.com/Strong-Lab/Viral_Classification_in_Metagenomics/f583cbff6b022ce3a7e3870003e22e14769566fa/README.md","retrieved_at":"2026-09-16T20:30:16.571343+00:00","review_method":"automated_source_review","review_scope":"Explanatory metadata only; no new numerical result or independent reproduction.","url":"https://github.com/Strong-Lab/Viral_Classification_in_Metagenomics/blob/f583cbff6b022ce3a7e3870003e22e14769566fa/README.md","version":"f583cbff6b022ce3a7e3870003e22e14769566fa"}}
{"id":"evidence-task-final-a-cammiq-2022-41467-2022-33869-moesm1-esm-pdf","kind":"source","name":"cammiq-2022__41467_2022_33869_MOESM1_ESM.pdf","description":"Primary-source artifact inspected for task-specific methodological metadata. The hash identifies the exact downloaded artifact or named member of its supplementary archive.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_member":"41467_2022_33869_MOESM1_ESM.pdf","artifact_sha256":"910aed130f3b4648b0758bdcc6b82d1e2d38ddb320ea80d670c98c565930610b","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC9616933/supplementaryFiles","retrieved_at":"2026-09-16T21:08:58.951180+00:00","review_method":"automated_source_review","review_scope":"Task metadata and methodological context; no numerical result changed or independently reproduced.","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC9616933/supplementaryFiles","version":"Retrieved 2026-09-16; sha256:910aed130f3b4648b0758bdcc6b82d1e2d38ddb320ea80d670c98c565930610b"}}
{"id":"evidence-task-final-a-cell-dino-2025-pcbi-1013828-s001-pdf","kind":"source","name":"cell-dino-2025__pcbi.1013828.s001.pdf","description":"Primary-source artifact inspected for task-specific methodological metadata. The hash identifies the exact downloaded artifact or named member of its supplementary archive.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_member":"pcbi.1013828.s001.pdf","artifact_sha256":"e19d3d5c8d9dc1abcaf699c571349d869a1a7fa8e293660f49f67819f4b4d283","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12826486/supplementaryFiles","retrieved_at":"2026-09-16T21:06:02.800324+00:00","review_method":"automated_source_review","review_scope":"Task metadata and methodological context; no numerical result changed or independently reproduced.","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12826486/supplementaryFiles","version":"Retrieved 2026-09-16; sha256:e19d3d5c8d9dc1abcaf699c571349d869a1a7fa8e293660f49f67819f4b4d283"}}
{"id":"evidence-task-final-a-clathrin-dataset-clathrin0-6-csv","kind":"source","name":"clathrin__Dataset__Clathrin0.6.csv","description":"Primary-source artifact inspected for task-specific methodological metadata. The hash identifies the exact downloaded artifact or named member of its supplementary archive.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"0c85f84ca65d8b4c2fbeaa92e33a2015ce3c7096bfa2b948d73c8716f8b8c739","artifact_url":"https://raw.githubusercontent.com/lawankorn-m/Clathrin/9a8f55bc008401180152864560d8c8528600fe71/Dataset/Clathrin0.6.csv","retrieved_at":"2026-09-16T21:12:58.536073+00:00","review_method":"automated_source_review","review_scope":"Task metadata and methodological context; no numerical result changed or independently reproduced.","url":"https://raw.githubusercontent.com/lawankorn-m/Clathrin/9a8f55bc008401180152864560d8c8528600fe71/Dataset/Clathrin0.6.csv","version":"9a8f55bc008401180152864560d8c8528600fe71"}}
{"id":"evidence-task-final-a-clathrin-dataset-clathrin0-7-csv","kind":"source","name":"clathrin__Dataset__Clathrin0.7.csv","description":"Primary-source artifact inspected for task-specific methodological metadata. The hash identifies the exact downloaded artifact or named member of its supplementary archive.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"ba5444b11ed53cc1770e59b1e4b4a6078808ada83765edd418b644359d00184b","artifact_url":"https://raw.githubusercontent.com/lawankorn-m/Clathrin/9a8f55bc008401180152864560d8c8528600fe71/Dataset/Clathrin0.7.csv","retrieved_at":"2026-09-16T21:12:59.067575+00:00","review_method":"automated_source_review","review_scope":"Task metadata and methodological context; no numerical result changed or independently reproduced.","url":"https://raw.githubusercontent.com/lawankorn-m/Clathrin/9a8f55bc008401180152864560d8c8528600fe71/Dataset/Clathrin0.7.csv","version":"9a8f55bc008401180152864560d8c8528600fe71"}}
{"id":"evidence-task-final-a-clathrin-dataset-clathrin1-0-csv","kind":"source","name":"clathrin__Dataset__Clathrin1.0.csv","description":"Primary-source artifact inspected for task-specific methodological metadata. The hash identifies the exact downloaded artifact or named member of its supplementary archive.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"260c306da68247a283e55c97b96de2750a98b459c91408b85d161f720cad6a79","artifact_url":"https://raw.githubusercontent.com/lawankorn-m/Clathrin/9a8f55bc008401180152864560d8c8528600fe71/Dataset/Clathrin1.0.csv","retrieved_at":"2026-09-16T21:12:59.571978+00:00","review_method":"automated_source_review","review_scope":"Task metadata and methodological context; no numerical result changed or independently reproduced.","url":"https://raw.githubusercontent.com/lawankorn-m/Clathrin/9a8f55bc008401180152864560d8c8528600fe71/Dataset/Clathrin1.0.csv","version":"9a8f55bc008401180152864560d8c8528600fe71"}}
{"id":"evidence-task-final-a-clathrin-readme-md","kind":"source","name":"clathrin__README.md","description":"Primary-source artifact inspected for task-specific methodological metadata. The hash identifies the exact downloaded artifact or named member of its supplementary archive.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"341256ca55892a34c03617a6748f0b285f9b0bc4bc1d68c6c462c66efce59e50","artifact_url":"https://raw.githubusercontent.com/lawankorn-m/Clathrin/9a8f55bc008401180152864560d8c8528600fe71/README.md","retrieved_at":"2026-09-16T21:12:59.799010+00:00","review_method":"automated_source_review","review_scope":"Task metadata and methodological context; no numerical result changed or independently reproduced.","url":"https://raw.githubusercontent.com/lawankorn-m/Clathrin/9a8f55bc008401180152864560d8c8528600fe71/README.md","version":"9a8f55bc008401180152864560d8c8528600fe71"}}
{"id":"evidence-task-final-a-deepinteraware-2025-advs-12-2412533-s001-pdf","kind":"source","name":"deepinteraware-2025__ADVS-12-2412533-s001.pdf","description":"Primary-source artifact inspected for task-specific methodological metadata. The hash identifies the exact downloaded artifact or named member of its supplementary archive.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_member":"ADVS-12-2412533-s001.pdf","artifact_sha256":"3c66d0d9561d744ca0e42a5d33c020719e6f46d7ed27e37a25c19c2bd8d89345","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11967782/supplementaryFiles","retrieved_at":"2026-09-16T21:05:56.618962+00:00","review_method":"automated_source_review","review_scope":"Task metadata and methodological context; no numerical result changed or independently reproduced.","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11967782/supplementaryFiles","version":"Retrieved 2026-09-16; sha256:3c66d0d9561d744ca0e42a5d33c020719e6f46d7ed27e37a25c19c2bd8d89345"}}
{"id":"evidence-task-final-a-enbed-2024-vbae117-supplementary-data-pdf","kind":"source","name":"enbed-2024__vbae117_supplementary_data.pdf","description":"Primary-source artifact inspected for task-specific methodological metadata. The hash identifies the exact downloaded artifact or named member of its supplementary archive.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_member":"vbae117_supplementary_data.pdf","artifact_sha256":"a169bbebba2298e9c98c33c28053c1a0b42d0c3e88e5a8795934725c4408cc32","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11341122/supplementaryFiles","retrieved_at":"2026-09-16T21:05:56.086158+00:00","review_method":"automated_source_review","review_scope":"Task metadata and methodological context; no numerical result changed or independently reproduced.","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11341122/supplementaryFiles","version":"Retrieved 2026-09-16; sha256:a169bbebba2298e9c98c33c28053c1a0b42d0c3e88e5a8795934725c4408cc32"}}
{"id":"evidence-task-final-a-gse108394-soft","kind":"source","name":"GSE108394.soft","description":"Primary-source artifact inspected for task-specific methodological metadata. The hash identifies the exact downloaded artifact or named member of its supplementary archive.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"e4b44c1b50d2a0e7a764630ea901d7848c72dc55cc66e9faf8b706154719fc17","artifact_url":"https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE108394&targ=self&form=text&view=quick","retrieved_at":"2026-09-16T21:11:57.422218+00:00","review_method":"automated_source_review","review_scope":"Task metadata and methodological context; no numerical result changed or independently reproduced.","url":"https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE108394&targ=self&form=text&view=quick","version":"Retrieved 2026-09-16; sha256:e4b44c1b50d2a0e7a764630ea901d7848c72dc55cc66e9faf8b706154719fc17"}}
{"id":"evidence-task-final-a-gse117872-soft","kind":"source","name":"GSE117872.soft","description":"Primary-source artifact inspected for task-specific methodological metadata. The hash identifies the exact downloaded artifact or named member of its supplementary archive.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"f615587bbf6e89a29811400538966cdf008dc4a7719091fc44448ba4a4aa8f5c","artifact_url":"https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE117872&targ=self&form=text&view=quick","retrieved_at":"2026-09-16T21:11:57.754129+00:00","review_method":"automated_source_review","review_scope":"Task metadata and methodological context; no numerical result changed or independently reproduced.","url":"https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE117872&targ=self&form=text&view=quick","version":"Retrieved 2026-09-16; sha256:f615587bbf6e89a29811400538966cdf008dc4a7719091fc44448ba4a4aa8f5c"}}
{"id":"evidence-task-final-a-gse127298-soft","kind":"source","name":"GSE127298.soft","description":"Primary-source artifact inspected for task-specific methodological metadata. The hash identifies the exact downloaded artifact or named member of its supplementary archive.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"40a50eb13981d4d9ca3dfcd987c6b4b90289385e6aaeea645c5672a2dc6a9f19","artifact_url":"https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE127298&targ=self&form=text&view=quick","retrieved_at":"2026-09-16T21:11:57.862686+00:00","review_method":"automated_source_review","review_scope":"Task metadata and methodological context; no numerical result changed or independently reproduced.","url":"https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE127298&targ=self&form=text&view=quick","version":"Retrieved 2026-09-16; sha256:40a50eb13981d4d9ca3dfcd987c6b4b90289385e6aaeea645c5672a2dc6a9f19"}}
{"id":"evidence-task-final-a-gse134839-soft","kind":"source","name":"GSE134839.soft","description":"Primary-source artifact inspected for task-specific methodological metadata. The hash identifies the exact downloaded artifact or named member of its supplementary archive.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"333ba9b2538c6ba1f41ae08d60b41099f26a443d2b473cbd10104d887ef20064","artifact_url":"https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE134839&targ=self&form=text&view=quick","retrieved_at":"2026-09-16T21:11:57.450992+00:00","review_method":"automated_source_review","review_scope":"Task metadata and methodological context; no numerical result changed or independently reproduced.","url":"https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE134839&targ=self&form=text&view=quick","version":"Retrieved 2026-09-16; sha256:333ba9b2538c6ba1f41ae08d60b41099f26a443d2b473cbd10104d887ef20064"}}
{"id":"evidence-task-final-a-gse140440-soft","kind":"source","name":"GSE140440.soft","description":"Primary-source artifact inspected for task-specific methodological metadata. The hash identifies the exact downloaded artifact or named member of its supplementary archive.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"7b6ca08ec013e2e79b367727f6ce4ff46dfb7ef9aa166bc4907dd02fbbb472e5","artifact_url":"https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE140440&targ=self&form=text&view=quick","retrieved_at":"2026-09-16T21:11:57.808765+00:00","review_method":"automated_source_review","review_scope":"Task metadata and methodological context; no numerical result changed or independently reproduced.","url":"https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE140440&targ=self&form=text&view=quick","version":"Retrieved 2026-09-16; sha256:7b6ca08ec013e2e79b367727f6ce4ff46dfb7ef9aa166bc4907dd02fbbb472e5"}}
{"id":"evidence-task-final-a-gse147326-soft","kind":"source","name":"GSE147326.soft","description":"Primary-source artifact inspected for task-specific methodological metadata. The hash identifies the exact downloaded artifact or named member of its supplementary archive.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"6882efe837a56064645705857c81f21e6f82b773a8a7fd9679180f71bba9bb74","artifact_url":"https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE147326&targ=self&form=text&view=quick","retrieved_at":"2026-09-16T21:11:57.732001+00:00","review_method":"automated_source_review","review_scope":"Task metadata and methodological context; no numerical result changed or independently reproduced.","url":"https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE147326&targ=self&form=text&view=quick","version":"Retrieved 2026-09-16; sha256:6882efe837a56064645705857c81f21e6f82b773a8a7fd9679180f71bba9bb74"}}
{"id":"evidence-task-final-a-gse149214-soft","kind":"source","name":"GSE149214.soft","description":"Primary-source artifact inspected for task-specific methodological metadata. The hash identifies the exact downloaded artifact or named member of its supplementary archive.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"801532dfc1983e4de5b4183d39e81f74825fa0b22632bb79322524d3f6139d1e","artifact_url":"https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE149214&targ=self&form=text&view=quick","retrieved_at":"2026-09-16T21:11:57.291347+00:00","review_method":"automated_source_review","review_scope":"Task metadata and methodological context; no numerical result changed or independently reproduced.","url":"https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE149214&targ=self&form=text&view=quick","version":"Retrieved 2026-09-16; sha256:801532dfc1983e4de5b4183d39e81f74825fa0b22632bb79322524d3f6139d1e"}}
{"id":"evidence-task-final-a-gse164614-soft","kind":"source","name":"GSE164614.soft","description":"Primary-source artifact inspected for task-specific methodological metadata. The hash identifies the exact downloaded artifact or named member of its supplementary archive.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"132e5ca2603b5e5d91c16e708f8928245086d43d13f8f56ec5512373c4248e2c","artifact_url":"https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE164614&targ=self&form=text&view=quick","retrieved_at":"2026-09-16T21:11:57.455257+00:00","review_method":"automated_source_review","review_scope":"Task metadata and methodological context; no numerical result changed or independently reproduced.","url":"https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE164614&targ=self&form=text&view=quick","version":"Retrieved 2026-09-16; sha256:132e5ca2603b5e5d91c16e708f8928245086d43d13f8f56ec5512373c4248e2c"}}
{"id":"evidence-task-final-a-gse230538-soft","kind":"source","name":"GSE230538.soft","description":"Primary-source artifact inspected for task-specific methodological metadata. The hash identifies the exact downloaded artifact or named member of its supplementary archive.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"63d3b23a3c1266d5cdbd04ca1f4766774ab8427e70b41dbc80e230275286cf0d","artifact_url":"https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE230538&targ=self&form=text&view=quick","retrieved_at":"2026-09-16T21:11:57.435478+00:00","review_method":"automated_source_review","review_scope":"Task metadata and methodological context; no numerical result changed or independently reproduced.","url":"https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE230538&targ=self&form=text&view=quick","version":"Retrieved 2026-09-16; sha256:63d3b23a3c1266d5cdbd04ca1f4766774ab8427e70b41dbc80e230275286cf0d"}}
{"id":"evidence-task-final-a-hi-enhancer-2025-btaf441-supplementary-data-docx","kind":"source","name":"hi-enhancer-2025__btaf441_supplementary_data.docx","description":"Primary-source artifact inspected for task-specific methodological metadata. The hash identifies the exact downloaded artifact or named member of its supplementary archive.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_member":"btaf441_supplementary_data.docx","artifact_sha256":"26de5de88996a9da7723ba4036eb4cf234a77fae3bbab833edffbc7d99ec56a5","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12758598/supplementaryFiles","retrieved_at":"2026-09-16T21:05:56.618187+00:00","review_method":"automated_source_review","review_scope":"Task metadata and methodological context; no numerical result changed or independently reproduced.","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12758598/supplementaryFiles","version":"Retrieved 2026-09-16; sha256:26de5de88996a9da7723ba4036eb4cf234a77fae3bbab833edffbc7d99ec56a5"}}
{"id":"evidence-task-final-a-mrna-lm-2025-gkaf044-supplemental-file-pdf","kind":"source","name":"mrna-lm-2025__gkaf044_Supplemental_File.pdf","description":"Primary-source artifact inspected for task-specific methodological metadata. The hash identifies the exact downloaded artifact or named member of its supplementary archive.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_member":"gkaf044_Supplemental_File.pdf","artifact_sha256":"bf1e156bb09c90a9e18101a45332e5a7a8f3b7fd363f6ffbb7760020b4c8f40f","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11962594/supplementaryFiles","retrieved_at":"2026-09-16T21:08:59.937487+00:00","review_method":"automated_source_review","review_scope":"Task metadata and methodological context; no numerical result changed or independently reproduced.","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11962594/supplementaryFiles","version":"Retrieved 2026-09-16; sha256:bf1e156bb09c90a9e18101a45332e5a7a8f3b7fd363f6ffbb7760020b4c8f40f"}}
{"id":"evidence-task-final-a-plantcad2-2025-media-1-xlsx","kind":"source","name":"plantcad2-2025__media-1.xlsx","description":"Primary-source artifact inspected for task-specific methodological metadata. The hash identifies the exact downloaded artifact or named member of its supplementary archive.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_member":"media-1.xlsx","artifact_sha256":"f551e80bf044a0ea8ccc2e6f443fdf3690ca5e4a54d0e4bb4b2e6470ffc944be","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12425018/supplementaryFiles","retrieved_at":"2026-09-16T21:05:57.965868+00:00","review_method":"automated_source_review","review_scope":"Task metadata and methodological context; no numerical result changed or independently reproduced.","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12425018/supplementaryFiles","version":"Retrieved 2026-09-16; sha256:f551e80bf044a0ea8ccc2e6f443fdf3690ca5e4a54d0e4bb4b2e6470ffc944be"}}
{"id":"evidence-task-final-a-pmc6731122-xml","kind":"source","name":"PMC6731122.xml","description":"Primary-source artifact inspected for task-specific methodological metadata. The hash identifies the exact downloaded artifact or named member of its supplementary archive.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"2136d6c7589d2573e33c69f2b1eb0f2b76c696da06ee1c14427ccbaa45a33d5f","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC6731122/fullTextXML","retrieved_at":"2026-09-16T21:08:48.254469+00:00","review_method":"automated_source_review","review_scope":"Task metadata and methodological context; no numerical result changed or independently reproduced.","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC6731122/fullTextXML","version":"Retrieved 2026-09-16; sha256:2136d6c7589d2573e33c69f2b1eb0f2b76c696da06ee1c14427ccbaa45a33d5f"}}
{"id":"evidence-task-final-a-pmc7331607-xml","kind":"source","name":"PMC7331607.xml","description":"Primary-source artifact inspected for task-specific methodological metadata. The hash identifies the exact downloaded artifact or named member of its supplementary archive.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"d87b7b7a67445c04db4f1fbb0cc52914b1bccd04a23d4ff6d66e2be5373291c6","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC7331607/fullTextXML","retrieved_at":"2026-09-16T21:08:47.387344+00:00","review_method":"automated_source_review","review_scope":"Task metadata and methodological context; no numerical result changed or independently reproduced.","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC7331607/fullTextXML","version":"Retrieved 2026-09-16; sha256:d87b7b7a67445c04db4f1fbb0cc52914b1bccd04a23d4ff6d66e2be5373291c6"}}
{"id":"evidence-task-final-a-pmc7912887-xml","kind":"source","name":"PMC7912887.xml","description":"Primary-source artifact inspected for task-specific methodological metadata. The hash identifies the exact downloaded artifact or named member of its supplementary archive.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"3504a0fa3b62698b7eff78418e682bf2484236ed05645597a8be4fa2e97b3efc","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC7912887/fullTextXML","retrieved_at":"2026-09-16T21:08:40.609030+00:00","review_method":"automated_source_review","review_scope":"Task metadata and methodological context; no numerical result changed or independently reproduced.","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC7912887/fullTextXML","version":"Retrieved 2026-09-16; sha256:3504a0fa3b62698b7eff78418e682bf2484236ed05645597a8be4fa2e97b3efc"}}
{"id":"evidence-task-final-a-pmc9556750-xml","kind":"source","name":"PMC9556750.xml","description":"Primary-source artifact inspected for task-specific methodological metadata. The hash identifies the exact downloaded artifact or named member of its supplementary archive.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_sha256":"71fd84bba280e9ac7b1009600432245122e73f6a01916a743ef4faeb56bb26f0","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC9556750/fullTextXML","retrieved_at":"2026-09-16T21:08:44.485140+00:00","review_method":"automated_source_review","review_scope":"Task metadata and methodological context; no numerical result changed or independently reproduced.","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC9556750/fullTextXML","version":"Retrieved 2026-09-16; sha256:71fd84bba280e9ac7b1009600432245122e73f6a01916a743ef4faeb56bb26f0"}}
{"id":"evidence-task-final-a-rnaret-2026-42003-2026-9757-moesm2-esm-pdf","kind":"source","name":"rnaret-2026__42003_2026_9757_MOESM2_ESM.pdf","description":"Primary-source artifact inspected for task-specific methodological metadata. The hash identifies the exact downloaded artifact or named member of its supplementary archive.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_member":"42003_2026_9757_MOESM2_ESM.pdf","artifact_sha256":"8464ff052a60b946bd08fa18860f22b2f390fe93dd0549eb845c8e9a68cf5fcc","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC13111708/supplementaryFiles","retrieved_at":"2026-09-16T21:06:05.136898+00:00","review_method":"automated_source_review","review_scope":"Task metadata and methodological context; no numerical result changed or independently reproduced.","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC13111708/supplementaryFiles","version":"Retrieved 2026-09-16; sha256:8464ff052a60b946bd08fa18860f22b2f390fe93dd0549eb845c8e9a68cf5fcc"}}
{"id":"evidence-task-final-a-scxdr-2026-42003-2025-9418-moesm1-esm-pdf","kind":"source","name":"scxdr-2026__42003_2025_9418_MOESM1_ESM.pdf","description":"Primary-source artifact inspected for task-specific methodological metadata. The hash identifies the exact downloaded artifact or named member of its supplementary archive.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"artifact_member":"42003_2025_9418_MOESM1_ESM.pdf","artifact_sha256":"ad53da81235ba47f762c93ac5140108241b857160d978c63ff934e8e57283758","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12859067/supplementaryFiles","retrieved_at":"2026-09-16T21:05:58.741824+00:00","review_method":"automated_source_review","review_scope":"Task metadata and methodological context; no numerical result changed or independently reproduced.","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12859067/supplementaryFiles","version":"Retrieved 2026-09-16; sha256:ad53da81235ba47f762c93ac5140108241b857160d978c63ff934e8e57283758"}}
{"id":"evidence-task-final-b-birna-supplement","kind":"source","name":"birna journal Supplementary Information — pinned PDF","description":"Published supplementary PDF inspected for task-specific computational evaluation metadata.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12635123/supplementaryFiles","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12635123/supplementaryFiles","artifact_sha256":"7897d4dcf456d1f22b0631beabf7c5fd8678d0b8765cd40325195eb5addf4493","archive_member":"42003_2025_8982_MOESM2_ESM.pdf","archive_sha256":"f30f2e52b0ef64680b9a544b9db444d32ef302b7d9e09682b1ca339ae9086d23","version":"Published supplementary PDF 42003_2025_8982_MOESM2_ESM.pdf; sha256:7897d4dcf456d1f22b0631beabf7c5fd8678d0b8765cd40325195eb5addf4493","retrieved_at":"2026-09-16T21:06:10.816352+00:00","review_scope":"Task-specific evaluation passages and reporting scope checked; no numerical results reproduced."}}
{"id":"evidence-task-final-b-mrnabert-supplement","kind":"source","name":"mrnabert journal Supplementary Information — pinned PDF","description":"Published supplementary PDF inspected for task-specific computational evaluation metadata.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12644827/supplementaryFiles","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12644827/supplementaryFiles","artifact_sha256":"2820d9385389e41e9213223608b26c84bf6bffbba57ff48fd5dda17fb78a8aed","archive_member":"41467_2025_65340_MOESM1_ESM.pdf","archive_sha256":"c89fe42941b7cc5a4ec7007ba236d500c3a14d4c925ac6e92d58be1307666538","version":"Published supplementary PDF 41467_2025_65340_MOESM1_ESM.pdf; sha256:2820d9385389e41e9213223608b26c84bf6bffbba57ff48fd5dda17fb78a8aed","retrieved_at":"2026-09-16T21:06:12.558631+00:00","review_scope":"Task-specific evaluation passages and reporting scope checked; no numerical results reproduced."}}
{"id":"evidence-task-final-b-pst-supplement","kind":"source","name":"pst journal Supplementary Information — pinned PDF","description":"Published supplementary PDF inspected for task-specific computational evaluation metadata.","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12603367/supplementaryFiles","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12603367/supplementaryFiles","artifact_sha256":"f8c3375644f10aeb50c76f99071ec64b2b0dceadcee7f70a809adbcc8985d936","archive_member":"btaf582_supplementary_data.pdf","archive_sha256":"6fbc2e0c0f3a5d5128bd7a2f968b838de3485ce1ba6b090694c130c3e2051549","version":"Published supplementary PDF btaf582_supplementary_data.pdf; sha256:f8c3375644f10aeb50c76f99071ec64b2b0dceadcee7f70a809adbcc8985d936","retrieved_at":"2026-09-16T21:06:10.146345+00:00","review_scope":"Task-specific evaluation passages and reporting scope checked; no numerical results reproduced."}}
{"id":"fingerprint-scoring-2022","kind":"source","name":"Machine-Learning- and Knowledge-Based Scoring Functions Incorporating Ligand and Protein Fingerprints","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["molecular-interactions"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC9178954/","version":"PMC archival version PMC9178954.1","retrieved_at":"2026-09-15T23:29:32Z","doi":"10.1021/acsomega.2c02822","publication_status":"peer_reviewed","year":2022,"artifact_sha256":"47bd60c6392b801095fdb604de06c0d4bda6f555bae58e9955e491a5abf60576","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC9178954/fullTextXML","artifact_retrieved_at":"2026-09-16T10:44:03.439460+00:00","legacy_paper":{"id":"fingerprint-scoring-2022","title":"Machine-Learning- and Knowledge-Based Scoring Functions Incorporating Ligand and Protein Fingerprints","year":2022,"publication_status":"peer_reviewed","version":"PMC archival version PMC9178954.1","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC9178954/","primary_domain":"molecular-interactions","retrieved_utc":"2026-09-15T23:29:32Z","notes":"Primary full text verified using Europe PMC XML; venue: ACS Omega; PMC ID: PMC9178954.","doi":"10.1021/acsomega.2c02822"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"fujisan-2024","kind":"source","name":"Enhanced prediction of protein functional identity through the integration of sequence and structural features","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["proteins-complexes"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11609699/","version":"PMC11609699.1","retrieved_at":"2026-09-15T23:25:00Z","doi":"10.1016/j.csbj.2024.11.028","publication_status":"peer_reviewed","year":2024,"artifact_sha256":"db33e0542005ffae00cd644dfe697185b94c8823d5aee2768620a6db0c48e56f","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11609699/fullTextXML","artifact_retrieved_at":"2026-09-16T10:33:35.728Z","legacy_paper":{"id":"fujisan-2024","title":"Enhanced prediction of protein functional identity through the integration of sequence and structural features","year":2024,"publication_status":"peer_reviewed","version":"PMC11609699.1","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11609699/","primary_domain":"proteins-complexes","retrieved_utc":"2026-09-15T23:25:00Z","notes":"Primary full text via Europe PMC XML; venue: Computational and Structural Biotechnology Journal; PMC ID: PMC11609699.","doi":"10.1016/j.csbj.2024.11.028"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"fusion-breakpoint-foundation-models-2026","kind":"source","name":"Benchmarking genomic foundation models for binary classification of gene fusion breakpoints from DNA sequences","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["dna-genomes"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC13182013/","version":"journal full text in PMC","retrieved_at":"2026-09-15T23:33:26Z","doi":"10.1186/s13040-026-00553-1","publication_status":"peer_reviewed","year":2026,"artifact_sha256":"0f4d9de77f1e39cfd2164a20653d86370767da684dc22d17e09f589761abeb5f","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC13182013/fullTextXML","artifact_retrieved_at":"2026-09-16T10:38:57.558209+00:00","legacy_paper":{"id":"fusion-breakpoint-foundation-models-2026","title":"Benchmarking genomic foundation models for binary classification of gene fusion breakpoints from DNA sequences","year":2026,"publication_status":"peer_reviewed","version":"journal full text in PMC","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC13182013/","primary_domain":"dna-genomes","retrieved_utc":"2026-09-15T23:33:26Z","doi":"10.1186/s13040-026-00553-1","notes":"Numeric result checked against Table 2 in primary full-text XML; journal/source: BioData Mining."},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"genept-2024","kind":"source","name":"GenePT: A Simple But Effective Foundation Model for Genes and Cells Built From ChatGPT","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["cells-tissues"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC10614824/","version":"PMC archival version PMC10614824.2","retrieved_at":"2026-09-15T23:29:32Z","doi":"10.1101/2023.10.16.562533","publication_status":"preprint","year":2024,"artifact_sha256":"230a2ec55458d9243eaeeebf3244df7409eb02d47f4b809ee56a06dcb6fdd047","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC10614824/fullTextXML","artifact_retrieved_at":"2026-09-16T10:44:03.399274+00:00","legacy_paper":{"id":"genept-2024","title":"GenePT: A Simple But Effective Foundation Model for Genes and Cells Built From ChatGPT","year":2024,"publication_status":"preprint","version":"PMC archival version PMC10614824.2","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC10614824/","primary_domain":"cells-tissues","retrieved_utc":"2026-09-15T23:29:32Z","notes":"Primary full text verified using Europe PMC XML; venue: bioRxiv; PMC ID: PMC10614824.","doi":"10.1101/2023.10.16.562533"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"genomeocean-2025","kind":"source","name":"GenomeOcean: An Efficient Genome Foundation Model Trained on Large-Scale Metagenomic Assemblies","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["microbes-communities"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11838515/","version":"preprint archived 2025-02-05","retrieved_at":"2026-09-15T23:29:32Z","doi":"10.1101/2025.01.30.635558","publication_status":"preprint","year":2025,"artifact_sha256":"3cc0df52522fccda23e3958f069c916b87ee50bb5c9a992fa37e25256546e145","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11838515/fullTextXML","artifact_retrieved_at":"2026-09-16T10:33:55.224Z","legacy_paper":{"id":"genomeocean-2025","title":"GenomeOcean: An Efficient Genome Foundation Model Trained on Large-Scale Metagenomic Assemblies","year":2025,"publication_status":"preprint","version":"preprint archived 2025-02-05","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11838515/","primary_domain":"microbes-communities","retrieved_utc":"2026-09-15T23:29:32Z","notes":"Primary full text verified using Europe PMC XML; venue: bioRxiv; PMC ID: PMC11838515.","doi":"10.1101/2025.01.30.635558"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"genomic-tokenizer-selection-2025","kind":"source","name":"The impact of tokenizer selection in genomic language models","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["dna-genomes"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12453675/","version":"journal full text in PMC","retrieved_at":"2026-09-15T23:33:26Z","doi":"10.1093/bioinformatics/btaf456","publication_status":"peer_reviewed","year":2025,"artifact_sha256":"0a01c36fdd63f3f6db509777e61c3f87e8a298c810f8aef7974915aaa0655342","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12453675/fullTextXML","artifact_retrieved_at":"2026-09-16T10:38:57.558210+00:00","legacy_paper":{"id":"genomic-tokenizer-selection-2025","title":"The impact of tokenizer selection in genomic language models","year":2025,"publication_status":"peer_reviewed","version":"journal full text in PMC","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12453675/","primary_domain":"dna-genomes","retrieved_utc":"2026-09-15T23:33:26Z","doi":"10.1093/bioinformatics/btaf456","notes":"Final Bioinformatics journal article Table 2, Caduceus (char) Regulatory MCC 0.778 checked directly; same study also has a bioRxiv manuscript."},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"gremln-2026","kind":"source","name":"GREmLN: A Cellular Graph Structure Aware Transcriptomics Foundation Model","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["cells-tissues"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC13060794/","version":"preprint version in PMC","retrieved_at":"2026-09-15T23:37:05Z","doi":"10.1101/2025.07.03.663009","publication_status":"preprint","year":2026,"artifact_sha256":"3a20c4ededb749fc3f1120baf16dcfebe3fcb30418a91c445cfd91a7b5fdf553","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC13060794/fullTextXML","artifact_retrieved_at":"2026-09-16T10:33:57.502Z","legacy_paper":{"id":"gremln-2026","title":"GREmLN: A Cellular Graph Structure Aware Transcriptomics Foundation Model","year":2026,"publication_status":"preprint","version":"preprint version in PMC","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC13060794/","primary_domain":"cells-tissues","retrieved_utc":"2026-09-15T23:37:05Z","notes":"Primary full text verified via Europe PMC fullTextXML; venue: bioRxiv; PMC ID: PMC13060794. Preprint; table labels metric F1; paper does not specify macro in this row.","doi":"10.1101/2025.07.03.663009"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"gsmformer-ppi-2026","kind":"source","name":"Multimodal graph, surface, and language-based model for protein protein interaction prediction","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["proteins-complexes"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12873117/","version":"journal full text in PMC","retrieved_at":"2026-09-15T23:33:26Z","doi":"10.1038/s41598-025-34758-x","publication_status":"peer_reviewed","year":2026,"artifact_sha256":"9b364b5d73d16f2787f93f78f17dbe98b954ab9c2c64c1df960eec2e615eb3b4","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12873117/fullTextXML","artifact_retrieved_at":"2026-09-16T10:38:57.558212+00:00","legacy_paper":{"id":"gsmformer-ppi-2026","title":"Multimodal graph, surface, and language-based model for protein protein interaction prediction","year":2026,"publication_status":"peer_reviewed","version":"journal full text in PMC","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12873117/","primary_domain":"proteins-complexes","retrieved_utc":"2026-09-15T23:33:26Z","doi":"10.1038/s41598-025-34758-x","notes":"Numeric result checked against Table 6 in primary full-text XML; journal/source: Scientific Reports."},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"hi-enhancer-2025","kind":"source","name":"Hi-Enhancer: a two-stage framework for prediction and localization of enhancers based on Blending-KAN and Stacking-Auto models","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["dna-genomes"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12758598/","version":"version of record","retrieved_at":"2026-09-15T23:37:05Z","doi":"10.1093/bioinformatics/btaf441","publication_status":"peer_reviewed","year":2025,"artifact_sha256":"c86488c9f60329b7a3c4370598e7a0a9e4c8c45d1758b87007bfc8242376b009","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12758598/fullTextXML","artifact_retrieved_at":"2026-09-16T10:41:06Z","legacy_paper":{"id":"hi-enhancer-2025","title":"Hi-Enhancer: a two-stage framework for prediction and localization of enhancers based on Blending-KAN and Stacking-Auto models","year":2025,"publication_status":"peer_reviewed","version":"version of record","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12758598/","primary_domain":"dna-genomes","retrieved_utc":"2026-09-15T23:37:05Z","notes":"Primary full text verified via Europe PMC fullTextXML; venue: Bioinformatics; PMC ID: PMC12758598. Task-specific enhancer predictor; not a DNA foundation model. Comparison values from older papers excluded.","doi":"10.1093/bioinformatics/btaf441"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"ibex-2025","kind":"source","name":"Conformation-aware structure prediction of antigen-recognizing immune proteins","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["molecular-interactions"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12710905/","version":"PMC archival version PMC12710905.1","retrieved_at":"2026-09-15T23:29:32Z","doi":"10.1080/19420862.2025.2602217","publication_status":"peer_reviewed","year":2025,"artifact_sha256":"caa1109bd5fe7f6be703aa9d4afd6f4f1522bcbce6b7361650eb59618c2a9e14","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12710905/fullTextXML","artifact_retrieved_at":"2026-09-16T10:44:03.426811+00:00","legacy_paper":{"id":"ibex-2025","title":"Conformation-aware structure prediction of antigen-recognizing immune proteins","year":2025,"publication_status":"peer_reviewed","version":"PMC archival version PMC12710905.1","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12710905/","primary_domain":"molecular-interactions","retrieved_utc":"2026-09-15T23:29:32Z","notes":"Primary full text verified using Europe PMC XML; venue: mAbs; PMC ID: PMC12710905.","doi":"10.1080/19420862.2025.2602217"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"icctax-2025","kind":"source","name":"ICCTax: a hierarchical taxonomic classifier for metagenomic sequences on a large language model","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["microbes-communities"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12619997/","version":"version of record","retrieved_at":"2026-09-15T23:29:32Z","doi":"10.1093/bioadv/vbaf257","publication_status":"peer_reviewed","year":2025,"artifact_sha256":"2ce0b48f1cde3aea7e561d92f4d7dc1525af7439ccd16f80bec0773e8812c8ec","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12619997/fullTextXML","artifact_retrieved_at":"2026-09-16T10:33:55.373Z","legacy_paper":{"id":"icctax-2025","title":"ICCTax: a hierarchical taxonomic classifier for metagenomic sequences on a large language model","year":2025,"publication_status":"peer_reviewed","version":"version of record","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12619997/","primary_domain":"microbes-communities","retrieved_utc":"2026-09-15T23:29:32Z","notes":"Primary full text verified using Europe PMC XML; venue: Bioinformatics Advances; PMC ID: PMC12619997.","doi":"10.1093/bioadv/vbaf257"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"insilico-perturbation-auprc-2025","kind":"source","name":"AUPRC: a metric for evaluating the performance of in-silico perturbation methods in identifying differentially expressed genes","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["cells-tissues"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12400816/","version":"PMC archival version PMC12400816.1","retrieved_at":"2026-09-15T23:37:05Z","doi":"10.1093/bib/bbaf426","publication_status":"peer_reviewed","year":2025,"artifact_sha256":"2715709d94f84744afa32cafdcaa72efd206d63af8c60afe7619b2cb90108b6b","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12400816/fullTextXML","artifact_retrieved_at":"2026-09-16T10:41:06Z","legacy_paper":{"id":"insilico-perturbation-auprc-2025","title":"AUPRC: a metric for evaluating the performance of in-silico perturbation methods in identifying differentially expressed genes","year":2025,"publication_status":"peer_reviewed","version":"PMC archival version PMC12400816.1","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12400816/","primary_domain":"cells-tissues","retrieved_utc":"2026-09-15T23:37:05Z","notes":"Primary full text verified via Europe PMC fullTextXML; venue: Briefings in Bioinformatics; PMC ID: PMC12400816. Paper benchmarks metrics and scGen perturbation method; no foundation-model result in this row.","doi":"10.1093/bib/bbaf426"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"ipromp-2025","kind":"source","name":"iPro-MP: a BERT-based model to predict multiple prokaryotic promoters","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["microbes-communities"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12516880/","version":"version of record","retrieved_at":"2026-09-15T23:29:32Z","doi":"10.1186/s13059-025-03819-9","publication_status":"peer_reviewed","year":2025,"artifact_sha256":"d21541ee1f7a168da8e4a7c0f0e133c970cbe7bc41118f43a929f08b2fd2afd1","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12516880/fullTextXML","artifact_retrieved_at":"2026-09-16T10:33:55.361Z","legacy_paper":{"id":"ipromp-2025","title":"iPro-MP: a BERT-based model to predict multiple prokaryotic promoters","year":2025,"publication_status":"peer_reviewed","version":"version of record","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12516880/","primary_domain":"microbes-communities","retrieved_utc":"2026-09-15T23:29:32Z","notes":"Primary full text verified using Europe PMC XML; venue: Genome Biology; PMC ID: PMC12516880.","doi":"10.1186/s13059-025-03819-9"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"kmetashot-2025","kind":"source","name":"kMetaShot: a fast and reliable taxonomy classifier for metagenome-assembled genomes","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["microbes-communities"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11695915/","version":"PMC archival version PMC11695915.1","retrieved_at":"2026-09-15T23:29:32Z","doi":"10.1093/bib/bbae680","publication_status":"peer_reviewed","year":2025,"artifact_sha256":"4584e93ea035c1170b8756a0a52cbe99fe72e70bd09b5f1dee639ee104f78247","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11695915/fullTextXML","artifact_retrieved_at":"2026-09-16T10:44:03.417367+00:00","legacy_paper":{"id":"kmetashot-2025","title":"kMetaShot: a fast and reliable taxonomy classifier for metagenome-assembled genomes","year":2025,"publication_status":"peer_reviewed","version":"PMC archival version PMC11695915.1","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11695915/","primary_domain":"microbes-communities","retrieved_utc":"2026-09-15T23:29:32Z","notes":"Primary full text verified using Europe PMC XML; venue: Briefings in Bioinformatics; PMC ID: PMC11695915.","doi":"10.1093/bib/bbae680"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"lambda-prophage-2026","kind":"source","name":"LAMBDA: A Prophage Detection Benchmark for Genomic Language Models","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["microbes-communities"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC13041943/","version":"PMC13041943.1","retrieved_at":"2026-09-15T23:25:00Z","doi":"10.64898/2026.03.26.714501","publication_status":"preprint","year":2026,"artifact_sha256":"22c2e218e87dce757907f6086a0e2ad37c13f785b34fff5bea7cfa1a6c276b16","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC13041943/fullTextXML","artifact_retrieved_at":"2026-09-16T10:33:36.240Z","legacy_paper":{"id":"lambda-prophage-2026","title":"LAMBDA: A Prophage Detection Benchmark for Genomic Language Models","year":2026,"publication_status":"preprint","version":"PMC13041943.1","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC13041943/","primary_domain":"microbes-communities","retrieved_utc":"2026-09-15T23:25:00Z","notes":"Primary full text via Europe PMC XML; venue: bioRxiv; PMC ID: PMC13041943.","doi":"10.64898/2026.03.26.714501"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"lazypipe-2020","kind":"source","name":"Novel NGS pipeline for virus discovery from a wide spectrum of hosts and sample types","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["microbes-communities"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC7772471/","version":"version of record","retrieved_at":"2026-09-15T23:29:32Z","doi":"10.1093/ve/veaa091","publication_status":"peer_reviewed","year":2020,"artifact_sha256":"77842d8e4f6b419e331ab5a01fdf8f9eb8604f259425d79602be896aad3d0ad1","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC7772471/fullTextXML","artifact_retrieved_at":"2026-09-16T10:44:03.412605+00:00","legacy_paper":{"id":"lazypipe-2020","title":"Novel NGS pipeline for virus discovery from a wide spectrum of hosts and sample types","year":2020,"publication_status":"peer_reviewed","version":"version of record","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC7772471/","primary_domain":"microbes-communities","retrieved_utc":"2026-09-15T23:29:32Z","notes":"Primary full text verified using Europe PMC XML; venue: Virus Evolution; PMC ID: PMC7772471.","doi":"10.1093/ve/veaa091"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"lemur-magnet-2024","kind":"source","name":"Lightweight taxonomic profiling of long-read metagenomic datasets with Lemur and Magnet","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["microbes-communities"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11185576/","version":"PMC archival version PMC11185576.2","retrieved_at":"2026-09-15T23:29:32Z","doi":"10.1101/2024.06.01.596961","publication_status":"preprint","year":2024,"artifact_sha256":"4afb9195da447916eb6f733816e3640741c7ade08ea8920d205c3be7b3cce27a","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11185576/fullTextXML","artifact_retrieved_at":"2026-09-16T10:44:03.420493+00:00","legacy_paper":{"id":"lemur-magnet-2024","title":"Lightweight taxonomic profiling of long-read metagenomic datasets with Lemur and Magnet","year":2024,"publication_status":"preprint","version":"PMC archival version PMC11185576.2","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11185576/","primary_domain":"microbes-communities","retrieved_utc":"2026-09-15T23:29:32Z","notes":"Primary full text verified using Europe PMC XML; venue: bioRxiv; PMC ID: PMC11185576.","doi":"10.1101/2024.06.01.596961"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"ligand-affinity-meta-model-2024","kind":"source","name":"Improved Prediction of Ligand–Protein Binding Affinities by Meta-modeling","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["molecular-interactions"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11632770/","version":"PMC archival version PMC11632770.1","retrieved_at":"2026-09-15T23:37:05Z","doi":"10.1021/acs.jcim.4c01116","publication_status":"peer_reviewed","year":2024,"artifact_sha256":"0be25fe75bc0b2eb8065136555763bbae5964ea3a8fdb8c5de79ff96445f6a29","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11632770/fullTextXML","artifact_retrieved_at":"2026-09-16T10:41:06Z","legacy_paper":{"id":"ligand-affinity-meta-model-2024","title":"Improved Prediction of Ligand–Protein Binding Affinities by Meta-modeling","year":2024,"publication_status":"peer_reviewed","version":"PMC archival version PMC11632770.1","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11632770/","primary_domain":"molecular-interactions","retrieved_utc":"2026-09-15T23:37:05Z","notes":"Primary full text verified via Europe PMC fullTextXML; venue: Journal of Chemical Information and Modeling; PMC ID: PMC11632770. Mixed prediction units across Table 4 comparators; only meta-model PCC recorded.","doi":"10.1021/acs.jcim.4c01116"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"lipp-2026","kind":"source","name":"The LiPP Benchmark Set for Modeling Lipid–Protein Complexes: Comparison of Co-Folding and Docking Methods","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["molecular-interactions"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC13292216/","version":"PMC13292216.1","retrieved_at":"2026-09-15T23:25:00Z","doi":"10.1021/acs.jcim.6c01457","publication_status":"peer_reviewed","year":2026,"artifact_sha256":"6ff34f2f709a14858a3753abf9f8f6efa1e7e3c351f15c70cf64264193a9414e","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC13292216/fullTextXML","artifact_retrieved_at":"2026-09-16T10:41:16.548973+00:00","legacy_paper":{"id":"lipp-2026","title":"The LiPP Benchmark Set for Modeling Lipid–Protein Complexes: Comparison of Co-Folding and Docking Methods","year":2026,"publication_status":"peer_reviewed","version":"PMC13292216.1","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC13292216/","primary_domain":"molecular-interactions","retrieved_utc":"2026-09-15T23:25:00Z","notes":"Primary full text via Europe PMC XML; venue: Journal of Chemical Information and Modeling; PMC ID: PMC13292216.","doi":"10.1021/acs.jcim.6c01457"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"lit-001","kind":"result","name":"Caduceus-Ph · AUC · Human 5mC","description":"","status":"source_checked","facets":{"areas":["dna-genomes"],"tasks":["Human 5mC detection"]},"source_ids":["dna-foundation-models-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-001"}],"attributes":{"printed_value":"0.783","numeric_value":"0.783","metric":"AUC","metric_direction":"unknown","unit":"unitless","uncertainty":null,"source_locator":"Table 3, Human 5mC row, Caduceus-Ph column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:35.327Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness.","evidence":"Table 3, Human 5mC row, Caduceus-Ph column; cell: 0.783","artifact_sha256":"5d8ca9bcf88cc1b38ad667906a2e4699b1aefa6d31c6f49259784930353f3202","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12663285/fullTextXML"},"legacy_id":"lit-001","legacy_row":{"id":"lit-001","paper_id":"dna-foundation-models-2025","domain_id":"dna-genomes","task":"Human 5mC detection","model":"Caduceus-Ph","model_version":"","dataset":"Human 5mC","dataset_version":"","split":"","metric":"AUC","value":"0.783","unit":"unitless","uncertainty":"","protocol":"Binary epigenetic-modification classification as reported in the paper.","source_locator":"Table 3, Human 5mC row, Caduceus-Ph column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12663285/","evaluation_origin":"independent_paper","reviewed_utc":"2026-09-15T23:25:00Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-002","kind":"result","name":"NT-v2 · AUC · Human 5mC","description":"","status":"source_checked","facets":{"areas":["dna-genomes"],"tasks":["Human 5mC detection"]},"source_ids":["dna-foundation-models-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-002"}],"attributes":{"printed_value":"0.7377","numeric_value":"0.7377","metric":"AUC","metric_direction":"unknown","unit":"unitless","uncertainty":null,"source_locator":"Table 3, Human 5mC row, NT-v2 column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:35.327Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness.","evidence":"Table 3, Human 5mC row, NT-v2 column; cell: 0.7377","artifact_sha256":"5d8ca9bcf88cc1b38ad667906a2e4699b1aefa6d31c6f49259784930353f3202","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12663285/fullTextXML"},"legacy_id":"lit-002","legacy_row":{"id":"lit-002","paper_id":"dna-foundation-models-2025","domain_id":"dna-genomes","task":"Human 5mC detection","model":"NT-v2","model_version":"","dataset":"Human 5mC","dataset_version":"","split":"","metric":"AUC","value":"0.7377","unit":"unitless","uncertainty":"","protocol":"Binary epigenetic-modification classification as reported in the paper.","source_locator":"Table 3, Human 5mC row, NT-v2 column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12663285/","evaluation_origin":"independent_paper","reviewed_utc":"2026-09-15T23:25:00Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-003","kind":"result","name":"ENBED · Accuracy · Genomic Benchmarks Mouse Enhancers","description":"","status":"source_checked","facets":{"areas":["dna-genomes"],"tasks":["Enhancer classification"]},"source_ids":["enbed-2024"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-003"}],"attributes":{"printed_value":"90.3","numeric_value":"90.3","metric":"Accuracy","metric_direction":"unknown","unit":"%","uncertainty":null,"source_locator":"Table 2, Mouse Enhancers row, ENBED column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:35.378Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness.","evidence":"Table 2, Mouse Enhancers row, ENBED column; cell: 90.3","artifact_sha256":"e95d4be70d32e61af5a92eda8ea66f25a2cc629e3f83e7b5241b13cde8bdb83b","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11341122/fullTextXML"},"legacy_id":"lit-003","legacy_row":{"id":"lit-003","paper_id":"enbed-2024","domain_id":"dna-genomes","task":"Enhancer classification","model":"ENBED","model_version":"","dataset":"Genomic Benchmarks Mouse Enhancers","dataset_version":"","split":"","metric":"Accuracy","value":"90.3","unit":"%","uncertainty":"","protocol":"Reported Genomic Benchmarks classification accuracy.","source_locator":"Table 2, Mouse Enhancers row, ENBED column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11341122/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:25:00Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-004","kind":"result","name":"ENBED (GRCh38) · Accuracy · Genomic Benchmarks Mouse Enhancers","description":"","status":"source_checked","facets":{"areas":["dna-genomes"],"tasks":["Enhancer classification"]},"source_ids":["enbed-2024"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-004"}],"attributes":{"printed_value":"81.1","numeric_value":"81.1","metric":"Accuracy","metric_direction":"unknown","unit":"%","uncertainty":null,"source_locator":"Table 2, Mouse Enhancers row, ENBED (GRCh38) column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:35.378Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness.","evidence":"Table 2, Mouse Enhancers row, ENBED (GRCh38) column; cell: 81.1","artifact_sha256":"e95d4be70d32e61af5a92eda8ea66f25a2cc629e3f83e7b5241b13cde8bdb83b","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11341122/fullTextXML"},"legacy_id":"lit-004","legacy_row":{"id":"lit-004","paper_id":"enbed-2024","domain_id":"dna-genomes","task":"Enhancer classification","model":"ENBED (GRCh38)","model_version":"","dataset":"Genomic Benchmarks Mouse Enhancers","dataset_version":"","split":"","metric":"Accuracy","value":"81.1","unit":"%","uncertainty":"","protocol":"ENBED trained on GRCh38; reported Genomic Benchmarks classification accuracy.","source_locator":"Table 2, Mouse Enhancers row, ENBED (GRCh38) column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11341122/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:25:00Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-005","kind":"result","name":"DNABERT-2 · Accuracy · KEx","description":"","status":"source_checked","facets":{"areas":["dna-genomes"],"tasks":["G-quadruplex classification"]},"source_ids":["quadruplex-llm-benchmark-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-005"}],"attributes":{"printed_value":"97.0","numeric_value":"97.0","metric":"Accuracy","metric_direction":"unknown","unit":"%","uncertainty":"± 0.5","source_locator":"Table 5, DNABERT-2 (117 M) row, Accuracy column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:35.379Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness.","evidence":"Table 5, DNABERT-2 (117 M) row, Accuracy column; cell: 97.0 ± 0.5","artifact_sha256":"c3d7c6d068d3c11a9c8255a932197ece3804d94e8b2d4f858bea373a1b6eb32f","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11953744/fullTextXML"},"legacy_id":"lit-005","legacy_row":{"id":"lit-005","paper_id":"quadruplex-llm-benchmark-2025","domain_id":"dna-genomes","task":"G-quadruplex classification","model":"DNABERT-2","model_version":"117M","dataset":"KEx","dataset_version":"","split":"","metric":"Accuracy","value":"97.0","unit":"%","uncertainty":"± 0.5","protocol":"Pretrained model evaluated on KEx as reported in Table 5.","source_locator":"Table 5, DNABERT-2 (117 M) row, Accuracy column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11953744/","evaluation_origin":"independent_paper","reviewed_utc":"2026-09-15T23:25:00Z"},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"lit-006","kind":"result","name":"Caduceus · Accuracy · KEx","description":"","status":"source_checked","facets":{"areas":["dna-genomes"],"tasks":["G-quadruplex classification"]},"source_ids":["quadruplex-llm-benchmark-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-006"}],"attributes":{"printed_value":"95.0","numeric_value":"95.0","metric":"Accuracy","metric_direction":"unknown","unit":"%","uncertainty":"± 0.5","source_locator":"Table 5, Caduceus (8 M) row, Accuracy column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:35.379Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness.","evidence":"Table 5, Caduceus (8 M) row, Accuracy column; cell: 95.0 ± 0.5","artifact_sha256":"c3d7c6d068d3c11a9c8255a932197ece3804d94e8b2d4f858bea373a1b6eb32f","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11953744/fullTextXML"},"legacy_id":"lit-006","legacy_row":{"id":"lit-006","paper_id":"quadruplex-llm-benchmark-2025","domain_id":"dna-genomes","task":"G-quadruplex classification","model":"Caduceus","model_version":"8M","dataset":"KEx","dataset_version":"","split":"","metric":"Accuracy","value":"95.0","unit":"%","uncertainty":"± 0.5","protocol":"Pretrained model evaluated on KEx as reported in Table 5.","source_locator":"Table 5, Caduceus (8 M) row, Accuracy column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11953744/","evaluation_origin":"independent_paper","reviewed_utc":"2026-09-15T23:25:00Z"},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"lit-007","kind":"result","name":"HyenaDNA · AUROC · DNALongBench ETGP","description":"","status":"source_checked","facets":{"areas":["dna-genomes"],"tasks":["Enhancer-target gene prediction"]},"source_ids":["dnalongbench-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-007"}],"attributes":{"printed_value":"0.828","numeric_value":"0.828","metric":"AUROC","metric_direction":"unknown","unit":"unitless","uncertainty":null,"source_locator":"Table 3, HyenaDNA row, ETGP column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.492545+00:00","notes":"ETGP is the first numeric column, separate from the six CMP cell-type columns and average; caption defines ETGP AUROC. This verifies the central score at its source location, not every metadata field or an experimental reproduction.","evidence":"{\"table_xml_id\": \"T3\", \"row_cells\": [\"HyenaDNA\", \"0.828\", \"0.139\", \"0.122\", \"0.099\", \"0.097\", \"0.118\", \"0.115\"], \"selected_cell_zero_based\": 1, \"selected_cell_xml\": \"<td align=\\\"center\\\" valign=\\\"middle\\\" rowspan=\\\"1\\\" colspan=\\\"1\\\">0.828</td>\", \"caption\": \"AUROC for enhancer-target gene prediction (ETGP) task and SCC scores for the contact map prediction (CMP) task. K562, HFF, H1hESC, GM12878, IMR90, and HCT116 represent different human cell types. The highest scores are highlighted in bold. “Avg” means the average score across different cell types. Notably, the Expert Model achieves the best performance on both ETGP and CMP tasks.\"}","artifact_sha256":"fa440a17cecf16a5d872d50a30910f7591b5f6f78e10a944c6bda5ea8d7e32dd","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11741265/fullTextXML"},"legacy_id":"lit-007","legacy_row":{"id":"lit-007","paper_id":"dnalongbench-2025","domain_id":"dna-genomes","task":"Enhancer-target gene prediction","model":"HyenaDNA","model_version":"","dataset":"DNALongBench ETGP","dataset_version":"","split":"","metric":"AUROC","value":"0.828","unit":"unitless","uncertainty":"","protocol":"Long-range ETGP benchmark; source table reports AUROC.","source_locator":"Table 3, HyenaDNA row, ETGP column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11741265/","evaluation_origin":"independent_paper","reviewed_utc":"2026-09-15T23:25:00Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-008","kind":"result","name":"Caduceus-Ph · AUROC · DNALongBench ETGP","description":"","status":"source_checked","facets":{"areas":["dna-genomes"],"tasks":["Enhancer-target gene prediction"]},"source_ids":["dnalongbench-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-008"}],"attributes":{"printed_value":"0.826","numeric_value":"0.826","metric":"AUROC","metric_direction":"unknown","unit":"unitless","uncertainty":null,"source_locator":"Table 3, Caduceus-Ph row, ETGP column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.493765+00:00","notes":"ETGP is the first numeric column, separate from the six CMP cell-type columns and average; caption defines ETGP AUROC. This verifies the central score at its source location, not every metadata field or an experimental reproduction.","evidence":"{\"table_xml_id\": \"T3\", \"row_cells\": [\"Caduceus-Ph\", \"0.826\", \"0.153\", \"0.130\", \"0.101\", \"0.138\", \"0.145\", \"0.133\"], \"selected_cell_zero_based\": 1, \"selected_cell_xml\": \"<td align=\\\"center\\\" valign=\\\"middle\\\" rowspan=\\\"1\\\" colspan=\\\"1\\\">0.826</td>\", \"caption\": \"AUROC for enhancer-target gene prediction (ETGP) task and SCC scores for the contact map prediction (CMP) task. K562, HFF, H1hESC, GM12878, IMR90, and HCT116 represent different human cell types. The highest scores are highlighted in bold. “Avg” means the average score across different cell types. Notably, the Expert Model achieves the best performance on both ETGP and CMP tasks.\"}","artifact_sha256":"fa440a17cecf16a5d872d50a30910f7591b5f6f78e10a944c6bda5ea8d7e32dd","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11741265/fullTextXML"},"legacy_id":"lit-008","legacy_row":{"id":"lit-008","paper_id":"dnalongbench-2025","domain_id":"dna-genomes","task":"Enhancer-target gene prediction","model":"Caduceus-Ph","model_version":"","dataset":"DNALongBench ETGP","dataset_version":"","split":"","metric":"AUROC","value":"0.826","unit":"unitless","uncertainty":"","protocol":"Long-range ETGP benchmark; source table reports AUROC.","source_locator":"Table 3, Caduceus-Ph row, ETGP column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11741265/","evaluation_origin":"independent_paper","reviewed_utc":"2026-09-15T23:25:00Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-009","kind":"result","name":"RiNALMo · Pearson R · mRNABench MRL-MPRA","description":"","status":"source_checked","facets":{"areas":["rna-transcriptomes"],"tasks":["Mean ribosome load from MPRA"]},"source_ids":["mrnabench-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-009"}],"attributes":{"printed_value":"0.74","numeric_value":"0.74","metric":"Pearson R","metric_direction":"unknown","unit":"unitless","uncertainty":null,"source_locator":"Table 2, RiNALMo row, MRL MPRA column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.497221+00:00","notes":"MRL MPRA is the second task under Local and uses R. Caption specifies mean over ten random seeds and selected best model per family, not a fully identified checkpoint. This verifies the central score at its source location, not every metadata field or an experimental reproduction.","evidence":"{\"table_xml_id\": \"T9\", \"row_cells\": [\"RiNALMo\", \"31.6\", \"0.74\", \"39.2\", \"79.5\", \"69.0\", \"0.53\", \"0.42\", \"0.47\", \"32.7\", \"34.7\"], \"selected_cell_zero_based\": 2, \"selected_cell_xml\": \"<td align=\\\"right\\\" valign=\\\"top\\\" rowspan=\\\"1\\\" colspan=\\\"1\\\">\\n<bold>\\n<underline>0.74</underline>\\n</bold>\\n</td>\", \"caption\": \"Linear probe results. Mean of metric over ten random seeds reported. Best model per model family reported, see Appendix D for selected models. Best model for each dataset is and best foundation model is underlined. Models not significantly worse under Wilcoxon signed-rank test at p=0.05 are bolded.\"}","artifact_sha256":"79f6264ee883535203c63a313547e7c57baa85585f76b42f8d899eb17fb7e600","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12265608/fullTextXML"},"legacy_id":"lit-009","legacy_row":{"id":"lit-009","paper_id":"mrnabench-2025","domain_id":"rna-transcriptomes","task":"Mean ribosome load from MPRA","model":"RiNALMo","model_version":"","dataset":"mRNABench MRL-MPRA","dataset_version":"","split":"","metric":"Pearson R","value":"0.74","unit":"unitless","uncertainty":"","protocol":"Linear probe; mean across ten random seeds.","source_locator":"Table 2, RiNALMo row, MRL MPRA column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12265608/","evaluation_origin":"independent_paper","reviewed_utc":"2026-09-15T23:25:00Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-010","kind":"result","name":"RNA-FM · Pearson R · mRNABench MRL-MPRA","description":"","status":"source_checked","facets":{"areas":["rna-transcriptomes"],"tasks":["Mean ribosome load from MPRA"]},"source_ids":["mrnabench-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-010"}],"attributes":{"printed_value":"0.49","numeric_value":"0.49","metric":"Pearson R","metric_direction":"unknown","unit":"unitless","uncertainty":null,"source_locator":"Table 2, RNA-FM row, MRL MPRA column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.500211+00:00","notes":"MRL MPRA is the second task under Local and uses R. Caption specifies mean over ten random seeds and selected best model per family, not a fully identified checkpoint. This verifies the central score at its source location, not every metadata field or an experimental reproduction.","evidence":"{\"table_xml_id\": \"T9\", \"row_cells\": [\"RNA-FM\", \"25.7\", \"0.49\", \"35.0\", \"74.3\", \"67.0\", \"0.47\", \"0.29\", \"0.49\", \"32.2\", \"32.2\"], \"selected_cell_zero_based\": 2, \"selected_cell_xml\": \"<td align=\\\"right\\\" valign=\\\"top\\\" rowspan=\\\"1\\\" colspan=\\\"1\\\">0.49</td>\", \"caption\": \"Linear probe results. Mean of metric over ten random seeds reported. Best model per model family reported, see Appendix D for selected models. Best model for each dataset is and best foundation model is underlined. Models not significantly worse under Wilcoxon signed-rank test at p=0.05 are bolded.\"}","artifact_sha256":"79f6264ee883535203c63a313547e7c57baa85585f76b42f8d899eb17fb7e600","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12265608/fullTextXML"},"legacy_id":"lit-010","legacy_row":{"id":"lit-010","paper_id":"mrnabench-2025","domain_id":"rna-transcriptomes","task":"Mean ribosome load from MPRA","model":"RNA-FM","model_version":"","dataset":"mRNABench MRL-MPRA","dataset_version":"","split":"","metric":"Pearson R","value":"0.49","unit":"unitless","uncertainty":"","protocol":"Linear probe; mean across ten random seeds.","source_locator":"Table 2, RNA-FM row, MRL MPRA column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12265608/","evaluation_origin":"independent_paper","reviewed_utc":"2026-09-15T23:25:00Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-011","kind":"result","name":"BPfold · F1 · PDB RNA set","description":"","status":"source_checked","facets":{"areas":["rna-transcriptomes"],"tasks":["RNA secondary structure"]},"source_ids":["bpfold-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-011"}],"attributes":{"printed_value":"0.814","numeric_value":"0.814","metric":"F1","metric_direction":"unknown","unit":"unitless","uncertainty":null,"source_locator":"Table 2, BPfold row, PDB F1 column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.502000+00:00","notes":"PDB is the second four-metric block; its F1 is numeric column six, not Rfam F1. This verifies the central score at its source location, not every metadata field or an experimental reproduction.","evidence":"{\"table_xml_id\": \"Tab2\", \"row_cells\": [\"BPfold\", \"0.694\", \"0.689\", \"0.660\", \"0.741\", \"0.817\", \"0.814\", \"0.840\", \"0.801\"], \"selected_cell_zero_based\": 6, \"selected_cell_xml\": \"<td colspan=\\\"1\\\" rowspan=\\\"1\\\">0.814</td>\", \"caption\": \"Family-wise evaluation of three DL methods (BPfold, SPOT-RNA, and MXfold2), three shallow learning methods (ContextFold, CONTRAfold, and EternaFold) and non-learning methods (LinearFold, RNAfold, SimFold, and RNAstructure) on Rfam12.3–14.10 (n = 10,791 RNAs) and PDB (n = 116 RNAs) datasets\"}","artifact_sha256":"976218bd172998a1a6e7ed1609ecb8cb2ee380fb48a8dc7b25bc05ea8b0a49af","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12216785/fullTextXML"},"legacy_id":"lit-011","legacy_row":{"id":"lit-011","paper_id":"bpfold-2025","domain_id":"rna-transcriptomes","task":"RNA secondary structure","model":"BPfold","model_version":"","dataset":"PDB RNA set","dataset_version":"116 RNAs","split":"","metric":"F1","value":"0.814","unit":"unitless","uncertainty":"","protocol":"Family-wise evaluation of canonical base-pair predictions.","source_locator":"Table 2, BPfold row, PDB F1 column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12216785/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:25:00Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-012","kind":"result","name":"RNAfold · F1 · PDB RNA set","description":"","status":"source_checked","facets":{"areas":["rna-transcriptomes"],"tasks":["RNA secondary structure"]},"source_ids":["bpfold-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-012"}],"attributes":{"printed_value":"0.747","numeric_value":"0.747","metric":"F1","metric_direction":"unknown","unit":"unitless","uncertainty":null,"source_locator":"Table 2, RNAfold row, PDB F1 column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.504220+00:00","notes":"PDB is the second four-metric block; its F1 is numeric column six, not Rfam F1. This verifies the central score at its source location, not every metadata field or an experimental reproduction.","evidence":"{\"table_xml_id\": \"Tab2\", \"row_cells\": [\"RNAfold\", \"0.656\", \"0.649\", \"0.599\", \"0.729\", \"0.749\", \"0.747\", \"0.776\", \"0.728\"], \"selected_cell_zero_based\": 6, \"selected_cell_xml\": \"<td colspan=\\\"1\\\" rowspan=\\\"1\\\">0.747</td>\", \"caption\": \"Family-wise evaluation of three DL methods (BPfold, SPOT-RNA, and MXfold2), three shallow learning methods (ContextFold, CONTRAfold, and EternaFold) and non-learning methods (LinearFold, RNAfold, SimFold, and RNAstructure) on Rfam12.3–14.10 (n = 10,791 RNAs) and PDB (n = 116 RNAs) datasets\"}","artifact_sha256":"976218bd172998a1a6e7ed1609ecb8cb2ee380fb48a8dc7b25bc05ea8b0a49af","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12216785/fullTextXML"},"legacy_id":"lit-012","legacy_row":{"id":"lit-012","paper_id":"bpfold-2025","domain_id":"rna-transcriptomes","task":"RNA secondary structure","model":"RNAfold","model_version":"","dataset":"PDB RNA set","dataset_version":"116 RNAs","split":"","metric":"F1","value":"0.747","unit":"unitless","uncertainty":"","protocol":"Family-wise evaluation of canonical base-pair predictions.","source_locator":"Table 2, RNAfold row, PDB F1 column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12216785/","evaluation_origin":"independent_paper","reviewed_utc":"2026-09-15T23:25:00Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-013","kind":"result","name":"TU-Fold (aug) · F1 · RNA8F","description":"","status":"source_checked","facets":{"areas":["rna-transcriptomes"],"tasks":["RNA secondary structure"]},"source_ids":["tu-fold-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-013"}],"attributes":{"printed_value":"0.947","numeric_value":"0.947","metric":"F1","metric_direction":"unknown","unit":"unitless","uncertainty":"± 0.002 standard deviation","source_locator":"Table 2, TU-Fold (aug) row, Overall F1 column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.505799+00:00","notes":"Overall is the first two-metric block. F1 is first numeric column; source cell includes uncertainty after the preserved central value. This verifies the central score at its source location, not every metadata field or an experimental reproduction.","evidence":"{\"table_xml_id\": \"tbl0020\", \"row_cells\": [\"TU-Fold (aug)\", \"0.947±0.002\", \"0.947±0.002\", \"0.973±0.003\", \"0.973±0.003\", \"0.797±0.016\", \"0.802±0.015\"], \"selected_cell_zero_based\": 1, \"selected_cell_xml\": \"<td colspan=\\\"1\\\" rowspan=\\\"1\\\"><bold>0.947</bold>±0.002</td>\", \"caption\": \"Main results: the F1 score and the INF score of each method. The largest score in each column is highlighted in bold. The score with an underline is the second largest one in each column. The numbers on the left and right of the ‘±’ suggest the mean and the standard deviation of this score when training and evaluating the model using 3 folds of the dataset.\"}","artifact_sha256":"5aa376d6466daee83fc307baa39fd48da0f185ff30a178624025032d4cbe597d","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12008525/fullTextXML"},"legacy_id":"lit-013","legacy_row":{"id":"lit-013","paper_id":"tu-fold-2025","domain_id":"rna-transcriptomes","task":"RNA secondary structure","model":"TU-Fold (aug)","model_version":"","dataset":"RNA8F","dataset_version":"","split":"","metric":"F1","value":"0.947","unit":"unitless","uncertainty":"± 0.002 standard deviation","protocol":"Three-fold training and evaluation; source reports mean and standard deviation.","source_locator":"Table 2, TU-Fold (aug) row, Overall F1 column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12008525/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:25:00Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"lit-014","kind":"result","name":"UFold · F1 · RNA8F","description":"","status":"source_checked","facets":{"areas":["rna-transcriptomes"],"tasks":["RNA secondary structure"]},"source_ids":["tu-fold-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-014"}],"attributes":{"printed_value":"0.938","numeric_value":"0.938","metric":"F1","metric_direction":"unknown","unit":"unitless","uncertainty":"± 0.004 standard deviation","source_locator":"Table 2, UFold row, Overall F1 column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.507031+00:00","notes":"Overall is the first two-metric block. F1 is first numeric column; source cell includes uncertainty after the preserved central value. This verifies the central score at its source location, not every metadata field or an experimental reproduction.","evidence":"{\"table_xml_id\": \"tbl0020\", \"row_cells\": [\"UFold\", \"0.938±0.004\", \"0.938±0.004\", \"0.976±0.003\", \"0.976±0.003\", \"0.719±0.030\", \"0.720±0.030\"], \"selected_cell_zero_based\": 1, \"selected_cell_xml\": \"<td colspan=\\\"1\\\" rowspan=\\\"1\\\">0.938±0.004</td>\", \"caption\": \"Main results: the F1 score and the INF score of each method. The largest score in each column is highlighted in bold. The score with an underline is the second largest one in each column. The numbers on the left and right of the ‘±’ suggest the mean and the standard deviation of this score when training and evaluating the model using 3 folds of the dataset.\"}","artifact_sha256":"5aa376d6466daee83fc307baa39fd48da0f185ff30a178624025032d4cbe597d","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12008525/fullTextXML"},"legacy_id":"lit-014","legacy_row":{"id":"lit-014","paper_id":"tu-fold-2025","domain_id":"rna-transcriptomes","task":"RNA secondary structure","model":"UFold","model_version":"","dataset":"RNA8F","dataset_version":"","split":"","metric":"F1","value":"0.938","unit":"unitless","uncertainty":"± 0.004 standard deviation","protocol":"Three-fold training and evaluation; source reports mean and standard deviation.","source_locator":"Table 2, UFold row, Overall F1 column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12008525/","evaluation_origin":"independent_paper","reviewed_utc":"2026-09-15T23:25:00Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"lit-015","kind":"result","name":"DEBFold · Median F1 · DEBFold TestSetβ","description":"","status":"source_checked","facets":{"areas":["rna-transcriptomes"],"tasks":["RNA secondary structure"]},"source_ids":["debfold-2024"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-015"}],"attributes":{"printed_value":"55.7","numeric_value":"55.7","metric":"Median F1","metric_direction":"unknown","unit":"%","uncertainty":null,"source_locator":"Table 1, DEBFold row, TestSetβ F1 (%) column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.509290+00:00","notes":"TestSet beta is the second four-column block. F1 (%) is its first column; caption reports test-set median F1. This verifies the central score at its source location, not every metadata field or an experimental reproduction.","evidence":"{\"table_xml_id\": \"tbl1\", \"row_cells\": [\"DEBFold\", \"64.9\", \"62.1\", \"67.9\", \"4\", \"55.7\", \"56.4\", \"56.8\", \"1\", \"77.9\", \"83.3\", \"73.2\", \"1\", \"64.9\", \"1\"], \"selected_cell_zero_based\": 5, \"selected_cell_xml\": \"<td style=\\\"border:none;\\\" align=\\\"left\\\" colspan=\\\"1\\\" rowspan=\\\"1\\\">55.7</td>\", \"caption\": \"Test Set Median F1 Score Performance Comparison between DEBFold and Other Available Thermodynamics-Based RNA Secondary Structure Prediction Tools on the Three Prepared Test Setsa\"}","artifact_sha256":"e8f960eafb7f00edfdd81d4fb75c6de838e9b872b7e18875fc7a5bff2a2f72b3","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11094721/fullTextXML"},"legacy_id":"lit-015","legacy_row":{"id":"lit-015","paper_id":"debfold-2024","domain_id":"rna-transcriptomes","task":"RNA secondary structure","model":"DEBFold","model_version":"","dataset":"DEBFold TestSetβ","dataset_version":"","split":"","metric":"Median F1","value":"55.7","unit":"%","uncertainty":"","protocol":"Median F1 on the prepared TestSetβ.","source_locator":"Table 1, DEBFold row, TestSetβ F1 (%) column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11094721/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:25:00Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-016","kind":"result","name":"RNAfold · Median F1 · DEBFold TestSetβ","description":"","status":"source_checked","facets":{"areas":["rna-transcriptomes"],"tasks":["RNA secondary structure"]},"source_ids":["debfold-2024"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-016"}],"attributes":{"printed_value":"52.3","numeric_value":"52.3","metric":"Median F1","metric_direction":"unknown","unit":"%","uncertainty":null,"source_locator":"Table 1, RNAfold row, TestSetβ F1 (%) column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.511166+00:00","notes":"TestSet beta is the second four-column block. F1 (%) is its first column; caption reports test-set median F1. This verifies the central score at its source location, not every metadata field or an experimental reproduction.","evidence":"{\"table_xml_id\": \"tbl1\", \"row_cells\": [\"RNAfold\", \"63.5\", \"53.9\", \"82.9\", \"8\", \"52.3\", \"39.9\", \"81.2\", \"8\", \"77.9\", \"83.3\", \"73.2\", \"1\", \"63.5\", \"8\"], \"selected_cell_zero_based\": 5, \"selected_cell_xml\": \"<td style=\\\"border:none;\\\" align=\\\"left\\\" colspan=\\\"1\\\" rowspan=\\\"1\\\">52.3</td>\", \"caption\": \"Test Set Median F1 Score Performance Comparison between DEBFold and Other Available Thermodynamics-Based RNA Secondary Structure Prediction Tools on the Three Prepared Test Setsa\"}","artifact_sha256":"e8f960eafb7f00edfdd81d4fb75c6de838e9b872b7e18875fc7a5bff2a2f72b3","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11094721/fullTextXML"},"legacy_id":"lit-016","legacy_row":{"id":"lit-016","paper_id":"debfold-2024","domain_id":"rna-transcriptomes","task":"RNA secondary structure","model":"RNAfold","model_version":"","dataset":"DEBFold TestSetβ","dataset_version":"","split":"","metric":"Median F1","value":"52.3","unit":"%","uncertainty":"","protocol":"Median F1 on the prepared TestSetβ.","source_locator":"Table 1, RNAfold row, TestSetβ F1 (%) column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11094721/","evaluation_origin":"independent_paper","reviewed_utc":"2026-09-15T23:25:00Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-017","kind":"result","name":"ESM-2 · Mean Spearman rho · ProteinGym substitution DMS: stability assays","description":"","status":"source_checked","facets":{"areas":["proteins-complexes"],"tasks":["Zero-shot substitution mutation effects: stability"]},"source_ids":["proteingym-2023"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-017"}],"attributes":{"printed_value":"0.488","numeric_value":"0.488","metric":"Mean Spearman rho","metric_direction":"unknown","unit":"unitless","uncertainty":null,"source_locator":"Table A7, ESM-2 (15B) row, Stability column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.517323+00:00","notes":"Table A7 is zero-shot substitution DMS grouped by function. Stability is the fifth numeric column; model-type row spans do not change its placement. This verifies the central score at its source location, not every metadata field or an experimental reproduction.","evidence":"{\"table_xml_id\": \"T11\", \"row_cells\": [\"ESM-2 (15B)\", \"0.405\", \"0.318\", \"0.425\", \"0.388\", \"0.488\", \"0.405\"], \"selected_cell_zero_based\": 5, \"selected_cell_xml\": \"<td align=\\\"center\\\" valign=\\\"top\\\" rowspan=\\\"1\\\" colspan=\\\"1\\\">0.488</td>\", \"caption\": \"ProteinGym - Zero-shot substitution DMS benchmark by function typeAverage Spearman’s rank correlation between model scores and experimental measurements on the ProteinGym substitution benchmark, separated into five functional categories (Activity, Binding, Organismal Fitness, Stability and Expression). ‘All’ is the average of all the categories.\"}","artifact_sha256":"4519641f13271bdd09b166e7d93232f22542489bc52a25b5a1628c3df8badce1","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC10723403/fullTextXML"},"legacy_id":"lit-017","legacy_row":{"id":"lit-017","paper_id":"proteingym-2023","domain_id":"proteins-complexes","task":"Zero-shot substitution mutation effects: stability","model":"ESM-2","model_version":"15B","dataset":"ProteinGym substitution DMS: stability assays","dataset_version":"","split":"","metric":"Mean Spearman rho","value":"0.488","unit":"unitless","uncertainty":"","protocol":"Zero-shot mutation scores; average Spearman across stability-category assays.","source_locator":"Table A7, ESM-2 (15B) row, Stability column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC10723403/","evaluation_origin":"independent_paper","reviewed_utc":"2026-09-15T23:25:00Z"},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-018","kind":"result","name":"ProteinMPNN · Mean Spearman rho · ProteinGym substitution DMS: stability assays","description":"","status":"source_checked","facets":{"areas":["proteins-complexes"],"tasks":["Zero-shot substitution mutation effects: stability"]},"source_ids":["proteingym-2023"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-018"}],"attributes":{"printed_value":"0.566","numeric_value":"0.566","metric":"Mean Spearman rho","metric_direction":"unknown","unit":"unitless","uncertainty":null,"source_locator":"Table A7, ProteinMPNN row, Stability column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.523422+00:00","notes":"Table A7 is zero-shot substitution DMS grouped by function. Stability is the fifth numeric column; model-type row spans do not change its placement. This verifies the central score at its source location, not every metadata field or an experimental reproduction.","evidence":"{\"table_xml_id\": \"T11\", \"row_cells\": [\"ProteinMPNN\", \"0.197\", \"0.165\", \"0.198\", \"0.165\", \"0.566\", \"0.258\"], \"selected_cell_zero_based\": 5, \"selected_cell_xml\": \"<td align=\\\"center\\\" valign=\\\"top\\\" rowspan=\\\"1\\\" colspan=\\\"1\\\">0.566</td>\", \"caption\": \"ProteinGym - Zero-shot substitution DMS benchmark by function typeAverage Spearman’s rank correlation between model scores and experimental measurements on the ProteinGym substitution benchmark, separated into five functional categories (Activity, Binding, Organismal Fitness, Stability and Expression). ‘All’ is the average of all the categories.\"}","artifact_sha256":"4519641f13271bdd09b166e7d93232f22542489bc52a25b5a1628c3df8badce1","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC10723403/fullTextXML"},"legacy_id":"lit-018","legacy_row":{"id":"lit-018","paper_id":"proteingym-2023","domain_id":"proteins-complexes","task":"Zero-shot substitution mutation effects: stability","model":"ProteinMPNN","model_version":"","dataset":"ProteinGym substitution DMS: stability assays","dataset_version":"","split":"","metric":"Mean Spearman rho","value":"0.566","unit":"unitless","uncertainty":"","protocol":"Zero-shot mutation scores; average Spearman across stability-category assays.","source_locator":"Table A7, ProteinMPNN row, Stability column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC10723403/","evaluation_origin":"independent_paper","reviewed_utc":"2026-09-15T23:25:00Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-019","kind":"result","name":"FUJISAN · AUROC · FUJISAN test sub-dataset","description":"","status":"source_checked","facets":{"areas":["proteins-complexes"],"tasks":["Enzyme functional identity prediction"]},"source_ids":["fujisan-2024"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-019"}],"attributes":{"printed_value":"0.9427","numeric_value":"0.9427","metric":"AUROC","metric_direction":"unknown","unit":"unitless","uncertainty":null,"source_locator":"Table 1, FUJISAN row, AUROC column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:35.728Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness.","evidence":"Table 1, FUJISAN row, AUROC column; cell: 0.9427","artifact_sha256":"db33e0542005ffae00cd644dfe697185b94c8823d5aee2768620a6db0c48e56f","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11609699/fullTextXML"},"legacy_id":"lit-019","legacy_row":{"id":"lit-019","paper_id":"fujisan-2024","domain_id":"proteins-complexes","task":"Enzyme functional identity prediction","model":"FUJISAN","model_version":"","dataset":"FUJISAN test sub-dataset","dataset_version":"","split":"","metric":"AUROC","value":"0.9427","unit":"unitless","uncertainty":"","protocol":"Sequence and structural feature integration; paper-reported test sub-dataset.","source_locator":"Table 1, FUJISAN row, AUROC column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11609699/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:25:00Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-020","kind":"result","name":"ESM2 · AUROC · FUJISAN test sub-dataset","description":"","status":"source_checked","facets":{"areas":["proteins-complexes"],"tasks":["Enzyme functional identity prediction"]},"source_ids":["fujisan-2024"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-020"}],"attributes":{"printed_value":"0.7991","numeric_value":"0.7991","metric":"AUROC","metric_direction":"unknown","unit":"unitless","uncertainty":null,"source_locator":"Table 1, ESM2 row, AUROC column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:35.728Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness.","evidence":"Table 1, ESM2 row, AUROC column; cell: 0.7991","artifact_sha256":"db33e0542005ffae00cd644dfe697185b94c8823d5aee2768620a6db0c48e56f","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11609699/fullTextXML"},"legacy_id":"lit-020","legacy_row":{"id":"lit-020","paper_id":"fujisan-2024","domain_id":"proteins-complexes","task":"Enzyme functional identity prediction","model":"ESM2","model_version":"","dataset":"FUJISAN test sub-dataset","dataset_version":"","split":"","metric":"AUROC","value":"0.7991","unit":"unitless","uncertainty":"","protocol":"Comparator evaluated on the paper-reported test sub-dataset.","source_locator":"Table 1, ESM2 row, AUROC column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11609699/","evaluation_origin":"independent_paper","reviewed_utc":"2026-09-15T23:25:00Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-021","kind":"result","name":"ESM-2 · R² · PRIME mutated RBD","description":"","status":"source_checked","facets":{"areas":["proteins-complexes"],"tasks":["Mutated RBD binding prediction"]},"source_ids":["prime-2026"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-021"}],"attributes":{"printed_value":"0.0248","numeric_value":"0.0248","metric":"R²","metric_direction":"unknown","unit":"unitless","uncertainty":"± 0.01","source_locator":"Table 1, ESM-2 8M / Mean / not fine-tuned row, Position-Stratified Split Binding R² column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.525183+00:00","notes":"Resolved model row spans and Mean/CLS subrows in JATS: selected Mean, not fine-tuned (cross), Position-Stratified Split > Binding > R-squared. Central value agrees; uncertainty is retained in evidence. This verifies the central score at its source location, not every metadata field or an experimental reproduction.","evidence":"{\"table_xml_id\": \"Tab1\", \"row_cells\": [\"Mean\", \"×\", \"0.6794 ± 0.02\", \"1.0777 ± 0.04\", \"0.6576 ± 0.04\", \"0.5807 ± 0.03\", \"0.0248 ± 0.01\", \"1.7519 ± 0.01\", \"0.0967 ± 0.02\", \"1.0221 ± 0.01\"], \"selected_cell_zero_based\": 6, \"selected_cell_xml\": \"<td align=\\\"left\\\" colspan=\\\"1\\\" rowspan=\\\"1\\\">0.0248 ± 0.01</td>\", \"caption\": \"Benchmarking PRIME across different model scales and validation regimes for mutated RBD binding and expression\"}","artifact_sha256":"f6aac4c25dd93026f87ce9a2e327c95faf4c3014d7f9ae04bb11f208ce047971","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC13425921/fullTextXML"},"legacy_id":"lit-021","legacy_row":{"id":"lit-021","paper_id":"prime-2026","domain_id":"proteins-complexes","task":"Mutated RBD binding prediction","model":"ESM-2","model_version":"8M","dataset":"PRIME mutated RBD","dataset_version":"","split":"position-stratified","metric":"R²","value":"0.0248","unit":"unitless","uncertainty":"± 0.01","protocol":"Frozen mean-pooled representation with downstream regression; position-stratified split.","source_locator":"Table 1, ESM-2 8M / Mean / not fine-tuned row, Position-Stratified Split Binding R² column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC13425921/","evaluation_origin":"independent_paper","reviewed_utc":"2026-09-15T23:25:00Z"},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract"}}}
{"id":"lit-022","kind":"result","name":"ESM-C · R² · PRIME mutated RBD","description":"","status":"source_checked","facets":{"areas":["proteins-complexes"],"tasks":["Mutated RBD binding prediction"]},"source_ids":["prime-2026"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-022"}],"attributes":{"printed_value":"-0.0162","numeric_value":"-0.0162","metric":"R²","metric_direction":"unknown","unit":"unitless","uncertainty":"± 0.01","source_locator":"Table 1, ESM-C 300M / Mean / not fine-tuned row, Position-Stratified Split Binding R² column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.526387+00:00","notes":"Resolved model row spans and Mean/CLS subrows in JATS: selected Mean, not fine-tuned (cross), Position-Stratified Split > Binding > R-squared. Central value agrees; uncertainty is retained in evidence. This verifies the central score at its source location, not every metadata field or an experimental reproduction.","evidence":"{\"table_xml_id\": \"Tab1\", \"row_cells\": [\"Mean\", \"×\", \"0.7155 ± 0.07\", \"1.0075 ± 0.13\", \"0.6546 ± 0.04\", \"0.5830 ± 0.03\", \"-0.0162 ± 0.01\", \"1.7874 ± 0.02\", \"0.0647 ± 0.02\", \"1.0410 ± 0.01\"], \"selected_cell_zero_based\": 6, \"selected_cell_xml\": \"<td align=\\\"left\\\" colspan=\\\"1\\\" rowspan=\\\"1\\\">-0.0162 ± 0.01</td>\", \"caption\": \"Benchmarking PRIME across different model scales and validation regimes for mutated RBD binding and expression\"}","artifact_sha256":"f6aac4c25dd93026f87ce9a2e327c95faf4c3014d7f9ae04bb11f208ce047971","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC13425921/fullTextXML"},"legacy_id":"lit-022","legacy_row":{"id":"lit-022","paper_id":"prime-2026","domain_id":"proteins-complexes","task":"Mutated RBD binding prediction","model":"ESM-C","model_version":"300M","dataset":"PRIME mutated RBD","dataset_version":"","split":"position-stratified","metric":"R²","value":"-0.0162","unit":"unitless","uncertainty":"± 0.01","protocol":"Frozen mean-pooled representation with downstream regression; position-stratified split.","source_locator":"Table 1, ESM-C 300M / Mean / not fine-tuned row, Position-Stratified Split Binding R² column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC13425921/","evaluation_origin":"independent_paper","reviewed_utc":"2026-09-15T23:25:00Z"},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract"}}}
{"id":"lit-023","kind":"result","name":"PST · Mean |Spearman rho| · ProteinShake VEP datasets","description":"","status":"source_checked","facets":{"areas":["proteins-complexes"],"tasks":["Zero-shot variant effect prediction"]},"source_ids":["pst-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-023"}],"attributes":{"printed_value":"0.501","numeric_value":"0.501","metric":"Mean |Spearman rho|","metric_direction":"unknown","unit":"unitless","uncertainty":null,"source_locator":"Table 2, PST row, Zero-shot VEP Mean |ρ| column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.527633+00:00","notes":"Zero-shot VEP is the last metric group; selected Mean absolute rho, not GO/EC/binding-site metrics. This verifies the central score at its source location, not every metadata field or an experimental reproduction.","evidence":"{\"table_xml_id\": \"btaf582-T2\", \"row_cells\": [\"PST\", \"0.650\", \"0.883\", \"0.704\", \"0.436\", \"0.797\", \"0.501\"], \"selected_cell_zero_based\": 6, \"selected_cell_xml\": \"<td rowspan=\\\"1\\\" colspan=\\\"1\\\">0.501</td>\", \"caption\": \"Comparison of PST and ESM-2 on ProteinShake tasks and VEP datasets.a\"}","artifact_sha256":"c21ad593de589a7188ca86a8b7ce617e301d48dd939ecc7da03efb22cbe8d7a3","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12603367/fullTextXML"},"legacy_id":"lit-023","legacy_row":{"id":"lit-023","paper_id":"pst-2025","domain_id":"proteins-complexes","task":"Zero-shot variant effect prediction","model":"PST","model_version":"","dataset":"ProteinShake VEP datasets","dataset_version":"","split":"","metric":"Mean |Spearman rho|","value":"0.501","unit":"unitless","uncertainty":"","protocol":"Zero-shot VEP; paper averages absolute Spearman correlations.","source_locator":"Table 2, PST row, Zero-shot VEP Mean |ρ| column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12603367/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:25:00Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-024","kind":"result","name":"ESM-2 · Mean |Spearman rho| · ProteinShake VEP datasets","description":"","status":"source_checked","facets":{"areas":["proteins-complexes"],"tasks":["Zero-shot variant effect prediction"]},"source_ids":["pst-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-024"}],"attributes":{"printed_value":"0.489","numeric_value":"0.489","metric":"Mean |Spearman rho|","metric_direction":"unknown","unit":"unitless","uncertainty":null,"source_locator":"Table 2, ESM-2 row, Zero-shot VEP Mean |ρ| column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.528656+00:00","notes":"Zero-shot VEP is the last metric group; selected Mean absolute rho, not GO/EC/binding-site metrics. This verifies the central score at its source location, not every metadata field or an experimental reproduction.","evidence":"{\"table_xml_id\": \"btaf582-T2\", \"row_cells\": [\"ESM-2\", \"0.648\", \"0.858\", \"0.698\", \"0.431\", \"0.791\", \"0.489\"], \"selected_cell_zero_based\": 6, \"selected_cell_xml\": \"<td rowspan=\\\"1\\\" colspan=\\\"1\\\">0.489</td>\", \"caption\": \"Comparison of PST and ESM-2 on ProteinShake tasks and VEP datasets.a\"}","artifact_sha256":"c21ad593de589a7188ca86a8b7ce617e301d48dd939ecc7da03efb22cbe8d7a3","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12603367/fullTextXML"},"legacy_id":"lit-024","legacy_row":{"id":"lit-024","paper_id":"pst-2025","domain_id":"proteins-complexes","task":"Zero-shot variant effect prediction","model":"ESM-2","model_version":"","dataset":"ProteinShake VEP datasets","dataset_version":"","split":"","metric":"Mean |Spearman rho|","value":"0.489","unit":"unitless","uncertainty":"","protocol":"Zero-shot VEP; paper averages absolute Spearman correlations.","source_locator":"Table 2, ESM-2 row, Zero-shot VEP Mean |ρ| column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12603367/","evaluation_origin":"independent_paper","reviewed_utc":"2026-09-15T23:25:00Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-025","kind":"result","name":"scGPT · F1-Score · M.S. single-cell dataset","description":"","status":"source_checked","facets":{"areas":["cells-tissues"],"tasks":["Cell-type identification"]},"source_ids":["single-cell-peft-2024"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-025"}],"attributes":{"printed_value":"0.734","numeric_value":"0.734","metric":"F1-Score","metric_direction":"unknown","unit":"unitless","uncertainty":null,"source_locator":"Table 2, M.S. / scGPT row, F1-Score column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.530269+00:00","notes":"Selected M.S. dataset block, first scGPT/Geneformer occurrences. F1-Score is last column; later dataset blocks deliberately excluded. This verifies the central score at its source location, not every metadata field or an experimental reproduction.","evidence":"{\"table_xml_id\": \"T2\", \"row_cells\": [\"M.S.\", \"scGPT\", \"0.595\", \"0.777\", \"0.728\", \"0.734\"], \"selected_cell_zero_based\": 5, \"selected_cell_xml\": \"<td align=\\\"center\\\" valign=\\\"bottom\\\" rowspan=\\\"1\\\" colspan=\\\"1\\\">0.734</td>\", \"caption\": \"Performance of cell type identification using native scLLMs and popular tools.Bold value represents the highest score among the methods\"}","artifact_sha256":"77a4a859010259eadf2187465db6ab385efa4927a5eadb95c1e01991044c283f","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC10862733/fullTextXML"},"legacy_id":"lit-025","legacy_row":{"id":"lit-025","paper_id":"single-cell-peft-2024","domain_id":"cells-tissues","task":"Cell-type identification","model":"scGPT","model_version":"","dataset":"M.S. single-cell dataset","dataset_version":"","split":"","metric":"F1-Score","value":"0.734","unit":"unitless","uncertainty":"","protocol":"Native scLLM cell-type identification as reported in Table 2.","source_locator":"Table 2, M.S. / scGPT row, F1-Score column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC10862733/","evaluation_origin":"independent_paper","reviewed_utc":"2026-09-15T23:25:00Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-026","kind":"result","name":"Geneformer · F1-Score · M.S. single-cell dataset","description":"","status":"source_checked","facets":{"areas":["cells-tissues"],"tasks":["Cell-type identification"]},"source_ids":["single-cell-peft-2024"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-026"}],"attributes":{"printed_value":"0.388","numeric_value":"0.388","metric":"F1-Score","metric_direction":"unknown","unit":"unitless","uncertainty":null,"source_locator":"Table 2, M.S. / Geneformer row, F1-Score column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.531756+00:00","notes":"Selected M.S. dataset block, first scGPT/Geneformer occurrences. F1-Score is last column; later dataset blocks deliberately excluded. This verifies the central score at its source location, not every metadata field or an experimental reproduction.","evidence":"{\"table_xml_id\": \"T2\", \"row_cells\": [\"\", \"Geneformer\", \"0.283\", \"0.235\", \"0.532\", \"0.388\"], \"selected_cell_zero_based\": 5, \"selected_cell_xml\": \"<td align=\\\"center\\\" valign=\\\"top\\\" rowspan=\\\"1\\\" colspan=\\\"1\\\">0.388</td>\", \"caption\": \"Performance of cell type identification using native scLLMs and popular tools.Bold value represents the highest score among the methods\"}","artifact_sha256":"77a4a859010259eadf2187465db6ab385efa4927a5eadb95c1e01991044c283f","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC10862733/fullTextXML"},"legacy_id":"lit-026","legacy_row":{"id":"lit-026","paper_id":"single-cell-peft-2024","domain_id":"cells-tissues","task":"Cell-type identification","model":"Geneformer","model_version":"","dataset":"M.S. single-cell dataset","dataset_version":"","split":"","metric":"F1-Score","value":"0.388","unit":"unitless","uncertainty":"","protocol":"Native scLLM cell-type identification as reported in Table 2.","source_locator":"Table 2, M.S. / Geneformer row, F1-Score column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC10862733/","evaluation_origin":"independent_paper","reviewed_utc":"2026-09-15T23:25:00Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-027","kind":"result","name":"C2S (GPT-2 Large) · Partial-label accuracy · L1000","description":"","status":"source_checked","facets":{"areas":["cells-tissues"],"tasks":["Combinatorial cell-label classification"]},"source_ids":["cell2sentence-2024"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-027"}],"attributes":{"printed_value":"0.631","numeric_value":"0.631","metric":"Partial-label accuracy","metric_direction":"unknown","unit":"unitless","uncertainty":"± 0.0031","source_locator":"Table 3, Partial label / C2S (GPT-2 Large) row, L1000 Acc column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.533640+00:00","notes":"Read inline small-caps/bold XML in document order, restoring Geneformer and GPT-2 Large labels. Selected Partial label (first block), L1000 > Acc, not AUROC or Full label. This verifies the central score at its source location, not every metadata field or an experimental reproduction.","evidence":"{\"table_xml_id\": \"T3\", \"row_cells\": [\"C2S (GPT-2 Large)\", \"0.639 ± 0.0049\", \"0.767 ± 0.0049\", \"0.631 ± 0.0031\", \"0.768 ± 0.0021\", \"0.575 ± 0.0035\", \"0.713 ± 0.0014\"], \"selected_cell_zero_based\": 3, \"selected_cell_xml\": \"<td align=\\\"center\\\" valign=\\\"middle\\\" rowspan=\\\"1\\\" colspan=\\\"1\\\"><bold>0.631</bold> ± <bold>0.0031</bold></td>\", \"caption\": \"Experimental results on downstream cell label classification. Cell labels are composed of multiple combinatorial metadata parts, including cell type, perturbations, and dosage information. Accuracy and area under ROC curve is computed on model predictions versus ground truth combinatorial labels, with partial credit given for partial misclassifications.\"}","artifact_sha256":"e088727d6e04857fccb7033a9b074e1850f775e86e7d2e99e603dde09558ab02","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11565894/fullTextXML"},"legacy_id":"lit-027","legacy_row":{"id":"lit-027","paper_id":"cell2sentence-2024","domain_id":"cells-tissues","task":"Combinatorial cell-label classification","model":"C2S (GPT-2 Large)","model_version":"GPT-2 Large","dataset":"L1000","dataset_version":"","split":"","metric":"Partial-label accuracy","value":"0.631","unit":"unitless","uncertainty":"± 0.0031","protocol":"Partial-credit labels including cell type, perturbation, and dose.","source_locator":"Table 3, Partial label / C2S (GPT-2 Large) row, L1000 Acc column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11565894/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:25:00Z"},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"lit-028","kind":"result","name":"Geneformer · Partial-label accuracy · L1000","description":"","status":"source_checked","facets":{"areas":["cells-tissues"],"tasks":["Combinatorial cell-label classification"]},"source_ids":["cell2sentence-2024"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-028"}],"attributes":{"printed_value":"0.419","numeric_value":"0.419","metric":"Partial-label accuracy","metric_direction":"unknown","unit":"unitless","uncertainty":"± 0.0153","source_locator":"Table 3, Partial label / Geneformer row, L1000 Acc column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.535220+00:00","notes":"Read inline small-caps/bold XML in document order, restoring Geneformer and GPT-2 Large labels. Selected Partial label (first block), L1000 > Acc, not AUROC or Full label. This verifies the central score at its source location, not every metadata field or an experimental reproduction.","evidence":"{\"table_xml_id\": \"T3\", \"row_cells\": [\"Geneformer\", \"0.600 ± 0.0170\", \"0.722 ± 0.0145\", \"0.419 ± 0.0153\", \"0.632 ± 0.0181\", \"0.500 ± 0.0013\", \"0.649 ± 0.0025\"], \"selected_cell_zero_based\": 3, \"selected_cell_xml\": \"<td align=\\\"center\\\" valign=\\\"middle\\\" rowspan=\\\"1\\\" colspan=\\\"1\\\">0.419 ± 0.0153</td>\", \"caption\": \"Experimental results on downstream cell label classification. Cell labels are composed of multiple combinatorial metadata parts, including cell type, perturbations, and dosage information. Accuracy and area under ROC curve is computed on model predictions versus ground truth combinatorial labels, with partial credit given for partial misclassifications.\"}","artifact_sha256":"e088727d6e04857fccb7033a9b074e1850f775e86e7d2e99e603dde09558ab02","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11565894/fullTextXML"},"legacy_id":"lit-028","legacy_row":{"id":"lit-028","paper_id":"cell2sentence-2024","domain_id":"cells-tissues","task":"Combinatorial cell-label classification","model":"Geneformer","model_version":"","dataset":"L1000","dataset_version":"","split":"","metric":"Partial-label accuracy","value":"0.419","unit":"unitless","uncertainty":"± 0.0153","protocol":"Partial-credit labels including cell type, perturbation, and dose.","source_locator":"Table 3, Partial label / Geneformer row, L1000 Acc column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11565894/","evaluation_origin":"independent_paper","reviewed_utc":"2026-09-15T23:25:00Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"lit-029","kind":"result","name":"scGPT · F1 · hPancreas","description":"","status":"source_checked","facets":{"areas":["cells-tissues"],"tasks":["Cell-type annotation"]},"source_ids":["scelmo-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-029"}],"attributes":{"printed_value":"0.550","numeric_value":"0.550","metric":"F1","metric_direction":"unknown","unit":"unitless","uncertainty":null,"source_locator":"Table 1, hPancreas zero-shot / scGPT (z) row, F1 column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.537541+00:00","notes":"Selected first hPancreas zero-shot block and F1 last column. Caption says some scores are copied from GenePT; this is source checking of the reported table, not independent experimental evidence. This verifies the central score at its source location, not every metadata field or an experimental reproduction.","evidence":"{\"table_xml_id\": \"T1\", \"row_cells\": [\"scGPT (z)\", \"0.770\", \"0.610\", \"0.560\", \"0.550\"], \"selected_cell_zero_based\": 4, \"selected_cell_xml\": \"<td align=\\\"right\\\" valign=\\\"top\\\" rowspan=\\\"1\\\" colspan=\\\"1\\\">0.550</td>\", \"caption\": \"Scores of cell-type annotation task under different settings. Parts of the results are directly extracted from GenePT. Here PCA represents principal component analysis, and scELMo+random emb represents fine-tuning scELMo with random numbers as meaningless gene embeddings. Average ranks of all methods across datasets are summarized in Extended Data Figure 6 (b). We boldfaced the highest score of each metric for each dataset.\"}","artifact_sha256":"ef75f0d63a567f5e9d7132fd847f44838a82a9741ae55323437e1d1812d86316","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12393277/fullTextXML"},"legacy_id":"lit-029","legacy_row":{"id":"lit-029","paper_id":"scelmo-2025","domain_id":"cells-tissues","task":"Cell-type annotation","model":"scGPT","model_version":"","dataset":"hPancreas","dataset_version":"","split":"","metric":"F1","value":"0.550","unit":"unitless","uncertainty":"","protocol":"Zero-shot setting; source caption says some comparator rows come from GenePT.","source_locator":"Table 1, hPancreas zero-shot / scGPT (z) row, F1 column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12393277/","evaluation_origin":"paper_compilation","reviewed_utc":"2026-09-15T23:25:00Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-030","kind":"result","name":"Geneformer · F1 · hPancreas","description":"","status":"source_checked","facets":{"areas":["cells-tissues"],"tasks":["Cell-type annotation"]},"source_ids":["scelmo-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-030"}],"attributes":{"printed_value":"0.270","numeric_value":"0.270","metric":"F1","metric_direction":"unknown","unit":"unitless","uncertainty":null,"source_locator":"Table 1, hPancreas zero-shot / Geneformer (z) row, F1 column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.539630+00:00","notes":"Selected first hPancreas zero-shot block and F1 last column. Caption says some scores are copied from GenePT; this is source checking of the reported table, not independent experimental evidence. This verifies the central score at its source location, not every metadata field or an experimental reproduction.","evidence":"{\"table_xml_id\": \"T1\", \"row_cells\": [\"Geneformer (z)\", \"0.500\", \"0.250\", \"0.340\", \"0.270\"], \"selected_cell_zero_based\": 4, \"selected_cell_xml\": \"<td align=\\\"right\\\" valign=\\\"top\\\" rowspan=\\\"1\\\" colspan=\\\"1\\\">0.270</td>\", \"caption\": \"Scores of cell-type annotation task under different settings. Parts of the results are directly extracted from GenePT. Here PCA represents principal component analysis, and scELMo+random emb represents fine-tuning scELMo with random numbers as meaningless gene embeddings. Average ranks of all methods across datasets are summarized in Extended Data Figure 6 (b). We boldfaced the highest score of each metric for each dataset.\"}","artifact_sha256":"ef75f0d63a567f5e9d7132fd847f44838a82a9741ae55323437e1d1812d86316","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12393277/fullTextXML"},"legacy_id":"lit-030","legacy_row":{"id":"lit-030","paper_id":"scelmo-2025","domain_id":"cells-tissues","task":"Cell-type annotation","model":"Geneformer","model_version":"","dataset":"hPancreas","dataset_version":"","split":"","metric":"F1","value":"0.270","unit":"unitless","uncertainty":"","protocol":"Zero-shot setting; source caption says some comparator rows come from GenePT.","source_locator":"Table 1, hPancreas zero-shot / Geneformer (z) row, F1 column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12393277/","evaluation_origin":"paper_compilation","reviewed_utc":"2026-09-15T23:25:00Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-031","kind":"result","name":"scRegNet (Geneformer backbone) · AUROC · hESC cell-type-specific GRN","description":"","status":"source_checked","facets":{"areas":["cells-tissues"],"tasks":["Gene-regulatory link prediction"]},"source_ids":["scregnet-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-031"}],"attributes":{"printed_value":"0.89","numeric_value":"0.89","metric":"AUROC","metric_direction":"unknown","unit":"unitless","uncertainty":"± 0.00 as printed","source_locator":"Table 2, scRegNet (w/ Geneformer) row, hESC AUROC entry","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.541287+00:00","notes":"Read break elements: cells contain AUROC on first line then AUPRC. Selected hESC (first cell type), first line. Caption specifies 500 most-variable genes and 50 independent evaluations. This verifies the central score at its source location, not every metadata field or an experimental reproduction.","evidence":"{\"table_xml_id\": \"T2\", \"row_cells\": [\"scRegNet (w/ Geneformer)\", \"AUROCAUPRC\", \"0.89±0.000.62±0.00\", \"0.90±0.000.84±0.00\", \"0.81±0.000.17±0.00\", \"0.93±0.000.86±0.00\", \"0.92±0.000.94±0.00\", \"0.93±0.000.94±0.00\", \"0.88±0.000.88±0.00\"], \"selected_cell_zero_based\": 2, \"selected_cell_xml\": \"<td align=\\\"center\\\" valign=\\\"top\\\" rowspan=\\\"1\\\" colspan=\\\"1\\\"><bold>0.89</bold>±0.00<break /><bold>0.62</bold>±0.00</td>\", \"caption\": \"Link prediction performance on seven scRNA-seq datasets with 500 most-variable genes. Each dataset includes a cell-type-specific ground-truth network. The values reported are averages from 50 independent evaluations per cell type. scRegNet utilizing the three backbone models—scBERT, Geneformer, and scFoundation—consistently outperforms the baselines.\"}","artifact_sha256":"65b3272d47bb9c4ee1e7a965169bef63add9dbeb31508d4076e5145b761af4ec","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11838224/fullTextXML"},"legacy_id":"lit-031","legacy_row":{"id":"lit-031","paper_id":"scregnet-2025","domain_id":"cells-tissues","task":"Gene-regulatory link prediction","model":"scRegNet (Geneformer backbone)","model_version":"","dataset":"hESC cell-type-specific GRN","dataset_version":"","split":"","metric":"AUROC","value":"0.89","unit":"unitless","uncertainty":"± 0.00 as printed","protocol":"TFs plus 500 variable genes; mean from 50 independent evaluations.","source_locator":"Table 2, scRegNet (w/ Geneformer) row, hESC AUROC entry","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11838224/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:25:00Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"lit-032","kind":"result","name":"scRegNet (scBERT backbone) · AUROC · hESC cell-type-specific GRN","description":"","status":"source_checked","facets":{"areas":["cells-tissues"],"tasks":["Gene-regulatory link prediction"]},"source_ids":["scregnet-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-032"}],"attributes":{"printed_value":"0.88","numeric_value":"0.88","metric":"AUROC","metric_direction":"unknown","unit":"unitless","uncertainty":"± 0.00 as printed","source_locator":"Table 2, scRegNet (w/ scBERT) row, hESC AUROC entry","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.542807+00:00","notes":"Read break elements: cells contain AUROC on first line then AUPRC. Selected hESC (first cell type), first line. Caption specifies 500 most-variable genes and 50 independent evaluations. This verifies the central score at its source location, not every metadata field or an experimental reproduction.","evidence":"{\"table_xml_id\": \"T2\", \"row_cells\": [\"scRegNet (w/ scBERT)\", \"AUROCAUPRC\", \"0.88±0.000.61±0.00\", \"0.90±0.000.83±0.00\", \"0.75±0.010.12±0.01\", \"0.92±0.000.84±0.00\", \"0.92±0.000.94±0.00\", \"0.92±0.000.93±0.00\", \"0.85±0.010.85±0.01\"], \"selected_cell_zero_based\": 2, \"selected_cell_xml\": \"<td align=\\\"center\\\" valign=\\\"top\\\" rowspan=\\\"1\\\" colspan=\\\"1\\\">0.88±0.00<break />0.61±0.00</td>\", \"caption\": \"Link prediction performance on seven scRNA-seq datasets with 500 most-variable genes. Each dataset includes a cell-type-specific ground-truth network. The values reported are averages from 50 independent evaluations per cell type. scRegNet utilizing the three backbone models—scBERT, Geneformer, and scFoundation—consistently outperforms the baselines.\"}","artifact_sha256":"65b3272d47bb9c4ee1e7a965169bef63add9dbeb31508d4076e5145b761af4ec","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11838224/fullTextXML"},"legacy_id":"lit-032","legacy_row":{"id":"lit-032","paper_id":"scregnet-2025","domain_id":"cells-tissues","task":"Gene-regulatory link prediction","model":"scRegNet (scBERT backbone)","model_version":"","dataset":"hESC cell-type-specific GRN","dataset_version":"","split":"","metric":"AUROC","value":"0.88","unit":"unitless","uncertainty":"± 0.00 as printed","protocol":"TFs plus 500 variable genes; mean from 50 independent evaluations.","source_locator":"Table 2, scRegNet (w/ scBERT) row, hESC AUROC entry","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11838224/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:25:00Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"lit-033","kind":"result","name":"ProkBERT-mini · Accuracy · E. coli sigma70 promoter dataset","description":"","status":"source_checked","facets":{"areas":["microbes-communities"],"tasks":["E. coli sigma70 promoter prediction"]},"source_ids":["prokbert-2024"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-033"}],"attributes":{"printed_value":"0.87","numeric_value":"0.87","metric":"Accuracy","metric_direction":"unknown","unit":"unitless","uncertainty":null,"source_locator":"Table 3, ProkBERT-mini row, Accuracy column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:36.197Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness.","evidence":"Table 3, ProkBERT-mini row, Accuracy column; cell: 0.87","artifact_sha256":"8610e2a54aa877c8dc565a9cdb6e82099f284c5e0907a52cab18d994ea732436","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC10810988/fullTextXML"},"legacy_id":"lit-033","legacy_row":{"id":"lit-033","paper_id":"prokbert-2024","domain_id":"microbes-communities","task":"E. coli sigma70 promoter prediction","model":"ProkBERT-mini","model_version":"","dataset":"E. coli sigma70 promoter dataset","dataset_version":"","split":"","metric":"Accuracy","value":"0.87","unit":"unitless","uncertainty":"","protocol":"Promoter versus non-promoter classification.","source_locator":"Table 3, ProkBERT-mini row, Accuracy column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC10810988/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:25:00Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-034","kind":"result","name":"Promotech · Accuracy · E. coli sigma70 promoter dataset","description":"","status":"source_checked","facets":{"areas":["microbes-communities"],"tasks":["E. coli sigma70 promoter prediction"]},"source_ids":["prokbert-2024"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-034"}],"attributes":{"printed_value":"0.71","numeric_value":"0.71","metric":"Accuracy","metric_direction":"unknown","unit":"unitless","uncertainty":null,"source_locator":"Table 3, Promotech row, Accuracy column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:36.197Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness.","evidence":"Table 3, Promotech row, Accuracy column; cell: 0.71","artifact_sha256":"8610e2a54aa877c8dc565a9cdb6e82099f284c5e0907a52cab18d994ea732436","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC10810988/fullTextXML"},"legacy_id":"lit-034","legacy_row":{"id":"lit-034","paper_id":"prokbert-2024","domain_id":"microbes-communities","task":"E. coli sigma70 promoter prediction","model":"Promotech","model_version":"","dataset":"E. coli sigma70 promoter dataset","dataset_version":"","split":"","metric":"Accuracy","value":"0.71","unit":"unitless","uncertainty":"","protocol":"Promoter versus non-promoter classification.","source_locator":"Table 3, Promotech row, Accuracy column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC10810988/","evaluation_origin":"independent_paper","reviewed_utc":"2026-09-15T23:25:00Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-035","kind":"result","name":"Eco70PromBERT · Promoter-class F1 · Independent E. coli sigma70 test dataset","description":"","status":"source_checked","facets":{"areas":["microbes-communities"],"tasks":["E. coli sigma70 promoter prediction"]},"source_ids":["cyaprombert-2022"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-035"}],"attributes":{"printed_value":"0.91","numeric_value":"0.91","metric":"Promoter-class F1","metric_direction":"unknown","unit":"unitless","uncertainty":null,"source_locator":"TABLE 3, Eco70PromBERT (BERT-base + 1bp tokenizer) row, F1 score Promoter column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.544033+00:00","notes":"F1 score is the third two-column group; selected Promoter subcolumn, not AUROC or precision. This verifies the central score at its source location, not every metadata field or an experimental reproduction.","evidence":"{\"table_xml_id\": \"T3\", \"row_cells\": [\"Eco70PromBERT (BERT-base + 1bp tokenizer)\", \"0.92\", \"0.90\", \"0.91\", \"0.91\", \"0.91\", \"0.91\", \"110\", \"108\"], \"selected_cell_zero_based\": 5, \"selected_cell_xml\": \"<td align=\\\"left\\\" rowspan=\\\"1\\\" colspan=\\\"1\\\">0.91</td>\", \"caption\": \"Performance of Eco70PromBERT and popular promoter prediction models for E.coli using an independent dataset (σ70 promoters and non-promoters).\"}","artifact_sha256":"74278ccd77b2bc00a3f4434546545e8bdec8b0652a0e5d1862ec0f91decccd8d","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC9745317/fullTextXML"},"legacy_id":"lit-035","legacy_row":{"id":"lit-035","paper_id":"cyaprombert-2022","domain_id":"microbes-communities","task":"E. coli sigma70 promoter prediction","model":"Eco70PromBERT","model_version":"","dataset":"Independent E. coli sigma70 test dataset","dataset_version":"","split":"independent test","metric":"Promoter-class F1","value":"0.91","unit":"unitless","uncertainty":"","protocol":"BERT-base with 1bp tokenizer; 110 promoters and 108 non-promoters.","source_locator":"TABLE 3, Eco70PromBERT (BERT-base + 1bp tokenizer) row, F1 score Promoter column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC9745317/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:25:00Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-036","kind":"result","name":"iPro70-FMWin · Promoter-class F1 · Independent E. coli sigma70 test dataset","description":"","status":"source_checked","facets":{"areas":["microbes-communities"],"tasks":["E. coli sigma70 promoter prediction"]},"source_ids":["cyaprombert-2022"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-036"}],"attributes":{"printed_value":"0.90","numeric_value":"0.90","metric":"Promoter-class F1","metric_direction":"unknown","unit":"unitless","uncertainty":null,"source_locator":"TABLE 3, iPro70-FMWin row, F1 score Promoter column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.544926+00:00","notes":"F1 score is the third two-column group; selected Promoter subcolumn, not AUROC or precision. This verifies the central score at its source location, not every metadata field or an experimental reproduction.","evidence":"{\"table_xml_id\": \"T3\", \"row_cells\": [\"iPro70-FMWin\", \"0.90\", \"0.90\", \"0.93\", \"0.88\", \"0.90\", \"0.91\", \"110\", \"108\"], \"selected_cell_zero_based\": 5, \"selected_cell_xml\": \"<td align=\\\"left\\\" rowspan=\\\"1\\\" colspan=\\\"1\\\">0.90</td>\", \"caption\": \"Performance of Eco70PromBERT and popular promoter prediction models for E.coli using an independent dataset (σ70 promoters and non-promoters).\"}","artifact_sha256":"74278ccd77b2bc00a3f4434546545e8bdec8b0652a0e5d1862ec0f91decccd8d","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC9745317/fullTextXML"},"legacy_id":"lit-036","legacy_row":{"id":"lit-036","paper_id":"cyaprombert-2022","domain_id":"microbes-communities","task":"E. coli sigma70 promoter prediction","model":"iPro70-FMWin","model_version":"","dataset":"Independent E. coli sigma70 test dataset","dataset_version":"","split":"independent test","metric":"Promoter-class F1","value":"0.90","unit":"unitless","uncertainty":"","protocol":"Compared on the same independent test dataset; 110 promoters and 108 non-promoters.","source_locator":"TABLE 3, iPro70-FMWin row, F1 score Promoter column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC9745317/","evaluation_origin":"independent_paper","reviewed_utc":"2026-09-15T23:25:00Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-037","kind":"result","name":"EVO2 · MCC · LAMBDA genome-wide prophage test","description":"","status":"source_checked","facets":{"areas":["microbes-communities"],"tasks":["Genome-wide prophage detection"]},"source_ids":["lambda-prophage-2026"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-037"}],"attributes":{"printed_value":"0.680","numeric_value":"0.680","metric":"MCC","metric_direction":"unknown","unit":"unitless","uncertainty":null,"source_locator":"Table 5, EVO2 row, MCC column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:36.240Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness.","evidence":"Table 5, EVO2 row, MCC column; cell: 0.680","artifact_sha256":"22c2e218e87dce757907f6086a0e2ad37c13f785b34fff5bea7cfa1a6c276b16","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC13041943/fullTextXML"},"legacy_id":"lit-037","legacy_row":{"id":"lit-037","paper_id":"lambda-prophage-2026","domain_id":"microbes-communities","task":"Genome-wide prophage detection","model":"EVO2","model_version":"","dataset":"LAMBDA genome-wide prophage test","dataset_version":"","split":"","metric":"MCC","value":"0.680","unit":"unitless","uncertainty":"","protocol":"Genomic language model fine-tuned for prophage detection; genome-wide evaluation.","source_locator":"Table 5, EVO2 row, MCC column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC13041943/","evaluation_origin":"independent_paper","reviewed_utc":"2026-09-15T23:25:00Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-038","kind":"result","name":"geNomad · MCC · LAMBDA genome-wide prophage test","description":"","status":"source_checked","facets":{"areas":["microbes-communities"],"tasks":["Genome-wide prophage detection"]},"source_ids":["lambda-prophage-2026"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-038"}],"attributes":{"printed_value":"0.794","numeric_value":"0.794","metric":"MCC","metric_direction":"unknown","unit":"unitless","uncertainty":null,"source_locator":"Table 5, geNomad row, MCC column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:36.240Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness.","evidence":"Table 5, geNomad row, MCC column; cell: 0.794","artifact_sha256":"22c2e218e87dce757907f6086a0e2ad37c13f785b34fff5bea7cfa1a6c276b16","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC13041943/fullTextXML"},"legacy_id":"lit-038","legacy_row":{"id":"lit-038","paper_id":"lambda-prophage-2026","domain_id":"microbes-communities","task":"Genome-wide prophage detection","model":"geNomad","model_version":"","dataset":"LAMBDA genome-wide prophage test","dataset_version":"","split":"","metric":"MCC","value":"0.794","unit":"unitless","uncertainty":"","protocol":"Traditional specialist comparator; genome-wide evaluation.","source_locator":"Table 5, geNomad row, MCC column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC13041943/","evaluation_origin":"independent_paper","reviewed_utc":"2026-09-15T23:25:00Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-039","kind":"result","name":"NABAS+ · F1 score · CAMI II Toy human gastrooral sample19-new","description":"","status":"source_checked","facets":{"areas":["microbes-communities"],"tasks":["Metagenomic taxonomic classification"]},"source_ids":["nabas-plus-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-039"}],"attributes":{"printed_value":"0.719","numeric_value":"0.719","metric":"F1 score","metric_direction":"unknown","unit":"unitless","uncertainty":null,"source_locator":"Table 3, Sample19-new / NABAS+ row, F1 score column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.546107+00:00","notes":"Selected Sample19-new explicitly, not Sample19-old, and F1 rather than precision/recall. This verifies the central score at its source location, not every metadata field or an experimental reproduction.","evidence":"{\"table_xml_id\": \"tbl3\", \"row_cells\": [\"Sample19-new\", \"NABAS+\", \"0.719\", \"0.719\", \"0.719\"], \"selected_cell_zero_based\": 4, \"selected_cell_xml\": \"<td rowspan=\\\"1\\\" colspan=\\\"1\\\">0.719</td>\", \"caption\": \"Performance of the classifier on the original and newly generated sample19\"}","artifact_sha256":"49903e751beb86f6744825d2fdb3ea2fbe327b52b1ce68c323bfa8b66dae71ec","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12231603/fullTextXML"},"legacy_id":"lit-039","legacy_row":{"id":"lit-039","paper_id":"nabas-plus-2025","domain_id":"microbes-communities","task":"Metagenomic taxonomic classification","model":"NABAS+","model_version":"","dataset":"CAMI II Toy human gastrooral sample19-new","dataset_version":"","split":"","metric":"F1 score","value":"0.719","unit":"unitless","uncertainty":"","protocol":"Newly generated sample19 used for classifier comparison.","source_locator":"Table 3, Sample19-new / NABAS+ row, F1 score column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12231603/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:25:00Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-040","kind":"result","name":"MetaPhlAn3 · F1 score · CAMI II Toy human gastrooral sample19-new","description":"","status":"source_checked","facets":{"areas":["microbes-communities"],"tasks":["Metagenomic taxonomic classification"]},"source_ids":["nabas-plus-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-040"}],"attributes":{"printed_value":"0.753","numeric_value":"0.753","metric":"F1 score","metric_direction":"unknown","unit":"unitless","uncertainty":null,"source_locator":"Table 3, Sample19-new / MetaPhlAn3 row, F1 score column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.547023+00:00","notes":"Selected Sample19-new explicitly, not Sample19-old, and F1 rather than precision/recall. This verifies the central score at its source location, not every metadata field or an experimental reproduction.","evidence":"{\"table_xml_id\": \"tbl3\", \"row_cells\": [\"Sample19-new\", \"MetaPhlAn3\", \"0.778\", \"0.729\", \"0.753\"], \"selected_cell_zero_based\": 4, \"selected_cell_xml\": \"<td rowspan=\\\"1\\\" colspan=\\\"1\\\">0.753</td>\", \"caption\": \"Performance of the classifier on the original and newly generated sample19\"}","artifact_sha256":"49903e751beb86f6744825d2fdb3ea2fbe327b52b1ce68c323bfa8b66dae71ec","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12231603/fullTextXML"},"legacy_id":"lit-040","legacy_row":{"id":"lit-040","paper_id":"nabas-plus-2025","domain_id":"microbes-communities","task":"Metagenomic taxonomic classification","model":"MetaPhlAn3","model_version":"","dataset":"CAMI II Toy human gastrooral sample19-new","dataset_version":"","split":"","metric":"F1 score","value":"0.753","unit":"unitless","uncertainty":"","protocol":"Newly generated sample19 used for classifier comparison.","source_locator":"Table 3, Sample19-new / MetaPhlAn3 row, F1 score column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12231603/","evaluation_origin":"independent_paper","reviewed_utc":"2026-09-15T23:25:00Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-041","kind":"result","name":"Chai-1 · Success rate, ligand all-atom RMSD <2 Å · LiPP lipid–protein complexes","description":"","status":"source_checked","facets":{"areas":["molecular-interactions"],"tasks":["Lipid–protein binding pose"]},"source_ids":["lipp-2026"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-041"}],"attributes":{"printed_value":"60.7","numeric_value":"60.7","metric":"Success rate, ligand all-atom RMSD <2 Å","metric_direction":"unknown","unit":"%","uncertainty":"95% CI 55.2–66.0","source_locator":"Table 2, Chai-1 row, LiPP (N=331) % Success Rate column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.548973+00:00","notes":"JATS label is bare 2, which caused original parser miss. LiPP N=331 full-set column selected, not N=36 test subset. Caption success is lipid all-atom RMSD <2 Angstrom; PB-valid is a separate table. This verifies the central score at its source location, not every metadata field or an experimental reproduction.","evidence":"{\"table_xml_id\": \"tbl2\", \"row_cells\": [\"Chai-1\", \"60.7 {55.2–66.0}\", \"36.1 {20.8–53.7}\", \"77\", \"-\"], \"selected_cell_zero_based\": 1, \"selected_cell_xml\": \"<td align=\\\"left\\\" colspan=\\\"1\\\" rowspan=\\\"1\\\">60.7 {55.2–66.0}</td>\", \"caption\": \"Success Rates (Success Defined Only by Lipid Pose All-Atom RMSD Cutoff Values Less Than 2 Å) of the Five Computational Methods Used in This Study on Lipid–Protein Complexes (via LiPP Benchmark Set) Compared to Protein-Small Molecule Complexes (via PoseBusters Benchmark Set)\"}","artifact_sha256":"6ff34f2f709a14858a3753abf9f8f6efa1e7e3c351f15c70cf64264193a9414e","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC13292216/fullTextXML"},"legacy_id":"lit-041","legacy_row":{"id":"lit-041","paper_id":"lipp-2026","domain_id":"molecular-interactions","task":"Lipid–protein binding pose","model":"Chai-1","model_version":"","dataset":"LiPP lipid–protein complexes","dataset_version":"331 complexes","split":"","metric":"Success rate, ligand all-atom RMSD <2 Å","value":"60.7","unit":"%","uncertainty":"95% CI 55.2–66.0","protocol":"Top-scoring pose; all-atom lipid RMSD below 2 Å.","source_locator":"Table 2, Chai-1 row, LiPP (N=331) % Success Rate column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC13292216/","evaluation_origin":"independent_paper","reviewed_utc":"2026-09-15T23:25:00Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"lit-042","kind":"result","name":"DiffDock-L · Success rate, ligand all-atom RMSD <2 Å · LiPP lipid–protein complexes","description":"","status":"source_checked","facets":{"areas":["molecular-interactions"],"tasks":["Lipid–protein binding pose"]},"source_ids":["lipp-2026"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-042"}],"attributes":{"printed_value":"46.8","numeric_value":"46.8","metric":"Success rate, ligand all-atom RMSD <2 Å","metric_direction":"unknown","unit":"%","uncertainty":"95% CI 41.3–52.3","source_locator":"Table 2, DiffDock-L row, LiPP (N=331) % Success Rate column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.550691+00:00","notes":"JATS label is bare 2, which caused original parser miss. LiPP N=331 full-set column selected, not N=36 test subset. Caption success is lipid all-atom RMSD <2 Angstrom; PB-valid is a separate table. This verifies the central score at its source location, not every metadata field or an experimental reproduction.","evidence":"{\"table_xml_id\": \"tbl2\", \"row_cells\": [\"DiffDock-L\", \"46.8 {41.3–52.3}\", \"30.5 {16.3–48.1}\", \"-\", \"50\"], \"selected_cell_zero_based\": 1, \"selected_cell_xml\": \"<td align=\\\"left\\\" colspan=\\\"1\\\" rowspan=\\\"1\\\">46.8 {41.3–52.3}</td>\", \"caption\": \"Success Rates (Success Defined Only by Lipid Pose All-Atom RMSD Cutoff Values Less Than 2 Å) of the Five Computational Methods Used in This Study on Lipid–Protein Complexes (via LiPP Benchmark Set) Compared to Protein-Small Molecule Complexes (via PoseBusters Benchmark Set)\"}","artifact_sha256":"6ff34f2f709a14858a3753abf9f8f6efa1e7e3c351f15c70cf64264193a9414e","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC13292216/fullTextXML"},"legacy_id":"lit-042","legacy_row":{"id":"lit-042","paper_id":"lipp-2026","domain_id":"molecular-interactions","task":"Lipid–protein binding pose","model":"DiffDock-L","model_version":"","dataset":"LiPP lipid–protein complexes","dataset_version":"331 complexes","split":"","metric":"Success rate, ligand all-atom RMSD <2 Å","value":"46.8","unit":"%","uncertainty":"95% CI 41.3–52.3","protocol":"Top-scoring pose; all-atom lipid RMSD below 2 Å.","source_locator":"Table 2, DiffDock-L row, LiPP (N=331) % Success Rate column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC13292216/","evaluation_origin":"independent_paper","reviewed_utc":"2026-09-15T23:25:00Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"lit-043","kind":"result","name":"DiffDock-NMDN · Forward-screening success rate · CASF-2016 blind docked poses","description":"","status":"source_checked","facets":{"areas":["molecular-interactions"],"tasks":["Protein–ligand virtual screening"]},"source_ids":["nmdn-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-043"}],"attributes":{"printed_value":"66.7","numeric_value":"66.7","metric":"Forward-screening success rate","metric_direction":"unknown","unit":"%","uncertainty":null,"source_locator":"Table 2, DiffDock-NMDN / NMDN row, success rate (%) column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.552697+00:00","notes":"All scoring functions share DiffDock-NMDN poses via rowspan. Selected forward-screening success percentage, not docking pose success or scoring-power correlation. This verifies the central score at its source location, not every metadata field or an experimental reproduction.","evidence":"{\"table_xml_id\": \"tbl2\", \"row_cells\": [\"DiffDock-NMDN\", \"NMDN\", \"35.85\", \"66.7\", \"0.376\", \"0.458\"], \"selected_cell_zero_based\": 3, \"selected_cell_xml\": \"<td style=\\\"border:none;\\\" align=\\\"center\\\" colspan=\\\"1\\\" rowspan=\\\"1\\\"><bold>66.7</bold></td>\", \"caption\": \"Performance of Representative Scoring Functions on the CASF-2016 Using the DiffDock-NMDN Blind Docked Posesa\"}","artifact_sha256":"194b21478aaedd9a7384cabb8b0040ca5b6a4938f4d275627b86a6b787affc20","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11815853/fullTextXML"},"legacy_id":"lit-043","legacy_row":{"id":"lit-043","paper_id":"nmdn-2025","domain_id":"molecular-interactions","task":"Protein–ligand virtual screening","model":"DiffDock-NMDN","model_version":"","dataset":"CASF-2016 blind docked poses","dataset_version":"","split":"","metric":"Forward-screening success rate","value":"66.7","unit":"%","uncertainty":"","protocol":"NMDN scoring on DiffDock-NMDN blind docked poses; not ligand-pose RMSD.","source_locator":"Table 2, DiffDock-NMDN / NMDN row, success rate (%) column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11815853/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:25:00Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-044","kind":"result","name":"Vina · Forward-screening success rate · CASF-2016 blind docked poses","description":"","status":"source_checked","facets":{"areas":["molecular-interactions"],"tasks":["Protein–ligand virtual screening"]},"source_ids":["nmdn-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-044"}],"attributes":{"printed_value":"42.1","numeric_value":"42.1","metric":"Forward-screening success rate","metric_direction":"unknown","unit":"%","uncertainty":null,"source_locator":"Table 2, Vina scoring row, success rate (%) column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.554518+00:00","notes":"All scoring functions share DiffDock-NMDN poses via rowspan. Selected forward-screening success percentage, not docking pose success or scoring-power correlation. This verifies the central score at its source location, not every metadata field or an experimental reproduction.","evidence":"{\"table_xml_id\": \"tbl2\", \"row_cells\": [\"Vina38\", \"15.18\", \"42.1\", \"0.340\", \"0.323\"], \"selected_cell_zero_based\": 2, \"selected_cell_xml\": \"<td style=\\\"border:none;\\\" align=\\\"center\\\" colspan=\\\"1\\\" rowspan=\\\"1\\\">42.1</td>\", \"caption\": \"Performance of Representative Scoring Functions on the CASF-2016 Using the DiffDock-NMDN Blind Docked Posesa\"}","artifact_sha256":"194b21478aaedd9a7384cabb8b0040ca5b6a4938f4d275627b86a6b787affc20","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11815853/fullTextXML"},"legacy_id":"lit-044","legacy_row":{"id":"lit-044","paper_id":"nmdn-2025","domain_id":"molecular-interactions","task":"Protein–ligand virtual screening","model":"Vina","model_version":"","dataset":"CASF-2016 blind docked poses","dataset_version":"","split":"","metric":"Forward-screening success rate","value":"42.1","unit":"%","uncertainty":"","protocol":"Vina scoring on the same DiffDock-NMDN blind docked poses; not ligand-pose RMSD.","source_locator":"Table 2, Vina scoring row, success rate (%) column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11815853/","evaluation_origin":"independent_paper","reviewed_utc":"2026-09-15T23:25:00Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-045","kind":"result","name":"Boltz-1 · Median ligand RMSD · PLINDER-L95","description":"","status":"source_checked","facets":{"areas":["molecular-interactions"],"tasks":["Protein–ligand pose prediction"]},"source_ids":["boltz-stereochemistry-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-045"}],"attributes":{"printed_value":"1.393","numeric_value":"1.393","metric":"Median ligand RMSD","metric_direction":"unknown","unit":"Å","uncertainty":null,"source_locator":"Table 1, Boltz-1 row, Ligand RMSD (Å) column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.555674+00:00","notes":"JATS label is bare 1. Selected Ligand RMSD column in Plinder-L95, not Protein RMSD; retained method row without refinement settings. This verifies the central score at its source location, not every metadata field or an experimental reproduction.","evidence":"{\"table_xml_id\": \"tbl1\", \"row_cells\": [\"Boltz-1\", \"\", \"0.4041\", \"1.393\", \"60.58\", \"0.0283\", \"0.4381\"], \"selected_cell_zero_based\": 3, \"selected_cell_xml\": \"<td align=\\\"center\\\" colspan=\\\"1\\\" rowspan=\\\"1\\\">1.393</td>\", \"caption\": \"Evaluation Metrics for Protein and Ligand Structure Prediction Performance across All Entries in the Plinder-L95 Data Set\"}","artifact_sha256":"78a77b9a0ab8bfa371f5b9baef3f443f4590d6e71cf864d67e90e9ebdfa7fc1b","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12658688/fullTextXML"},"legacy_id":"lit-045","legacy_row":{"id":"lit-045","paper_id":"boltz-stereochemistry-2025","domain_id":"molecular-interactions","task":"Protein–ligand pose prediction","model":"Boltz-1","model_version":"","dataset":"PLINDER-L95","dataset_version":"","split":"","metric":"Median ligand RMSD","value":"1.393","unit":"Å","uncertainty":"","protocol":"All entries; authors note this dataset contains structures seen during model training.","source_locator":"Table 1, Boltz-1 row, Ligand RMSD (Å) column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12658688/","evaluation_origin":"independent_paper","reviewed_utc":"2026-09-15T23:25:00Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-046","kind":"result","name":"DiffDock · Median ligand RMSD · PLINDER-L95","description":"","status":"source_checked","facets":{"areas":["molecular-interactions"],"tasks":["Protein–ligand pose prediction"]},"source_ids":["boltz-stereochemistry-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-046"}],"attributes":{"printed_value":"1.342","numeric_value":"1.342","metric":"Median ligand RMSD","metric_direction":"unknown","unit":"Å","uncertainty":null,"source_locator":"Table 1, DiffDock row, Ligand RMSD (Å) column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.556572+00:00","notes":"JATS label is bare 1. Selected Ligand RMSD column in Plinder-L95, not Protein RMSD; retained method row without refinement settings. This verifies the central score at its source location, not every metadata field or an experimental reproduction.","evidence":"{\"table_xml_id\": \"tbl1\", \"row_cells\": [\"DiffDock\", \"\", \"\", \"1.342\", \"100\", \"0\", \"0\"], \"selected_cell_zero_based\": 3, \"selected_cell_xml\": \"<td align=\\\"center\\\" colspan=\\\"1\\\" rowspan=\\\"1\\\">1.342</td>\", \"caption\": \"Evaluation Metrics for Protein and Ligand Structure Prediction Performance across All Entries in the Plinder-L95 Data Set\"}","artifact_sha256":"78a77b9a0ab8bfa371f5b9baef3f443f4590d6e71cf864d67e90e9ebdfa7fc1b","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12658688/fullTextXML"},"legacy_id":"lit-046","legacy_row":{"id":"lit-046","paper_id":"boltz-stereochemistry-2025","domain_id":"molecular-interactions","task":"Protein–ligand pose prediction","model":"DiffDock","model_version":"","dataset":"PLINDER-L95","dataset_version":"","split":"","metric":"Median ligand RMSD","value":"1.342","unit":"Å","uncertainty":"","protocol":"All entries; rigid-protein docking comparator; authors note this dataset contains structures seen during model training.","source_locator":"Table 1, DiffDock row, Ligand RMSD (Å) column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12658688/","evaluation_origin":"independent_paper","reviewed_utc":"2026-09-15T23:25:00Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-047","kind":"result","name":"Boltz-2 · Pearson R · SARS-CoV-2 Mpro ligands","description":"","status":"source_checked","facets":{"areas":["molecular-interactions"],"tasks":["Ligand potency prediction using generated poses"]},"source_ids":["mpro-pose-affinity-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-047"}],"attributes":{"printed_value":"0.800","numeric_value":"0.800","metric":"Pearson R","metric_direction":"unknown","unit":"unitless","uncertainty":"± 0.027","source_locator":"Table 3, Boltz-2 row, Pearson’s R column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.557756+00:00","notes":"JATS label is bare 3. Selected SARS-CoV-2 Mpro potency Pearson R, not MERS-CoV table 2 or Boltz-2-Internal row; uncertainty retained in evidence. This verifies the central score at its source location, not every metadata field or an experimental reproduction.","evidence":"{\"table_xml_id\": \"tbl3\", \"row_cells\": [\"Boltz-2\", \"0.716 ± 0.036\", \"0.909 ± 0.043\", \"0.800 ± 0.027\", \"9.54 × 10–59\", \"0.532 ± 0.043\", \"0.598 ± 0.024\", \"0.799 ± 0.012\"], \"selected_cell_zero_based\": 3, \"selected_cell_xml\": \"<td align=\\\"center\\\" colspan=\\\"1\\\" rowspan=\\\"1\\\">0.800 ± 0.027</td>\", \"caption\": \"Statistical Performance Metrics of Potency Prediction for SARS-CoV-2 Mpro Using Different Ligand Pose Generation Protocols\"}","artifact_sha256":"c356a1c65a0033e5ae18a05d4afab5495856c5b6869328ff49e13547a4801a57","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12801289/fullTextXML"},"legacy_id":"lit-047","legacy_row":{"id":"lit-047","paper_id":"mpro-pose-affinity-2025","domain_id":"molecular-interactions","task":"Ligand potency prediction using generated poses","model":"Boltz-2","model_version":"","dataset":"SARS-CoV-2 Mpro ligands","dataset_version":"","split":"","metric":"Pearson R","value":"0.800","unit":"unitless","uncertainty":"± 0.027","protocol":"Potency prediction using Boltz-2 ligand-pose generation protocol; see paper scoring pipeline.","source_locator":"Table 3, Boltz-2 row, Pearson’s R column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12801289/","evaluation_origin":"independent_paper","reviewed_utc":"2026-09-15T23:25:00Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"lit-048","kind":"result","name":"DiffDock · Pearson R · SARS-CoV-2 Mpro ligands","description":"","status":"source_checked","facets":{"areas":["molecular-interactions"],"tasks":["Ligand potency prediction using generated poses"]},"source_ids":["mpro-pose-affinity-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-048"}],"attributes":{"printed_value":"0.695","numeric_value":"0.695","metric":"Pearson R","metric_direction":"unknown","unit":"unitless","uncertainty":"± 0.037","source_locator":"Table 3, DiffDock row, Pearson’s R column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:41:16.558815+00:00","notes":"JATS label is bare 3. Selected SARS-CoV-2 Mpro potency Pearson R, not MERS-CoV table 2 or Boltz-2-Internal row; uncertainty retained in evidence. This verifies the central score at its source location, not every metadata field or an experimental reproduction.","evidence":"{\"table_xml_id\": \"tbl3\", \"row_cells\": [\"DiffDock\", \"0.973 ± 0.044\", \"1.192 ± 0.045\", \"0.695 ± 0.037\", \"9.93 × 10–39\", \"0.195 ± 0.061\", \"0.512 ± 0.028\", \"0.756 ± 0.014\"], \"selected_cell_zero_based\": 3, \"selected_cell_xml\": \"<td align=\\\"center\\\" colspan=\\\"1\\\" rowspan=\\\"1\\\">0.695 ± 0.037</td>\", \"caption\": \"Statistical Performance Metrics of Potency Prediction for SARS-CoV-2 Mpro Using Different Ligand Pose Generation Protocols\"}","artifact_sha256":"c356a1c65a0033e5ae18a05d4afab5495856c5b6869328ff49e13547a4801a57","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12801289/fullTextXML"},"legacy_id":"lit-048","legacy_row":{"id":"lit-048","paper_id":"mpro-pose-affinity-2025","domain_id":"molecular-interactions","task":"Ligand potency prediction using generated poses","model":"DiffDock","model_version":"","dataset":"SARS-CoV-2 Mpro ligands","dataset_version":"","split":"","metric":"Pearson R","value":"0.695","unit":"unitless","uncertainty":"± 0.037","protocol":"Potency prediction using DiffDock ligand-pose generation plus paper scoring pipeline; not a native DiffDock affinity score.","source_locator":"Table 3, DiffDock row, Pearson’s R column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12801289/","evaluation_origin":"independent_paper","reviewed_utc":"2026-09-15T23:25:00Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"lit-b3-003","kind":"result","name":"Mouse-Geneformer · F1 · Human thymus scRNA-seq","description":"","status":"source_checked","facets":{"areas":["cells-tissues"],"tasks":["Human thymus cell-type classification"]},"source_ids":["mouse-geneformer-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b3-003"}],"attributes":{"printed_value":"48.57","numeric_value":"48.57","metric":"F1","metric_direction":"unknown","unit":"%","uncertainty":null,"source_locator":"Table 4, h/ Thymus row, Mouse-Geneformer Zero-shot F1 column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:44:03.392488+00:00","notes":"h/ Thymus row, four human cell types; zero-shot model blocks use Acc then F1, not fine-tuned scores. Ortholog-based conversion evaluated, not mouse cell annotation. Source check verifies central value and table context, not experiment reproduction or all metadata.","evidence":"{\"table_xml_id\": \"pgen.1011420.t004\", \"row_cells\": [\"h/ Thymus\", \"4\", \"82.74\", \"48.57\", \"95.59\", \"87.81\", \"87.27\", \"74.48\", \"95.44\", \"87.79\", \"92.04\", \"82.37\"], \"selected_cell_zero_based\": 3, \"selected_cell_xml\": \"<td align=\\\"left\\\" rowspan=\\\"1\\\" colspan=\\\"1\\\">48.57</td>\", \"caption\": \"Human cell type classification using mouse-Geneformer via ortholog-based gene name conversion, compared to native human models (human-Geneformer and scGPT).\"}","artifact_sha256":"ef6c68f5c9b47c2f89598ddf647f05b72e4155e8b33609ebff23936a84bbd41d","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11964219/fullTextXML"},"legacy_id":"lit-b3-003","legacy_row":{"id":"lit-b3-003","paper_id":"mouse-geneformer-2025","domain_id":"cells-tissues","task":"Human thymus cell-type classification","model":"Mouse-Geneformer","model_version":"","dataset":"Human thymus scRNA-seq","dataset_version":"","split":"","metric":"F1","value":"48.57","unit":"%","uncertainty":"","protocol":"Ortholog-based gene conversion; zero-shot mouse model on human cells.","source_locator":"Table 4, h/ Thymus row, Mouse-Geneformer Zero-shot F1 column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11964219/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:29:32Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-b3-004","kind":"result","name":"Human-Geneformer · F1 · Human thymus scRNA-seq","description":"","status":"source_checked","facets":{"areas":["cells-tissues"],"tasks":["Human thymus cell-type classification"]},"source_ids":["mouse-geneformer-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b3-004"}],"attributes":{"printed_value":"74.48","numeric_value":"74.48","metric":"F1","metric_direction":"unknown","unit":"%","uncertainty":null,"source_locator":"Table 4, h/ Thymus row, Human-Geneformer Zero-shot F1 column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:44:03.394073+00:00","notes":"h/ Thymus row, four human cell types; zero-shot model blocks use Acc then F1, not fine-tuned scores. Ortholog-based conversion evaluated, not mouse cell annotation. Source check verifies central value and table context, not experiment reproduction or all metadata.","evidence":"{\"table_xml_id\": \"pgen.1011420.t004\", \"row_cells\": [\"h/ Thymus\", \"4\", \"82.74\", \"48.57\", \"95.59\", \"87.81\", \"87.27\", \"74.48\", \"95.44\", \"87.79\", \"92.04\", \"82.37\"], \"selected_cell_zero_based\": 7, \"selected_cell_xml\": \"<td align=\\\"left\\\" rowspan=\\\"1\\\" colspan=\\\"1\\\">74.48</td>\", \"caption\": \"Human cell type classification using mouse-Geneformer via ortholog-based gene name conversion, compared to native human models (human-Geneformer and scGPT).\"}","artifact_sha256":"ef6c68f5c9b47c2f89598ddf647f05b72e4155e8b33609ebff23936a84bbd41d","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11964219/fullTextXML"},"legacy_id":"lit-b3-004","legacy_row":{"id":"lit-b3-004","paper_id":"mouse-geneformer-2025","domain_id":"cells-tissues","task":"Human thymus cell-type classification","model":"Human-Geneformer","model_version":"","dataset":"Human thymus scRNA-seq","dataset_version":"","split":"","metric":"F1","value":"74.48","unit":"%","uncertainty":"","protocol":"Native human model; zero-shot setting.","source_locator":"Table 4, h/ Thymus row, Human-Geneformer Zero-shot F1 column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11964219/","evaluation_origin":"independent_paper","reviewed_utc":"2026-09-15T23:29:32Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-b3-005","kind":"result","name":"scLLMDA · F1 · MosA1 reference → WholeBrainA query","description":"","status":"source_checked","facets":{"areas":["cells-tissues"],"tasks":["Cross-platform scATAC cell-type annotation"]},"source_ids":["scatac-llmda-2026"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b3-005"}],"attributes":{"printed_value":"0.6525","numeric_value":"0.6525","metric":"F1","metric_direction":"unknown","unit":"unitless","uncertainty":null,"source_locator":"Table 2, scLLMDA row, Ref: MosA1 / Q: WholeBrainA F1 column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:44:03.395850+00:00","notes":"First reference/query block MosA1 to WholeBrainA, F1 second column in block; direction of transfer is part of protocol. Source check verifies central value and table context, not experiment reproduction or all metadata.","evidence":"{\"table_xml_id\": \"pcbi.1014226.t002\", \"row_cells\": [\"scLLMDA\", \"0.7228\", \"0.6525\", \"0.7691\", \"0.3993\", \"0.7114\", \"0.6253\", \"0.7044\", \"0.3630\", \"0.7012\", \"0.6490\", \"0.7583\", \"0.4580\"], \"selected_cell_zero_based\": 2, \"selected_cell_xml\": \"<td align=\\\"left\\\" rowspan=\\\"1\\\" colspan=\\\"1\\\">\\n<bold>0.6525</bold>\\n</td>\", \"caption\": \"Cell type annotation between snATAC-seq and sciATAC-seq platforms.\"}","artifact_sha256":"f1cdc7d54c6b2d491e4a74a44a1188a3679c555262c988e95a1c0de4614fe3cb","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC13132462/fullTextXML"},"legacy_id":"lit-b3-005","legacy_row":{"id":"lit-b3-005","paper_id":"scatac-llmda-2026","domain_id":"cells-tissues","task":"Cross-platform scATAC cell-type annotation","model":"scLLMDA","model_version":"","dataset":"MosA1 reference → WholeBrainA query","dataset_version":"","split":"","metric":"F1","value":"0.6525","unit":"unitless","uncertainty":"","protocol":"Cross-platform reference-query cell-type annotation.","source_locator":"Table 2, scLLMDA row, Ref: MosA1 / Q: WholeBrainA F1 column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC13132462/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:29:32Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-b3-006","kind":"result","name":"MINGLE · F1 · MosA1 reference → WholeBrainA query","description":"","status":"source_checked","facets":{"areas":["cells-tissues"],"tasks":["Cross-platform scATAC cell-type annotation"]},"source_ids":["scatac-llmda-2026"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b3-006"}],"attributes":{"printed_value":"0.6256","numeric_value":"0.6256","metric":"F1","metric_direction":"unknown","unit":"unitless","uncertainty":null,"source_locator":"Table 2, MINGLE row, Ref: MosA1 / Q: WholeBrainA F1 column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:44:03.397272+00:00","notes":"First reference/query block MosA1 to WholeBrainA, F1 second column in block; direction of transfer is part of protocol. Source check verifies central value and table context, not experiment reproduction or all metadata.","evidence":"{\"table_xml_id\": \"pcbi.1014226.t002\", \"row_cells\": [\"MINGLE\", \"0.7143\", \"0.6256\", \"0.7095\", \"0.32\", \"0.7097\", \"0.6243\", \"0.6459\", \"0.2973\", \"0.6665\", \"0.6022\", \"0.7033\", \"0.3756\"], \"selected_cell_zero_based\": 2, \"selected_cell_xml\": \"<td align=\\\"left\\\" rowspan=\\\"1\\\" colspan=\\\"1\\\">0.6256</td>\", \"caption\": \"Cell type annotation between snATAC-seq and sciATAC-seq platforms.\"}","artifact_sha256":"f1cdc7d54c6b2d491e4a74a44a1188a3679c555262c988e95a1c0de4614fe3cb","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC13132462/fullTextXML"},"legacy_id":"lit-b3-006","legacy_row":{"id":"lit-b3-006","paper_id":"scatac-llmda-2026","domain_id":"cells-tissues","task":"Cross-platform scATAC cell-type annotation","model":"MINGLE","model_version":"","dataset":"MosA1 reference → WholeBrainA query","dataset_version":"","split":"","metric":"F1","value":"0.6256","unit":"unitless","uncertainty":"","protocol":"Cross-platform reference-query comparator.","source_locator":"Table 2, MINGLE row, Ref: MosA1 / Q: WholeBrainA F1 column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC13132462/","evaluation_origin":"independent_paper","reviewed_utc":"2026-09-15T23:29:32Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-b3-011","kind":"result","name":"GenePT-w · Adjusted Rand Index · Aorta single-cell dataset","description":"","status":"source_checked","facets":{"areas":["cells-tissues"],"tasks":["Cell-type structure in frozen embeddings"]},"source_ids":["genept-2024"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b3-011"}],"attributes":{"printed_value":"0.54","numeric_value":"0.54","metric":"Adjusted Rand Index","metric_direction":"unknown","unit":"unitless","uncertainty":null,"source_locator":"Table 2, Aorta / Cell type row, GenePT-w ARI column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:44:03.399274+00:00","notes":"First Cell type row belongs to Aorta, not preceding Phenotype row or later organs. ARI is first in each three-metric method block, not AMI/ASW. Source check verifies central value and table context, not experiment reproduction or all metadata.","evidence":"{\"table_xml_id\": \"T6\", \"row_cells\": [\"\", \"Cell type\", \"0.21\", \"0.31\", \"−0.04\", \"0.47\", \"0.64\", \"0.18\", \"0.54\", \"0.60\", \"0.03\", \"0.31\", \"0.47\", \"0.04\"], \"selected_cell_zero_based\": 8, \"selected_cell_xml\": \"<td align=\\\"center\\\" valign=\\\"top\\\" rowspan=\\\"1\\\" colspan=\\\"1\\\">\\n<bold>0.54</bold>\\n</td>\", \"caption\": \"Assessing the Association Between Different Latent Cell Representations and Biological Annotations.This analysis involves datasets representing cells from circulatory systems (Aorta and Artery), bone tissues (Bones, Myeloid), the Pancreas, and immune cells collected from healthy individuals and patients with Multiple Sclerosis. We utilized pretrained Geneformer and scGPT embeddings for this task. The Adjusted Rand Index (ARI) and Adjusted Mutual Information (AMI) were computed to compare the labels derived from k-means clustering with the true annotations of the original samples (higher values indicate better alignment); the Average Silhouette Width (ASW) was calculated using the true annotations of original samples to assess the cohesion and separation of the clusters.\"}","artifact_sha256":"230a2ec55458d9243eaeeebf3244df7409eb02d47f4b809ee56a06dcb6fdd047","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC10614824/fullTextXML"},"legacy_id":"lit-b3-011","legacy_row":{"id":"lit-b3-011","paper_id":"genept-2024","domain_id":"cells-tissues","task":"Cell-type structure in frozen embeddings","model":"GenePT-w","model_version":"","dataset":"Aorta single-cell dataset","dataset_version":"","split":"","metric":"Adjusted Rand Index","value":"0.54","unit":"unitless","uncertainty":"","protocol":"k-means on pretrained cell embeddings; agreement with original cell-type labels.","source_locator":"Table 2, Aorta / Cell type row, GenePT-w ARI column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC10614824/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:29:32Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-b3-012","kind":"result","name":"scGPT · Adjusted Rand Index · Aorta single-cell dataset","description":"","status":"source_checked","facets":{"areas":["cells-tissues"],"tasks":["Cell-type structure in frozen embeddings"]},"source_ids":["genept-2024"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b3-012"}],"attributes":{"printed_value":"0.47","numeric_value":"0.47","metric":"Adjusted Rand Index","metric_direction":"unknown","unit":"unitless","uncertainty":null,"source_locator":"Table 2, Aorta / Cell type row, scGPT ARI column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:44:03.402215+00:00","notes":"First Cell type row belongs to Aorta, not preceding Phenotype row or later organs. ARI is first in each three-metric method block, not AMI/ASW. Source check verifies central value and table context, not experiment reproduction or all metadata.","evidence":"{\"table_xml_id\": \"T6\", \"row_cells\": [\"\", \"Cell type\", \"0.21\", \"0.31\", \"−0.04\", \"0.47\", \"0.64\", \"0.18\", \"0.54\", \"0.60\", \"0.03\", \"0.31\", \"0.47\", \"0.04\"], \"selected_cell_zero_based\": 5, \"selected_cell_xml\": \"<td align=\\\"center\\\" valign=\\\"top\\\" rowspan=\\\"1\\\" colspan=\\\"1\\\">0.47</td>\", \"caption\": \"Assessing the Association Between Different Latent Cell Representations and Biological Annotations.This analysis involves datasets representing cells from circulatory systems (Aorta and Artery), bone tissues (Bones, Myeloid), the Pancreas, and immune cells collected from healthy individuals and patients with Multiple Sclerosis. We utilized pretrained Geneformer and scGPT embeddings for this task. The Adjusted Rand Index (ARI) and Adjusted Mutual Information (AMI) were computed to compare the labels derived from k-means clustering with the true annotations of the original samples (higher values indicate better alignment); the Average Silhouette Width (ASW) was calculated using the true annotations of original samples to assess the cohesion and separation of the clusters.\"}","artifact_sha256":"230a2ec55458d9243eaeeebf3244df7409eb02d47f4b809ee56a06dcb6fdd047","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC10614824/fullTextXML"},"legacy_id":"lit-b3-012","legacy_row":{"id":"lit-b3-012","paper_id":"genept-2024","domain_id":"cells-tissues","task":"Cell-type structure in frozen embeddings","model":"scGPT","model_version":"","dataset":"Aorta single-cell dataset","dataset_version":"","split":"","metric":"Adjusted Rand Index","value":"0.47","unit":"unitless","uncertainty":"","protocol":"k-means on pretrained cell embeddings; agreement with original cell-type labels.","source_locator":"Table 2, Aorta / Cell type row, scGPT ARI column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC10614824/","evaluation_origin":"independent_paper","reviewed_utc":"2026-09-15T23:29:32Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-b3-013","kind":"result","name":"Best frozen single-cell foundation model · Balanced accuracy · AIDA v2 PBMC cohort","description":"","status":"source_checked","facets":{"areas":["cells-tissues"],"tasks":["Donor-aware age-class prediction"]},"source_ids":["single-cell-aging-probes-2026"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b3-013"}],"attributes":{"printed_value":"0.322","numeric_value":"0.322","metric":"Balanced accuracy","metric_direction":"unknown","unit":"unitless","uncertainty":"± 0.008 standard deviation","source_locator":"Table 2, AIDA v2 row, scFM BA ± SD column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:44.421Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness.","evidence":"Table 2, AIDA v2 row, scFM BA ± SD column; cell: 0.322 ± 0.008","artifact_sha256":"d6cfb13933ceefed630954f804e7dc979b747bc4aec4c8c5518232f25772736a","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC13407579/fullTextXML"},"legacy_id":"lit-b3-013","legacy_row":{"id":"lit-b3-013","paper_id":"single-cell-aging-probes-2026","domain_id":"cells-tissues","task":"Donor-aware age-class prediction","model":"Best frozen single-cell foundation model","model_version":"","dataset":"AIDA v2 PBMC cohort","dataset_version":"622 donors","split":"","metric":"Balanced accuracy","value":"0.322","unit":"unitless","uncertainty":"± 0.008 standard deviation","protocol":"Same donor-aware splits and logistic-regression probe as expression PCA; text names Geneformer as best model on AIDA v2.","source_locator":"Table 2, AIDA v2 row, scFM BA ± SD column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC13407579/","evaluation_origin":"independent_paper","reviewed_utc":"2026-09-15T23:29:32Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"lit-b3-014","kind":"result","name":"Gene-expression PCA · Balanced accuracy · AIDA v2 PBMC cohort","description":"","status":"source_checked","facets":{"areas":["cells-tissues"],"tasks":["Donor-aware age-class prediction"]},"source_ids":["single-cell-aging-probes-2026"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b3-014"}],"attributes":{"printed_value":"0.384","numeric_value":"0.384","metric":"Balanced accuracy","metric_direction":"unknown","unit":"unitless","uncertainty":null,"source_locator":"Table 2, AIDA v2 row, Gene-expr BA column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:44.421Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness.","evidence":"Table 2, AIDA v2 row, Gene-expr BA column; cell: 0.384","artifact_sha256":"d6cfb13933ceefed630954f804e7dc979b747bc4aec4c8c5518232f25772736a","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC13407579/fullTextXML"},"legacy_id":"lit-b3-014","legacy_row":{"id":"lit-b3-014","paper_id":"single-cell-aging-probes-2026","domain_id":"cells-tissues","task":"Donor-aware age-class prediction","model":"Gene-expression PCA","model_version":"","dataset":"AIDA v2 PBMC cohort","dataset_version":"622 donors","split":"","metric":"Balanced accuracy","value":"0.384","unit":"unitless","uncertainty":"","protocol":"Fifty-component gene-expression PCA with the same donor-aware probe splits.","source_locator":"Table 2, AIDA v2 row, Gene-expr BA column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC13407579/","evaluation_origin":"independent_paper","reviewed_utc":"2026-09-15T23:29:32Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-b3-015","kind":"result","name":"scaLR · Cell-type accuracy · PBMCs-BS","description":"","status":"source_checked","facets":{"areas":["cells-tissues"],"tasks":["PBMC cell-type classification"]},"source_ids":["scalr-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b3-015"}],"attributes":{"printed_value":"0.942","numeric_value":"0.942","metric":"Cell-type accuracy","metric_direction":"unknown","unit":"unitless","uncertainty":null,"source_locator":"Table 2, scaLR row, Cell type Accuracy column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:44:03.403844+00:00","notes":"PBMCs-BS all-feature/all-sample cell-type accuracy block, not cell-state accuracy or time. Footnote letters on model labels excluded from identity. Source check verifies central value and table context, not experiment reproduction or all metadata.","evidence":"{\"table_xml_id\": \"TB2\", \"row_cells\": [\"scaLRa\", \"0.942\", \"27:50\", \"9.778\", \"0.840\", \"40:50\", \"8.679\"], \"selected_cell_zero_based\": 1, \"selected_cell_xml\": \"<td align=\\\"left\\\" rowspan=\\\"1\\\" colspan=\\\"1\\\">\\n<bold>0.942</bold>\\n</td>\", \"caption\": \"Accuracy, wall-clock time, and memory usage of different pipelines executed using all features and samples from the PBMCs-BS dataset.\"}","artifact_sha256":"829afab6a4e30997c608745d3eb280105b8c5c4e601020ffdc55c866144527ca","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12121358/fullTextXML"},"legacy_id":"lit-b3-015","legacy_row":{"id":"lit-b3-015","paper_id":"scalr-2025","domain_id":"cells-tissues","task":"PBMC cell-type classification","model":"scaLR","model_version":"","dataset":"PBMCs-BS","dataset_version":"","split":"","metric":"Cell-type accuracy","value":"0.942","unit":"unitless","uncertainty":"","protocol":"All features and samples from PBMCs-BS.","source_locator":"Table 2, scaLR row, Cell type Accuracy column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12121358/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:29:32Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-b3-016","kind":"result","name":"scVI + scANVI · Cell-type accuracy · PBMCs-BS","description":"","status":"source_checked","facets":{"areas":["cells-tissues"],"tasks":["PBMC cell-type classification"]},"source_ids":["scalr-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b3-016"}],"attributes":{"printed_value":"0.939","numeric_value":"0.939","metric":"Cell-type accuracy","metric_direction":"unknown","unit":"unitless","uncertainty":null,"source_locator":"Table 2, Svi-tools (scVI & scANVI) row, Cell type Accuracy column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:44:03.405089+00:00","notes":"PBMCs-BS all-feature/all-sample cell-type accuracy block, not cell-state accuracy or time. Footnote letters on model labels excluded from identity. Source check verifies central value and table context, not experiment reproduction or all metadata.","evidence":"{\"table_xml_id\": \"TB2\", \"row_cells\": [\"Svi-tools(scVI & scANVI)b\", \"0.939\", \"53:52\", \"23.914\", \"0.870\", \"54:24\", \"23.645\"], \"selected_cell_zero_based\": 1, \"selected_cell_xml\": \"<td align=\\\"left\\\" rowspan=\\\"1\\\" colspan=\\\"1\\\">0.939</td>\", \"caption\": \"Accuracy, wall-clock time, and memory usage of different pipelines executed using all features and samples from the PBMCs-BS dataset.\"}","artifact_sha256":"829afab6a4e30997c608745d3eb280105b8c5c4e601020ffdc55c866144527ca","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12121358/fullTextXML"},"legacy_id":"lit-b3-016","legacy_row":{"id":"lit-b3-016","paper_id":"scalr-2025","domain_id":"cells-tissues","task":"PBMC cell-type classification","model":"scVI + scANVI","model_version":"","dataset":"PBMCs-BS","dataset_version":"","split":"","metric":"Cell-type accuracy","value":"0.939","unit":"unitless","uncertainty":"","protocol":"All features and samples from PBMCs-BS; comparison pipeline combines scVI and scANVI.","source_locator":"Table 2, Svi-tools (scVI & scANVI) row, Cell type Accuracy column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12121358/","evaluation_origin":"independent_paper","reviewed_utc":"2026-09-15T23:29:32Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-b3-017","kind":"result","name":"scXDR · AUC · scXDR transfer scenario 2","description":"","status":"source_checked","facets":{"areas":["cells-tissues"],"tasks":["Cross-dataset single-cell drug response transfer"]},"source_ids":["scxdr-2026"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b3-017"}],"attributes":{"printed_value":"0.8248","numeric_value":"0.8248","metric":"AUC","metric_direction":"unknown","unit":"unitless","uncertainty":"± 0.1573 standard deviation","source_locator":"Table 2, scXDR row, Scenario 2 column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:50.056Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness.","evidence":"Table 2, scXDR row, Scenario 2 column; cell: 0.8248 0.1573 ±","artifact_sha256":"47b5925e9887d87fc8288d29288802b1d67d54f064d913151df92171f7c68d33","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12859067/fullTextXML"},"legacy_id":"lit-b3-017","legacy_row":{"id":"lit-b3-017","paper_id":"scxdr-2026","domain_id":"cells-tissues","task":"Cross-dataset single-cell drug response transfer","model":"scXDR","model_version":"","dataset":"scXDR transfer scenario 2","dataset_version":"","split":"","metric":"AUC","value":"0.8248","unit":"unitless","uncertainty":"± 0.1573 standard deviation","protocol":"Single-cell-to-single-cell transfer; source scenario 2.","source_locator":"Table 2, scXDR row, Scenario 2 column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12859067/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:29:32Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"lit-b3-018","kind":"result","name":"scVI · AUC · scXDR transfer scenario 2","description":"","status":"source_checked","facets":{"areas":["cells-tissues"],"tasks":["Cross-dataset single-cell drug response transfer"]},"source_ids":["scxdr-2026"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b3-018"}],"attributes":{"printed_value":"0.6970","numeric_value":"0.6970","metric":"AUC","metric_direction":"unknown","unit":"unitless","uncertainty":"± 0.2463 standard deviation","source_locator":"Table 2, scVI row, Scenario 2 column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:50.056Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness.","evidence":"Table 2, scVI row, Scenario 2 column; cell: 0.6970 ± 0.2463","artifact_sha256":"47b5925e9887d87fc8288d29288802b1d67d54f064d913151df92171f7c68d33","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12859067/fullTextXML"},"legacy_id":"lit-b3-018","legacy_row":{"id":"lit-b3-018","paper_id":"scxdr-2026","domain_id":"cells-tissues","task":"Cross-dataset single-cell drug response transfer","model":"scVI","model_version":"","dataset":"scXDR transfer scenario 2","dataset_version":"","split":"","metric":"AUC","value":"0.6970","unit":"unitless","uncertainty":"± 0.2463 standard deviation","protocol":"Single-cell-to-single-cell transfer; source scenario 2.","source_locator":"Table 2, scVI row, Scenario 2 column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12859067/","evaluation_origin":"independent_paper","reviewed_utc":"2026-09-15T23:29:32Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"lit-b3-019","kind":"result","name":"CAMMiQ · L1 abundance error · HumanGut-all strain-level query","description":"","status":"source_checked","facets":{"areas":["microbes-communities"],"tasks":["Strain-level abundance quantification"]},"source_ids":["cammiq-2022"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b3-019"}],"attributes":{"printed_value":"0.0517","numeric_value":"0.0517","metric":"L1 abundance error","metric_direction":"unknown","unit":"unitless","uncertainty":null,"source_locator":"Table 5, B. L1 Err. / HumanGut-all row, CAMMiQ L1 Err. column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:44:03.408237+00:00","notes":"HumanGut-all in B. L1 Err. block (second occurrence), not strain count or C. L2 Err. Numbers are errors; lower is better. Source check verifies central value and table context, not experiment reproduction or all metadata.","evidence":"{\"table_xml_id\": \"Tab5\", \"row_cells\": [\"\", \"HumanGut-all\", \"0.0517\", \"0.2841\", \"0.2426\", \"0.4004\", \"0.2811\", \"0.4439\"], \"selected_cell_zero_based\": 2, \"selected_cell_xml\": \"<td colspan=\\\"1\\\" rowspan=\\\"1\\\">0.0517</td>\", \"caption\": \"CAMMiQ’s strain level performance compared to Kraken2, KrakenUniq, CLARK, Centrifuge, and MetaPhlAn2, on the four strain-level queries\"}","artifact_sha256":"f0939647ed3de995d58254f79472a612c21b0e1b2560a82783302aa1a148dde3","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC9616933/fullTextXML"},"legacy_id":"lit-b3-019","legacy_row":{"id":"lit-b3-019","paper_id":"cammiq-2022","domain_id":"microbes-communities","task":"Strain-level abundance quantification","model":"CAMMiQ","model_version":"","dataset":"HumanGut-all strain-level query","dataset_version":"","split":"","metric":"L1 abundance error","value":"0.0517","unit":"unitless","uncertainty":"","protocol":"Strain-level quantification on the HumanGut-all synthetic query.","source_locator":"Table 5, B. L1 Err. / HumanGut-all row, CAMMiQ L1 Err. column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC9616933/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:29:32Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-b3-020","kind":"result","name":"Kraken2 · L1 abundance error · HumanGut-all strain-level query","description":"","status":"source_checked","facets":{"areas":["microbes-communities"],"tasks":["Strain-level abundance quantification"]},"source_ids":["cammiq-2022"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b3-020"}],"attributes":{"printed_value":"0.2841","numeric_value":"0.2841","metric":"L1 abundance error","metric_direction":"unknown","unit":"unitless","uncertainty":null,"source_locator":"Table 5, B. L1 Err. / HumanGut-all row, Kraken2 L1 Err. column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:44:03.411180+00:00","notes":"HumanGut-all in B. L1 Err. block (second occurrence), not strain count or C. L2 Err. Numbers are errors; lower is better. Source check verifies central value and table context, not experiment reproduction or all metadata.","evidence":"{\"table_xml_id\": \"Tab5\", \"row_cells\": [\"\", \"HumanGut-all\", \"0.0517\", \"0.2841\", \"0.2426\", \"0.4004\", \"0.2811\", \"0.4439\"], \"selected_cell_zero_based\": 3, \"selected_cell_xml\": \"<td colspan=\\\"1\\\" rowspan=\\\"1\\\">0.2841</td>\", \"caption\": \"CAMMiQ’s strain level performance compared to Kraken2, KrakenUniq, CLARK, Centrifuge, and MetaPhlAn2, on the four strain-level queries\"}","artifact_sha256":"f0939647ed3de995d58254f79472a612c21b0e1b2560a82783302aa1a148dde3","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC9616933/fullTextXML"},"legacy_id":"lit-b3-020","legacy_row":{"id":"lit-b3-020","paper_id":"cammiq-2022","domain_id":"microbes-communities","task":"Strain-level abundance quantification","model":"Kraken2","model_version":"","dataset":"HumanGut-all strain-level query","dataset_version":"","split":"","metric":"L1 abundance error","value":"0.2841","unit":"unitless","uncertainty":"","protocol":"Strain-level quantification on the HumanGut-all synthetic query.","source_locator":"Table 5, B. L1 Err. / HumanGut-all row, Kraken2 L1 Err. column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC9616933/","evaluation_origin":"independent_paper","reviewed_utc":"2026-09-15T23:29:32Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-b3-021","kind":"result","name":"Lazypipe-nt · Genus-level F1 · Simulated viral metagenome","description":"","status":"source_checked","facets":{"areas":["microbes-communities"],"tasks":["Simulated metagenome virus-taxon retrieval"]},"source_ids":["lazypipe-2020"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b3-021"}],"attributes":{"printed_value":"0.932","numeric_value":"0.932","metric":"Genus-level F1","metric_direction":"unknown","unit":"unitless","uncertainty":null,"source_locator":"Table 1, Lazypipe-nt / Genus row, F column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:44:03.412605+00:00","notes":"First Genus block selected using rank row span, not Species. Final F column is F score, not precision or recall. Source check verifies central value and table context, not experiment reproduction or all metadata.","evidence":"{\"table_xml_id\": \"veaa091-T1\", \"row_cells\": [\"Lazypipe-nt\", \"Genus\", \"41\", \"2\", \"4\", \"0.953\", \"0.911\", \"0.932\"], \"selected_cell_zero_based\": -1, \"selected_cell_xml\": \"<td rowspan=\\\"1\\\" colspan=\\\"1\\\">0.932</td>\", \"caption\": \"Accessing accuracy of virus taxon retrieval by different tools.\"}","artifact_sha256":"77842d8e4f6b419e331ab5a01fdf8f9eb8604f259425d79602be896aad3d0ad1","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC7772471/fullTextXML"},"legacy_id":"lit-b3-021","legacy_row":{"id":"lit-b3-021","paper_id":"lazypipe-2020","domain_id":"microbes-communities","task":"Simulated metagenome virus-taxon retrieval","model":"Lazypipe-nt","model_version":"","dataset":"Simulated viral metagenome","dataset_version":"","split":"","metric":"Genus-level F1","value":"0.932","unit":"unitless","uncertainty":"","protocol":"Genus-rank viral taxon retrieval.","source_locator":"Table 1, Lazypipe-nt / Genus row, F column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC7772471/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:29:32Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-b3-022","kind":"result","name":"Kraken2 · Genus-level F1 · Simulated viral metagenome","description":"","status":"source_checked","facets":{"areas":["microbes-communities"],"tasks":["Simulated metagenome virus-taxon retrieval"]},"source_ids":["lazypipe-2020"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b3-022"}],"attributes":{"printed_value":"0.627","numeric_value":"0.627","metric":"Genus-level F1","metric_direction":"unknown","unit":"unitless","uncertainty":null,"source_locator":"Table 1, Kraken2 / Genus row, F column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:44:03.413635+00:00","notes":"First Genus block selected using rank row span, not Species. Final F column is F score, not precision or recall. Source check verifies central value and table context, not experiment reproduction or all metadata.","evidence":"{\"table_xml_id\": \"veaa091-T1\", \"row_cells\": [\"Kraken2\", \"21\", \"1\", \"24\", \"0.955\", \"0.467\", \"0.627\"], \"selected_cell_zero_based\": -1, \"selected_cell_xml\": \"<td rowspan=\\\"1\\\" colspan=\\\"1\\\">0.627</td>\", \"caption\": \"Accessing accuracy of virus taxon retrieval by different tools.\"}","artifact_sha256":"77842d8e4f6b419e331ab5a01fdf8f9eb8604f259425d79602be896aad3d0ad1","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC7772471/fullTextXML"},"legacy_id":"lit-b3-022","legacy_row":{"id":"lit-b3-022","paper_id":"lazypipe-2020","domain_id":"microbes-communities","task":"Simulated metagenome virus-taxon retrieval","model":"Kraken2","model_version":"","dataset":"Simulated viral metagenome","dataset_version":"","split":"","metric":"Genus-level F1","value":"0.627","unit":"unitless","uncertainty":"","protocol":"Genus-rank viral taxon retrieval.","source_locator":"Table 1, Kraken2 / Genus row, F column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC7772471/","evaluation_origin":"independent_paper","reviewed_utc":"2026-09-15T23:29:32Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-b3-023","kind":"result","name":"NCD-gzip · Macro F1 · CAMI II Sample_0 10,000-read subsample","description":"","status":"source_checked","facets":{"areas":["microbes-communities"],"tasks":["CAMI II superkingdom read classification"]},"source_ids":["ncd-metagenomics-2026"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b3-023"}],"attributes":{"printed_value":"0.9804","numeric_value":"0.9804","metric":"Macro F1","metric_direction":"unknown","unit":"unitless","uncertainty":null,"source_locator":"Table 5, NCD Superkingdom row, F1 column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:44:03.414806+00:00","notes":"NCD rank-specific table 5, F1 column; Superkingdom and Phylum are different classification granularities. Source check verifies central value and table context, not experiment reproduction or all metadata.","evidence":"{\"table_xml_id\": \"table-5\", \"row_cells\": [\"Superkingdom\", \"0.9616\", \"1.0000\", \"0.9804\", \"0.9616\"], \"selected_cell_zero_based\": 3, \"selected_cell_xml\": \"<td rowspan=\\\"1\\\" colspan=\\\"1\\\">0.9804</td>\", \"caption\": \"Taxonomic classification on the CAMI II 10,000-read subsample (Sample_0).Metrics are macro-averaged (recall, precision, F1) and micro-averaged (accuracy). NCD uses genome fragmentation (‘Genome fragmentation’) and assigns every read; Kraken2 uses low-confidence assignments and leaves 61.4% unclassified.\"}","artifact_sha256":"10e9ba780c45e7787baff9b81ef7b45c014d7fe6c716d6c759e14d89a813dc1c","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12884959/fullTextXML"},"legacy_id":"lit-b3-023","legacy_row":{"id":"lit-b3-023","paper_id":"ncd-metagenomics-2026","domain_id":"microbes-communities","task":"CAMI II superkingdom read classification","model":"NCD-gzip","model_version":"","dataset":"CAMI II Sample_0 10,000-read subsample","dataset_version":"10,000 reads","split":"","metric":"Macro F1","value":"0.9804","unit":"unitless","uncertainty":"","protocol":"Superkingdom-level macro-averaged F1; NCD assigns every read.","source_locator":"Table 5, NCD Superkingdom row, F1 column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12884959/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:29:32Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-b3-024","kind":"result","name":"NCD-gzip · Macro F1 · CAMI II Sample_0 10,000-read subsample","description":"","status":"source_checked","facets":{"areas":["microbes-communities"],"tasks":["CAMI II phylum read classification"]},"source_ids":["ncd-metagenomics-2026"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b3-024"}],"attributes":{"printed_value":"0.1263","numeric_value":"0.1263","metric":"Macro F1","metric_direction":"unknown","unit":"unitless","uncertainty":null,"source_locator":"Table 5, NCD Phylum row, F1 column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:44:03.415788+00:00","notes":"NCD rank-specific table 5, F1 column; Superkingdom and Phylum are different classification granularities. Source check verifies central value and table context, not experiment reproduction or all metadata.","evidence":"{\"table_xml_id\": \"table-5\", \"row_cells\": [\"Phylum\", \"0.1336\", \"0.1637\", \"0.1263\", \"0.2709\"], \"selected_cell_zero_based\": 3, \"selected_cell_xml\": \"<td rowspan=\\\"1\\\" colspan=\\\"1\\\">0.1263</td>\", \"caption\": \"Taxonomic classification on the CAMI II 10,000-read subsample (Sample_0).Metrics are macro-averaged (recall, precision, F1) and micro-averaged (accuracy). NCD uses genome fragmentation (‘Genome fragmentation’) and assigns every read; Kraken2 uses low-confidence assignments and leaves 61.4% unclassified.\"}","artifact_sha256":"10e9ba780c45e7787baff9b81ef7b45c014d7fe6c716d6c759e14d89a813dc1c","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12884959/fullTextXML"},"legacy_id":"lit-b3-024","legacy_row":{"id":"lit-b3-024","paper_id":"ncd-metagenomics-2026","domain_id":"microbes-communities","task":"CAMI II phylum read classification","model":"NCD-gzip","model_version":"","dataset":"CAMI II Sample_0 10,000-read subsample","dataset_version":"10,000 reads","split":"","metric":"Macro F1","value":"0.1263","unit":"unitless","uncertainty":"","protocol":"Phylum-level macro-averaged F1; distinct taxonomic rank from the other row.","source_locator":"Table 5, NCD Phylum row, F1 column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12884959/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:29:32Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-b3-025","kind":"result","name":"VIBRANT · Average prophage F1 · 20 medium/high-complexity viral simulations","description":"","status":"source_checked","facets":{"areas":["microbes-communities"],"tasks":["Simulated prophage-contig detection"]},"source_ids":["viral-contig-simulation-2021"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b3-025"}],"attributes":{"printed_value":"0.169","numeric_value":"0.169","metric":"Average prophage F1","metric_direction":"unknown","unit":"unitless","uncertainty":null,"source_locator":"Table 3, Vibrant row, Prophage F1 column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:50.134Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness.","evidence":"Table 3, Vibrant row, Prophage F1 column; cell: 0.169","artifact_sha256":"93a24652edfa6d9f686862479df3addf50a2d5d432d2d5a31882075fa59dfdfd","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC8207588/fullTextXML"},"legacy_id":"lit-b3-025","legacy_row":{"id":"lit-b3-025","paper_id":"viral-contig-simulation-2021","domain_id":"microbes-communities","task":"Simulated prophage-contig detection","model":"VIBRANT","model_version":"","dataset":"20 medium/high-complexity viral simulations","dataset_version":"","split":"","metric":"Average prophage F1","value":"0.169","unit":"unitless","uncertainty":"","protocol":"Average across twenty medium- and high-complexity simulated communities.","source_locator":"Table 3, Vibrant row, Prophage F1 column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC8207588/","evaluation_origin":"independent_paper","reviewed_utc":"2026-09-15T23:29:32Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-b3-026","kind":"result","name":"VirSorter · Average prophage F1 · 20 medium/high-complexity viral simulations","description":"","status":"source_checked","facets":{"areas":["microbes-communities"],"tasks":["Simulated prophage-contig detection"]},"source_ids":["viral-contig-simulation-2021"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b3-026"}],"attributes":{"printed_value":"0.147","numeric_value":"0.147","metric":"Average prophage F1","metric_direction":"unknown","unit":"unitless","uncertainty":null,"source_locator":"Table 3, VirSorter row, Prophage F1 column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:50.134Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness.","evidence":"Table 3, VirSorter row, Prophage F1 column; cell: 0.147","artifact_sha256":"93a24652edfa6d9f686862479df3addf50a2d5d432d2d5a31882075fa59dfdfd","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC8207588/fullTextXML"},"legacy_id":"lit-b3-026","legacy_row":{"id":"lit-b3-026","paper_id":"viral-contig-simulation-2021","domain_id":"microbes-communities","task":"Simulated prophage-contig detection","model":"VirSorter","model_version":"","dataset":"20 medium/high-complexity viral simulations","dataset_version":"","split":"","metric":"Average prophage F1","value":"0.147","unit":"unitless","uncertainty":"","protocol":"Average across twenty medium- and high-complexity simulated communities.","source_locator":"Table 3, VirSorter row, Prophage F1 column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC8207588/","evaluation_origin":"independent_paper","reviewed_utc":"2026-09-15T23:29:32Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-b3-027","kind":"result","name":"GenomeOcean · F1 · GenomeOcean natural/artificial sequence test","description":"","status":"source_checked","facets":{"areas":["microbes-communities"],"tasks":["Natural vs artificial microbial genome sequence"]},"source_ids":["genomeocean-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b3-027"}],"attributes":{"printed_value":"99.03","numeric_value":"99.03","metric":"F1","metric_direction":"unknown","unit":"%","uncertainty":null,"source_locator":"Table 2, GenomeOcean row, F1 column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:55.224Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness.","evidence":"Table 2, GenomeOcean row, F1 column; cell: 99.03","artifact_sha256":"3cc0df52522fccda23e3958f069c916b87ee50bb5c9a992fa37e25256546e145","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11838515/fullTextXML"},"legacy_id":"lit-b3-027","legacy_row":{"id":"lit-b3-027","paper_id":"genomeocean-2025","domain_id":"microbes-communities","task":"Natural vs artificial microbial genome sequence","model":"GenomeOcean","model_version":"","dataset":"GenomeOcean natural/artificial sequence test","dataset_version":"","split":"","metric":"F1","value":"99.03","unit":"%","uncertainty":"","protocol":"Source reports natural-versus-artificial sequence classification.","source_locator":"Table 2, GenomeOcean row, F1 column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11838515/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:29:32Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-b3-028","kind":"result","name":"DNABERT-2 · F1 · GenomeOcean natural/artificial sequence test","description":"","status":"source_checked","facets":{"areas":["microbes-communities"],"tasks":["Natural vs artificial microbial genome sequence"]},"source_ids":["genomeocean-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b3-028"}],"attributes":{"printed_value":"85.12","numeric_value":"85.12","metric":"F1","metric_direction":"unknown","unit":"%","uncertainty":null,"source_locator":"Table 2, DNABERT-2 row, F1 column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:55.224Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness.","evidence":"Table 2, DNABERT-2 row, F1 column; cell: 85.12","artifact_sha256":"3cc0df52522fccda23e3958f069c916b87ee50bb5c9a992fa37e25256546e145","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11838515/fullTextXML"},"legacy_id":"lit-b3-028","legacy_row":{"id":"lit-b3-028","paper_id":"genomeocean-2025","domain_id":"microbes-communities","task":"Natural vs artificial microbial genome sequence","model":"DNABERT-2","model_version":"","dataset":"GenomeOcean natural/artificial sequence test","dataset_version":"","split":"","metric":"F1","value":"85.12","unit":"%","uncertainty":"","protocol":"Source reports natural-versus-artificial sequence classification.","source_locator":"Table 2, DNABERT-2 row, F1 column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11838515/","evaluation_origin":"independent_paper","reviewed_utc":"2026-09-15T23:29:32Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-b3-029","kind":"result","name":"kMetaShot · Genus-level F1 · Real mock community MAGs","description":"","status":"source_checked","facets":{"areas":["microbes-communities"],"tasks":["Mock-community MAG taxonomy classification"]},"source_ids":["kmetashot-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b3-029"}],"attributes":{"printed_value":"95.83","numeric_value":"95.83","metric":"Genus-level F1","metric_direction":"unknown","unit":"%","uncertainty":null,"source_locator":"Table 2, F1-score % row, Genus kMS column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:44:03.417367+00:00","notes":"Real mock sequencing table, F1-score percentage row; Genus is final three-column block, selecting kMS or Gtk rather than Species/Strain. Source check verifies central value and table context, not experiment reproduction or all metadata.","evidence":"{\"table_xml_id\": \"TB2\", \"row_cells\": [\"F1-score %\", \"28.13\", \"85.42\", \"86.24\", \"87.18\", \"87.72\", \"95.83\", \"89.80\", \"94.85\"], \"selected_cell_zero_based\": 6, \"selected_cell_xml\": \"<td align=\\\"left\\\" rowspan=\\\"1\\\" colspan=\\\"1\\\">95.83</td>\", \"caption\": \"Classification performance metrics measured on the real mock sequencing data.\"}","artifact_sha256":"4584e93ea035c1170b8756a0a52cbe99fe72e70bd09b5f1dee639ee104f78247","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11695915/fullTextXML"},"legacy_id":"lit-b3-029","legacy_row":{"id":"lit-b3-029","paper_id":"kmetashot-2025","domain_id":"microbes-communities","task":"Mock-community MAG taxonomy classification","model":"kMetaShot","model_version":"","dataset":"Real mock community MAGs","dataset_version":"","split":"","metric":"Genus-level F1","value":"95.83","unit":"%","uncertainty":"","protocol":"Genus classification of MAGs from MegaHIT contigs; uncorrected kMetaShot.","source_locator":"Table 2, F1-score % row, Genus kMS column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11695915/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:29:32Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-b3-030","kind":"result","name":"GTDB-Tk · Genus-level F1 · Real mock community MAGs","description":"","status":"source_checked","facets":{"areas":["microbes-communities"],"tasks":["Mock-community MAG taxonomy classification"]},"source_ids":["kmetashot-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b3-030"}],"attributes":{"printed_value":"89.80","numeric_value":"89.80","metric":"Genus-level F1","metric_direction":"unknown","unit":"%","uncertainty":null,"source_locator":"Table 2, F1-score % row, Genus Gtk column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:44:03.418914+00:00","notes":"Real mock sequencing table, F1-score percentage row; Genus is final three-column block, selecting kMS or Gtk rather than Species/Strain. Source check verifies central value and table context, not experiment reproduction or all metadata.","evidence":"{\"table_xml_id\": \"TB2\", \"row_cells\": [\"F1-score %\", \"28.13\", \"85.42\", \"86.24\", \"87.18\", \"87.72\", \"95.83\", \"89.80\", \"94.85\"], \"selected_cell_zero_based\": 7, \"selected_cell_xml\": \"<td align=\\\"left\\\" rowspan=\\\"1\\\" colspan=\\\"1\\\">89.80</td>\", \"caption\": \"Classification performance metrics measured on the real mock sequencing data.\"}","artifact_sha256":"4584e93ea035c1170b8756a0a52cbe99fe72e70bd09b5f1dee639ee104f78247","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11695915/fullTextXML"},"legacy_id":"lit-b3-030","legacy_row":{"id":"lit-b3-030","paper_id":"kmetashot-2025","domain_id":"microbes-communities","task":"Mock-community MAG taxonomy classification","model":"GTDB-Tk","model_version":"","dataset":"Real mock community MAGs","dataset_version":"","split":"","metric":"Genus-level F1","value":"89.80","unit":"%","uncertainty":"","protocol":"Genus classification of the same MAG set.","source_locator":"Table 2, F1-score % row, Genus Gtk column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11695915/","evaluation_origin":"independent_paper","reviewed_utc":"2026-09-15T23:29:32Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-b3-031","kind":"result","name":"Lemur · F1 · Zymo LOG 10%","description":"","status":"source_checked","facets":{"areas":["microbes-communities"],"tasks":["Long-read taxonomic profiling"]},"source_ids":["lemur-magnet-2024"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b3-031"}],"attributes":{"printed_value":"0.376","numeric_value":"0.376","metric":"F1","metric_direction":"unknown","unit":"unitless","uncertainty":null,"source_locator":"Table 3, LOG 10% / Lemur row, F1 column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:44:03.420493+00:00","notes":"LOG 10% first block, F1 column. Kraken 2 row inherits dataset via rowspan; not LOG 75% or abundance Spearman. Source check verifies central value and table context, not experiment reproduction or all metadata.","evidence":"{\"table_xml_id\": \"T3\", \"row_cells\": [\"LOG 10%\", \"Lemur\", \"0.500\", \"0.301\", \"0.376\", \"0.984\"], \"selected_cell_zero_based\": 4, \"selected_cell_xml\": \"<td align=\\\"right\\\" valign=\\\"top\\\" rowspan=\\\"1\\\" colspan=\\\"1\\\">0.376</td>\", \"caption\": \"Mean performance and standard deviation across 5 replicate runs of all methods on Zymo LOG, bold values show best performance. Magnet does not report relative abundance, so the Spearman’s ρ cannot be computed. Tools listed below the horizontal dashed lines (for LOG 10% and LOG 75%) focus on the taxonomic classification of reads.\"}","artifact_sha256":"4afb9195da447916eb6f733816e3640741c7ade08ea8920d205c3be7b3cce27a","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11185576/fullTextXML"},"legacy_id":"lit-b3-031","legacy_row":{"id":"lit-b3-031","paper_id":"lemur-magnet-2024","domain_id":"microbes-communities","task":"Long-read taxonomic profiling","model":"Lemur","model_version":"","dataset":"Zymo LOG 10%","dataset_version":"","split":"","metric":"F1","value":"0.376","unit":"unitless","uncertainty":"","protocol":"Mean across five replicate runs on Zymo LOG 10%.","source_locator":"Table 3, LOG 10% / Lemur row, F1 column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11185576/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:29:32Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-b3-032","kind":"result","name":"Kraken 2 · F1 · Zymo LOG 10%","description":"","status":"source_checked","facets":{"areas":["microbes-communities"],"tasks":["Long-read taxonomic profiling"]},"source_ids":["lemur-magnet-2024"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b3-032"}],"attributes":{"printed_value":"0.375","numeric_value":"0.375","metric":"F1","metric_direction":"unknown","unit":"unitless","uncertainty":null,"source_locator":"Table 3, LOG 10% / Kraken 2 row, F1 column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:44:03.421888+00:00","notes":"LOG 10% first block, F1 column. Kraken 2 row inherits dataset via rowspan; not LOG 75% or abundance Spearman. Source check verifies central value and table context, not experiment reproduction or all metadata.","evidence":"{\"table_xml_id\": \"T3\", \"row_cells\": [\"Kraken 2\", \"0.760\", \"0.249\", \"0.375\", \"0.910\"], \"selected_cell_zero_based\": 3, \"selected_cell_xml\": \"<td align=\\\"right\\\" valign=\\\"top\\\" rowspan=\\\"1\\\" colspan=\\\"1\\\">0.375</td>\", \"caption\": \"Mean performance and standard deviation across 5 replicate runs of all methods on Zymo LOG, bold values show best performance. Magnet does not report relative abundance, so the Spearman’s ρ cannot be computed. Tools listed below the horizontal dashed lines (for LOG 10% and LOG 75%) focus on the taxonomic classification of reads.\"}","artifact_sha256":"4afb9195da447916eb6f733816e3640741c7ade08ea8920d205c3be7b3cce27a","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11185576/fullTextXML"},"legacy_id":"lit-b3-032","legacy_row":{"id":"lit-b3-032","paper_id":"lemur-magnet-2024","domain_id":"microbes-communities","task":"Long-read taxonomic profiling","model":"Kraken 2","model_version":"","dataset":"Zymo LOG 10%","dataset_version":"","split":"","metric":"F1","value":"0.375","unit":"unitless","uncertainty":"","protocol":"Mean across five replicate runs on Zymo LOG 10%.","source_locator":"Table 3, LOG 10% / Kraken 2 row, F1 column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11185576/","evaluation_origin":"independent_paper","reviewed_utc":"2026-09-15T23:29:32Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-b3-033","kind":"result","name":"iPro-MP · Mean AUC · 23 independent prokaryotic promoter test sets","description":"","status":"source_checked","facets":{"areas":["microbes-communities"],"tasks":["Multi-species prokaryotic promoter detection"]},"source_ids":["ipromp-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b3-033"}],"attributes":{"printed_value":"0.935","numeric_value":"0.935","metric":"Mean AUC","metric_direction":"unknown","unit":"unitless","uncertainty":null,"source_locator":"Table 2, iPro-MP row, AUC column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:55.361Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness.","evidence":"Table 2, iPro-MP row, AUC column; cell: 0.935","artifact_sha256":"d21541ee1f7a168da8e4a7c0f0e133c970cbe7bc41118f43a929f08b2fd2afd1","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12516880/fullTextXML"},"legacy_id":"lit-b3-033","legacy_row":{"id":"lit-b3-033","paper_id":"ipromp-2025","domain_id":"microbes-communities","task":"Multi-species prokaryotic promoter detection","model":"iPro-MP","model_version":"","dataset":"23 independent prokaryotic promoter test sets","dataset_version":"23 test sets","split":"independent test","metric":"Mean AUC","value":"0.935","unit":"unitless","uncertainty":"","protocol":"Average over independent testing sets.","source_locator":"Table 2, iPro-MP row, AUC column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12516880/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:29:32Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-b3-034","kind":"result","name":"Prompt · Mean AUC · 23 independent prokaryotic promoter test sets","description":"","status":"source_checked","facets":{"areas":["microbes-communities"],"tasks":["Multi-species prokaryotic promoter detection"]},"source_ids":["ipromp-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b3-034"}],"attributes":{"printed_value":"0.835","numeric_value":"0.835","metric":"Mean AUC","metric_direction":"unknown","unit":"unitless","uncertainty":null,"source_locator":"Table 2, Prompt row, AUC column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:55.361Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness.","evidence":"Table 2, Prompt row, AUC column; cell: 0.835","artifact_sha256":"d21541ee1f7a168da8e4a7c0f0e133c970cbe7bc41118f43a929f08b2fd2afd1","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12516880/fullTextXML"},"legacy_id":"lit-b3-034","legacy_row":{"id":"lit-b3-034","paper_id":"ipromp-2025","domain_id":"microbes-communities","task":"Multi-species prokaryotic promoter detection","model":"Prompt","model_version":"","dataset":"23 independent prokaryotic promoter test sets","dataset_version":"23 test sets","split":"independent test","metric":"Mean AUC","value":"0.835","unit":"unitless","uncertainty":"","protocol":"Average over the same independent testing sets.","source_locator":"Table 2, Prompt row, AUC column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12516880/","evaluation_origin":"independent_paper","reviewed_utc":"2026-09-15T23:29:32Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-b3-035","kind":"result","name":"ICCTax · Genus macro AveP · ICCTax Complete dataset","description":"","status":"source_checked","facets":{"areas":["microbes-communities"],"tasks":["Hierarchical metagenomic taxonomy classification"]},"source_ids":["icctax-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b3-035"}],"attributes":{"printed_value":"67.20","numeric_value":"67.20","metric":"Genus macro AveP","metric_direction":"unknown","unit":"%","uncertainty":null,"source_locator":"Table 2, ICCTax row, Genus column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:55.373Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness.","evidence":"Table 2, ICCTax row, Genus column; cell: 67.20","artifact_sha256":"2ce0b48f1cde3aea7e561d92f4d7dc1525af7439ccd16f80bec0773e8812c8ec","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12619997/fullTextXML"},"legacy_id":"lit-b3-035","legacy_row":{"id":"lit-b3-035","paper_id":"icctax-2025","domain_id":"microbes-communities","task":"Hierarchical metagenomic taxonomy classification","model":"ICCTax","model_version":"","dataset":"ICCTax Complete dataset","dataset_version":"","split":"","metric":"Genus macro AveP","value":"67.20","unit":"%","uncertainty":"","protocol":"Macro average precision at genus rank on Complete dataset.","source_locator":"Table 2, ICCTax row, Genus column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12619997/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:29:32Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-b3-036","kind":"result","name":"Kraken2 · Genus macro AveP · ICCTax Complete dataset","description":"","status":"source_checked","facets":{"areas":["microbes-communities"],"tasks":["Hierarchical metagenomic taxonomy classification"]},"source_ids":["icctax-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b3-036"}],"attributes":{"printed_value":"70.56","numeric_value":"70.56","metric":"Genus macro AveP","metric_direction":"unknown","unit":"%","uncertainty":null,"source_locator":"Table 2, Kraken2 row, Genus column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:55.373Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness.","evidence":"Table 2, Kraken2 row, Genus column; cell: 70.56","artifact_sha256":"2ce0b48f1cde3aea7e561d92f4d7dc1525af7439ccd16f80bec0773e8812c8ec","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12619997/fullTextXML"},"legacy_id":"lit-b3-036","legacy_row":{"id":"lit-b3-036","paper_id":"icctax-2025","domain_id":"microbes-communities","task":"Hierarchical metagenomic taxonomy classification","model":"Kraken2","model_version":"","dataset":"ICCTax Complete dataset","dataset_version":"","split":"","metric":"Genus macro AveP","value":"70.56","unit":"%","uncertainty":"","protocol":"Macro average precision at genus rank on Complete dataset.","source_locator":"Table 2, Kraken2 row, Genus column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12619997/","evaluation_origin":"independent_paper","reviewed_utc":"2026-09-15T23:29:32Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-b3-037","kind":"result","name":"Chai-1 · AUC-ROC · Antibody–antigen GEP test set","description":"","status":"source_checked","facets":{"areas":["molecular-interactions"],"tasks":["Antibody–antigen interaction prediction using folded complexes"]},"source_ids":["antibody-flexibility-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b3-037"}],"attributes":{"printed_value":"0.86","numeric_value":"0.86","metric":"AUC-ROC","metric_direction":"unknown","unit":"unitless","uncertainty":"± 0.07","source_locator":"Table 5, Folded row, Chai-1 (no MSA) column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:55.400Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness.","evidence":"Table 5, Folded row, Chai-1 (no MSA) column; cell: 0.86 ± 0.07","artifact_sha256":"57e64694c69052ed0495570e12ebfb4bb6c0ad152219f23827cd4b1cb53450ef","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12530544/fullTextXML"},"legacy_id":"lit-b3-037","legacy_row":{"id":"lit-b3-037","paper_id":"antibody-flexibility-2025","domain_id":"molecular-interactions","task":"Antibody–antigen interaction prediction using folded complexes","model":"Chai-1","model_version":"","dataset":"Antibody–antigen GEP test set","dataset_version":"","split":"","metric":"AUC-ROC","value":"0.86","unit":"unitless","uncertainty":"± 0.07","protocol":"Interaction classifier evaluated using Chai-1-folded input complexes; this is pipeline AUC, not DockQ.","source_locator":"Table 5, Folded row, Chai-1 (no MSA) column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12530544/","evaluation_origin":"independent_paper","reviewed_utc":"2026-09-15T23:29:32Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"lit-b3-038","kind":"result","name":"Boltz-1 · AUC-ROC · Antibody–antigen GEP test set","description":"","status":"source_checked","facets":{"areas":["molecular-interactions"],"tasks":["Antibody–antigen interaction prediction using folded complexes"]},"source_ids":["antibody-flexibility-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b3-038"}],"attributes":{"printed_value":"0.85","numeric_value":"0.85","metric":"AUC-ROC","metric_direction":"unknown","unit":"unitless","uncertainty":"± 0.05","source_locator":"Table 5, Folded row, Boltz-1 (no MSA) column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:55.400Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness.","evidence":"Table 5, Folded row, Boltz-1 (no MSA) column; cell: 0.85 ± 0.05","artifact_sha256":"57e64694c69052ed0495570e12ebfb4bb6c0ad152219f23827cd4b1cb53450ef","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12530544/fullTextXML"},"legacy_id":"lit-b3-038","legacy_row":{"id":"lit-b3-038","paper_id":"antibody-flexibility-2025","domain_id":"molecular-interactions","task":"Antibody–antigen interaction prediction using folded complexes","model":"Boltz-1","model_version":"","dataset":"Antibody–antigen GEP test set","dataset_version":"","split":"","metric":"AUC-ROC","value":"0.85","unit":"unitless","uncertainty":"± 0.05","protocol":"Interaction classifier evaluated using Boltz-1-folded input complexes; this is pipeline AUC, not DockQ.","source_locator":"Table 5, Folded row, Boltz-1 (no MSA) column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12530544/","evaluation_origin":"independent_paper","reviewed_utc":"2026-09-15T23:29:32Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"lit-b3-039","kind":"result","name":"Boltz-1 · Top-1 ligand RMSD <2 Å rate · Boltz-1 structure test set","description":"","status":"source_checked","facets":{"areas":["molecular-interactions"],"tasks":["Protein–ligand pose prediction"]},"source_ids":["boltz1-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b3-039"}],"attributes":{"printed_value":"0.545","numeric_value":"0.545","metric":"Top-1 ligand RMSD <2 Å rate","metric_direction":"unknown","unit":"unitless","uncertainty":null,"source_locator":"Table 1, 3 recycling rounds / 200 steps row, L-RMSD <2Å top-1 column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:44:03.424864+00:00","notes":"3 recycling rounds and 200 steps; L-RMSD <2 Angstrom top-1 (last column), not oracle. Five samples generated; top-1 means highest-confidence candidate. Repeated reference rows are one evaluation, not independent experiments. Source check verifies central value and table context, not experiment reproduction or all metadata.","evidence":"{\"table_xml_id\": \"T4\", \"row_cells\": [\"3\", \"200\", \"0.729\", \"0.716\", \"0.654\", \"0.625\", \"0.621\", \"0.580\", \"0.581\", \"0.545\"], \"selected_cell_zero_based\": 9, \"selected_cell_xml\": \"<td align=\\\"center\\\" valign=\\\"top\\\" style=\\\"border-bottom: solid 1px\\\" rowspan=\\\"1\\\" colspan=\\\"1\\\">0.545</td>\", \"caption\": \"Ablation on the number of recycling rounds and sampling steps for Boltz-1 on the test set. We run the ablation study generating 5 samples and evaluating both the best (oracle) and highest confidence prediction (top-1) out of the 5 for every metric. All models used pre-computed MSAs with up to 4,096 sequences. It is worth noting that the metrics are noisy, so minor inconsistencies (e.g., lack of improvement with increased recycling rounds or diffusion steps) should not be overinterpreted. Moreover, there is a slight difference with the results in Figures 5 and 7 due to differences in MSA parameters as well as the set of structures passing all ablations.\"}","artifact_sha256":"1ebf712314d9a1c678ded989cc95a0c00c0331e5ad8c9f63194bc9780971d214","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11601547/fullTextXML"},"legacy_id":"lit-b3-039","legacy_row":{"id":"lit-b3-039","paper_id":"boltz1-2025","domain_id":"molecular-interactions","task":"Protein–ligand pose prediction","model":"Boltz-1","model_version":"3 recycling rounds; 200 diffusion steps","dataset":"Boltz-1 structure test set","dataset_version":"","split":"","metric":"Top-1 ligand RMSD <2 Å rate","value":"0.545","unit":"unitless","uncertainty":"","protocol":"Highest-confidence pose from five samples; precomputed MSAs up to 4,096 sequences.","source_locator":"Table 1, 3 recycling rounds / 200 steps row, L-RMSD <2Å top-1 column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11601547/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:29:32Z"},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-b3-040","kind":"result","name":"Ibex · Mean CDR H3 RMSD · ImmuneBuilder antibody test set","description":"","status":"source_checked","facets":{"areas":["molecular-interactions"],"tasks":["Antibody loop structure prediction"]},"source_ids":["ibex-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b3-040"}],"attributes":{"printed_value":"2.72","numeric_value":"2.72","metric":"Mean CDR H3 RMSD","metric_direction":"unknown","unit":"Å","uncertainty":null,"source_locator":"Table 1, Antibodies / Ibex row, CDR H3 column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:44:03.426811+00:00","notes":"First Antibodies block, CDR H3 mean RMSD in Angstrom; excludes later Nanobodies and TCR blocks. Source check verifies central value and table context, not experiment reproduction or all metadata.","evidence":"{\"table_xml_id\": \"t0001\", \"row_cells\": [\"\", \"Ibex\", \"0.61\", \"0.57\", \"2.72\", \"0.45\", \"0.57\", \"0.43\", \"0.98\", \"0.52\"], \"selected_cell_zero_based\": 4, \"selected_cell_xml\": \"<td align=\\\"center\\\" colspan=\\\"1\\\" rowspan=\\\"1\\\"><italic toggle=\\\"yes\\\">2.72</italic></td>\", \"caption\": \"Average RMSD in angstrom, evaluated separately for each region on the ImmuneBuilder test set of antibodies, nanobodies and TCRs.\"}","artifact_sha256":"caa1109bd5fe7f6be703aa9d4afd6f4f1522bcbce6b7361650eb59618c2a9e14","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12710905/fullTextXML"},"legacy_id":"lit-b3-040","legacy_row":{"id":"lit-b3-040","paper_id":"ibex-2025","domain_id":"molecular-interactions","task":"Antibody loop structure prediction","model":"Ibex","model_version":"","dataset":"ImmuneBuilder antibody test set","dataset_version":"","split":"","metric":"Mean CDR H3 RMSD","value":"2.72","unit":"Å","uncertainty":"","protocol":"Backbone RMSD after framework alignment; average over antibody test structures.","source_locator":"Table 1, Antibodies / Ibex row, CDR H3 column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12710905/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:29:32Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-b3-041","kind":"result","name":"Chai-1 · Mean CDR H3 RMSD · ImmuneBuilder antibody test set","description":"","status":"source_checked","facets":{"areas":["molecular-interactions"],"tasks":["Antibody loop structure prediction"]},"source_ids":["ibex-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b3-041"}],"attributes":{"printed_value":"2.65","numeric_value":"2.65","metric":"Mean CDR H3 RMSD","metric_direction":"unknown","unit":"Å","uncertainty":null,"source_locator":"Table 1, Antibodies / Chai-1 row, CDR H3 column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:44:03.428536+00:00","notes":"First Antibodies block, CDR H3 mean RMSD in Angstrom; excludes later Nanobodies and TCR blocks. Source check verifies central value and table context, not experiment reproduction or all metadata.","evidence":"{\"table_xml_id\": \"t0001\", \"row_cells\": [\"\", \"Chai-1\", \"0.67\", \"0.53\", \"2.65\", \"0.45\", \"0.53\", \"0.41\", \"1.20\", \"0.54\"], \"selected_cell_zero_based\": 4, \"selected_cell_xml\": \"<td align=\\\"center\\\" colspan=\\\"1\\\" rowspan=\\\"1\\\"><bold>2.65</bold></td>\", \"caption\": \"Average RMSD in angstrom, evaluated separately for each region on the ImmuneBuilder test set of antibodies, nanobodies and TCRs.\"}","artifact_sha256":"caa1109bd5fe7f6be703aa9d4afd6f4f1522bcbce6b7361650eb59618c2a9e14","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12710905/fullTextXML"},"legacy_id":"lit-b3-041","legacy_row":{"id":"lit-b3-041","paper_id":"ibex-2025","domain_id":"molecular-interactions","task":"Antibody loop structure prediction","model":"Chai-1","model_version":"","dataset":"ImmuneBuilder antibody test set","dataset_version":"","split":"","metric":"Mean CDR H3 RMSD","value":"2.65","unit":"Å","uncertainty":"","protocol":"Backbone RMSD after framework alignment; one seed and one diffusion trajectory.","source_locator":"Table 1, Antibodies / Chai-1 row, CDR H3 column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12710905/","evaluation_origin":"independent_paper","reviewed_utc":"2026-09-15T23:29:32Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-b3-042","kind":"result","name":"DEELIG · Pearson R · PDBbind core v2016","description":"","status":"source_checked","facets":{"areas":["molecular-interactions"],"tasks":["Protein–ligand binding affinity prediction"]},"source_ids":["deelig-2021"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b3-042"}],"attributes":{"printed_value":"0.889","numeric_value":"0.889","metric":"Pearson R","metric_direction":"unknown","unit":"unitless","uncertainty":null,"source_locator":"Table 2, DEELIG row, PDBbind v2016 column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:55.586Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness.","evidence":"Table 2, DEELIG row, PDBbind v2016 column; cell: 0.889","artifact_sha256":"5a7620c18d0622561004e1e25b5cfaf7399e93df3547eeefdd4cf6d300bb8aba","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC8274096/fullTextXML"},"legacy_id":"lit-b3-042","legacy_row":{"id":"lit-b3-042","paper_id":"deelig-2021","domain_id":"molecular-interactions","task":"Protein–ligand binding affinity prediction","model":"DEELIG","model_version":"","dataset":"PDBbind core v2016","dataset_version":"v2016","split":"","metric":"Pearson R","value":"0.889","unit":"unitless","uncertainty":"","protocol":"Source paper reports DEELIG on PDBbind core set.","source_locator":"Table 2, DEELIG row, PDBbind v2016 column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC8274096/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:29:32Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-b3-043","kind":"result","name":"TOPBP (Complex) · Pearson R · PDBbind core v2016","description":"","status":"source_checked","facets":{"areas":["molecular-interactions"],"tasks":["Protein–ligand binding affinity prediction"]},"source_ids":["deelig-2021"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b3-043"}],"attributes":{"printed_value":"0.861","numeric_value":"0.861","metric":"Pearson R","metric_direction":"unknown","unit":"unitless","uncertainty":null,"source_locator":"Table 2, TOPBP (Complex) row, PDBbind v2016 column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:44:03.429669+00:00","notes":"TOPBP Complex reference row; PDBbind v2016 core-set Pearson correlation. Third-party comparator with cited reference; do not infer an independent new run from table inclusion. Source check verifies central value and table context, not experiment reproduction or all metadata.","evidence":"{\"table_xml_id\": \"table2-11779322211030364\", \"row_cells\": [\"TOPBP (Complex) 31\", \"0.808\", \"0.861\"], \"selected_cell_zero_based\": 2, \"selected_cell_xml\": \"<td rowspan=\\\"1\\\" colspan=\\\"1\\\">0.861</td>\", \"caption\": \"Pearson correlations coefficient on PDBbind core set.\"}","artifact_sha256":"5a7620c18d0622561004e1e25b5cfaf7399e93df3547eeefdd4cf6d300bb8aba","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC8274096/fullTextXML"},"legacy_id":"lit-b3-043","legacy_row":{"id":"lit-b3-043","paper_id":"deelig-2021","domain_id":"molecular-interactions","task":"Protein–ligand binding affinity prediction","model":"TOPBP (Complex)","model_version":"","dataset":"PDBbind core v2016","dataset_version":"v2016","split":"","metric":"Pearson R","value":"0.861","unit":"unitless","uncertainty":"","protocol":"Source table compiles a previously published comparator; protocol equivalence is not established.","source_locator":"Table 2, TOPBP (Complex) row, PDBbind v2016 column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC8274096/","evaluation_origin":"paper_compilation","reviewed_utc":"2026-09-15T23:29:32Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-b3-044","kind":"result","name":"MolAS · RMSD ≤1 Å and PB-valid success · PoseBusters","description":"","status":"source_checked","facets":{"areas":["molecular-interactions"],"tasks":["Physically valid protein–ligand pose selection"]},"source_ids":["molas-2026"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b3-044"}],"attributes":{"printed_value":"36.69","numeric_value":"36.69","metric":"RMSD ≤1 Å and PB-valid success","metric_direction":"unknown","unit":"%","uncertainty":null,"source_locator":"Table 3, PoseBusters / Mixed / AutoDock row, MolAS success column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:44:03.432565+00:00","notes":"PoseBusters, Mixed, AutoDock row within jointly trained with/without relaxation block. Selected RMSD <=1 Angstrom AND PB-valid group; five-fold average success percentage, not <=2 Angstrom. Inline bold digit nodes joined in original order. Source check verifies central value and table context, not experiment reproduction or all metadata.","evidence":"{\"table_xml_id\": \"Tab3\", \"row_cells\": [\"PoseBusters\", \"Mixed\", \"AutoDock\", \"34.34\", \"36.69\", \"8.90\", \"51.17\", \"54.91\", \"11.87\"], \"selected_cell_zero_based\": 4, \"selected_cell_xml\": \"<td align=\\\"left\\\" colspan=\\\"1\\\" rowspan=\\\"1\\\"><bold>36</bold>.<bold>69</bold></td>\", \"caption\": \"Averaged 5-fold MolAS performance v.s. SBS across benchmarks\"}","artifact_sha256":"d556d47e0f7bbdc37eb62374b85ac9092dc7ff438fbae9e42892ac2cc784023f","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC13104262/fullTextXML"},"legacy_id":"lit-b3-044","legacy_row":{"id":"lit-b3-044","paper_id":"molas-2026","domain_id":"molecular-interactions","task":"Physically valid protein–ligand pose selection","model":"MolAS","model_version":"","dataset":"PoseBusters","dataset_version":"","split":"","metric":"RMSD ≤1 Å and PB-valid success","value":"36.69","unit":"%","uncertainty":"","protocol":"Averaged five-fold algorithm-selection performance on PoseBusters; joint RMSD and validity criterion.","source_locator":"Table 3, PoseBusters / Mixed / AutoDock row, MolAS success column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC13104262/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:29:32Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-b3-045","kind":"result","name":"Single best solver · RMSD ≤1 Å and PB-valid success · PoseBusters","description":"","status":"source_checked","facets":{"areas":["molecular-interactions"],"tasks":["Physically valid protein–ligand pose selection"]},"source_ids":["molas-2026"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b3-045"}],"attributes":{"printed_value":"34.34","numeric_value":"34.34","metric":"RMSD ≤1 Å and PB-valid success","metric_direction":"unknown","unit":"%","uncertainty":null,"source_locator":"Table 3, PoseBusters / Mixed / AutoDock row, SBS success column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:44:03.435454+00:00","notes":"PoseBusters, Mixed, AutoDock row within jointly trained with/without relaxation block. Selected RMSD <=1 Angstrom AND PB-valid group; five-fold average success percentage, not <=2 Angstrom. Inline bold digit nodes joined in original order. Source check verifies central value and table context, not experiment reproduction or all metadata.","evidence":"{\"table_xml_id\": \"Tab3\", \"row_cells\": [\"PoseBusters\", \"Mixed\", \"AutoDock\", \"34.34\", \"36.69\", \"8.90\", \"51.17\", \"54.91\", \"11.87\"], \"selected_cell_zero_based\": 3, \"selected_cell_xml\": \"<td align=\\\"left\\\" colspan=\\\"1\\\" rowspan=\\\"1\\\">34.34</td>\", \"caption\": \"Averaged 5-fold MolAS performance v.s. SBS across benchmarks\"}","artifact_sha256":"d556d47e0f7bbdc37eb62374b85ac9092dc7ff438fbae9e42892ac2cc784023f","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC13104262/fullTextXML"},"legacy_id":"lit-b3-045","legacy_row":{"id":"lit-b3-045","paper_id":"molas-2026","domain_id":"molecular-interactions","task":"Physically valid protein–ligand pose selection","model":"Single best solver","model_version":"","dataset":"PoseBusters","dataset_version":"","split":"","metric":"RMSD ≤1 Å and PB-valid success","value":"34.34","unit":"%","uncertainty":"","protocol":"Single best solver baseline under the same averaged five-fold selection test.","source_locator":"Table 3, PoseBusters / Mixed / AutoDock row, SBS success column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC13104262/","evaluation_origin":"independent_paper","reviewed_utc":"2026-09-15T23:29:32Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-b3-046","kind":"result","name":"AutoDock Vina holo · Docked frames best-matched RMSD <3 Å · α-synuclein Ligand 47 MD ensemble","description":"","status":"source_checked","facets":{"areas":["molecular-interactions"],"tasks":["Intrinsically disordered protein ensemble docking"]},"source_ids":["ensemble-idp-docking-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b3-046"}],"attributes":{"printed_value":"27.96","numeric_value":"27.96","metric":"Docked frames best-matched RMSD <3 Å","metric_direction":"unknown","unit":"%","uncertainty":null,"source_locator":"Table 2, Ligand 47 row, AutoDock Vina Holo Docking column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:56.275Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness.","evidence":"Table 2, Ligand 47 row, AutoDock Vina Holo Docking column; cell: 27.96 (98.30)","artifact_sha256":"d02d91cdde41cb76ec5c86b532dffc564879c69e764a8c6b7752460fbbfd24b7","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11785235/fullTextXML"},"legacy_id":"lit-b3-046","legacy_row":{"id":"lit-b3-046","paper_id":"ensemble-idp-docking-2025","domain_id":"molecular-interactions","task":"Intrinsically disordered protein ensemble docking","model":"AutoDock Vina holo","model_version":"","dataset":"α-synuclein Ligand 47 MD ensemble","dataset_version":"","split":"","metric":"Docked frames best-matched RMSD <3 Å","value":"27.96","unit":"%","uncertainty":"","protocol":"Fraction of docked frames within 3 Å of MD-observed bound pose; holo protein ensemble.","source_locator":"Table 2, Ligand 47 row, AutoDock Vina Holo Docking column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11785235/","evaluation_origin":"independent_paper","reviewed_utc":"2026-09-15T23:29:32Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-b3-047","kind":"result","name":"DiffDock holo · Docked frames best-matched RMSD <3 Å · α-synuclein Ligand 47 MD ensemble","description":"","status":"source_checked","facets":{"areas":["molecular-interactions"],"tasks":["Intrinsically disordered protein ensemble docking"]},"source_ids":["ensemble-idp-docking-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b3-047"}],"attributes":{"printed_value":"21.32","numeric_value":"21.32","metric":"Docked frames best-matched RMSD <3 Å","metric_direction":"unknown","unit":"%","uncertainty":null,"source_locator":"Table 2, Ligand 47 row, DiffDock Holo Docking column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:56.275Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness.","evidence":"Table 2, Ligand 47 row, DiffDock Holo Docking column; cell: 21.32 (97.21)","artifact_sha256":"d02d91cdde41cb76ec5c86b532dffc564879c69e764a8c6b7752460fbbfd24b7","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11785235/fullTextXML"},"legacy_id":"lit-b3-047","legacy_row":{"id":"lit-b3-047","paper_id":"ensemble-idp-docking-2025","domain_id":"molecular-interactions","task":"Intrinsically disordered protein ensemble docking","model":"DiffDock holo","model_version":"","dataset":"α-synuclein Ligand 47 MD ensemble","dataset_version":"","split":"","metric":"Docked frames best-matched RMSD <3 Å","value":"21.32","unit":"%","uncertainty":"","protocol":"Fraction of docked frames within 3 Å of MD-observed bound pose; holo protein ensemble.","source_locator":"Table 2, Ligand 47 row, DiffDock Holo Docking column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11785235/","evaluation_origin":"independent_paper","reviewed_utc":"2026-09-15T23:29:32Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-b3-048","kind":"result","name":"AK-score-ensemble · Pearson R · CASF-2016","description":"","status":"source_checked","facets":{"areas":["molecular-interactions"],"tasks":["Protein–ligand binding affinity scoring"]},"source_ids":["akscore-2020"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b3-048"}],"attributes":{"printed_value":"0.812","numeric_value":"0.812","metric":"Pearson R","metric_direction":"unknown","unit":"unitless","uncertainty":null,"source_locator":"Table 2, AK-score-ensemble / learning rate 0.0007 row, Scoring Pearson (R) column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:44:03.436853+00:00","notes":"CASF-2016 scoring Pearson R with learning rate0.0007. Single-model versus ensemble blocks kept distinct; ranking/docking scores not substituted. Source check verifies central value and table context, not experiment reproduction or all metadata.","evidence":"{\"table_xml_id\": \"ijms-21-08424-t002\", \"row_cells\": [\"AK-score-ensemble\", \"0.0007\", \"0.812\", \"0.670\", \"0.589\", \"0.698\", \"36.0\", \"51.4\", \"59.7\"], \"selected_cell_zero_based\": 2, \"selected_cell_xml\": \"<td align=\\\"center\\\" valign=\\\"middle\\\" style=\\\"border-bottom:solid thin\\\" rowspan=\\\"1\\\" colspan=\\\"1\\\">\\n<bold>0.812</bold>\\n</td>\", \"caption\": \"A comparison of prediction accuracy with the CASF-2016 dataset.\"}","artifact_sha256":"40cfd28dcd587599768ec99a6590ec593486475ff01c7b1d1f229b44aa91bf8d","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC7697539/fullTextXML"},"legacy_id":"lit-b3-048","legacy_row":{"id":"lit-b3-048","paper_id":"akscore-2020","domain_id":"molecular-interactions","task":"Protein–ligand binding affinity scoring","model":"AK-score-ensemble","model_version":"ensemble; learning rate 0.0007","dataset":"CASF-2016","dataset_version":"","split":"","metric":"Pearson R","value":"0.812","unit":"unitless","uncertainty":"","protocol":"CASF-2016 scoring-power evaluation.","source_locator":"Table 2, AK-score-ensemble / learning rate 0.0007 row, Scoring Pearson (R) column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC7697539/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:29:32Z"},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-b3-049","kind":"result","name":"AK-score-single · Pearson R · CASF-2016","description":"","status":"source_checked","facets":{"areas":["molecular-interactions"],"tasks":["Protein–ligand binding affinity scoring"]},"source_ids":["akscore-2020"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b3-049"}],"attributes":{"printed_value":"0.759","numeric_value":"0.759","metric":"Pearson R","metric_direction":"unknown","unit":"unitless","uncertainty":null,"source_locator":"Table 2, AK-score-single / learning rate 0.0007 row, Scoring Pearson (R) column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:44:03.437894+00:00","notes":"CASF-2016 scoring Pearson R with learning rate0.0007. Single-model versus ensemble blocks kept distinct; ranking/docking scores not substituted. Source check verifies central value and table context, not experiment reproduction or all metadata.","evidence":"{\"table_xml_id\": \"ijms-21-08424-t002\", \"row_cells\": [\"\", \"0.0007\", \"0.759\", \"0.616\", \"0.526\", \"0.640\", \"31.3\", \"47.1\", \"57.9\"], \"selected_cell_zero_based\": 2, \"selected_cell_xml\": \"<td align=\\\"center\\\" valign=\\\"middle\\\" rowspan=\\\"1\\\" colspan=\\\"1\\\">0.759</td>\", \"caption\": \"A comparison of prediction accuracy with the CASF-2016 dataset.\"}","artifact_sha256":"40cfd28dcd587599768ec99a6590ec593486475ff01c7b1d1f229b44aa91bf8d","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC7697539/fullTextXML"},"legacy_id":"lit-b3-049","legacy_row":{"id":"lit-b3-049","paper_id":"akscore-2020","domain_id":"molecular-interactions","task":"Protein–ligand binding affinity scoring","model":"AK-score-single","model_version":"single; learning rate 0.0007","dataset":"CASF-2016","dataset_version":"","split":"","metric":"Pearson R","value":"0.759","unit":"unitless","uncertainty":"","protocol":"CASF-2016 scoring-power evaluation.","source_locator":"Table 2, AK-score-single / learning rate 0.0007 row, Scoring Pearson (R) column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC7697539/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:29:32Z"},"missing_metadata":{"dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-b3-050","kind":"result","name":"PMF + ECFP + PF (LightGBM) · Pearson R · Fingerprint-scoring benchmark","description":"","status":"source_checked","facets":{"areas":["molecular-interactions"],"tasks":["Protein–ligand binding energy prediction"]},"source_ids":["fingerprint-scoring-2022"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b3-050"}],"attributes":{"printed_value":"0.79","numeric_value":"0.79","metric":"Pearson R","metric_direction":"unknown","unit":"unitless","uncertainty":null,"source_locator":"Table 1, PMF + ECFP + PF / LightGBM row, R column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:44:03.439460+00:00","notes":"Resolved two-row model rowspan: last LightGBM row is PMF+ECFP+PF; first LASSO row is PMF. Pearson R, not RMSE. Source check verifies central value and table context, not experiment reproduction or all metadata.","evidence":"{\"table_xml_id\": \"tbl1\", \"row_cells\": [\"LightGBM\", \"0.79\", \"1.64\"], \"selected_cell_zero_based\": 1, \"selected_cell_xml\": \"<td style=\\\"border:none;\\\" align=\\\"char\\\" char=\\\".\\\" colspan=\\\"1\\\" rowspan=\\\"1\\\">0.79</td>\", \"caption\": \"Pearson Correlation Coefficient (R) and RMSE between the Experimental Values and the Predicted Values by the Newly Developed Scoring Functions\"}","artifact_sha256":"47bd60c6392b801095fdb604de06c0d4bda6f555bae58e9955e491a5abf60576","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC9178954/fullTextXML"},"legacy_id":"lit-b3-050","legacy_row":{"id":"lit-b3-050","paper_id":"fingerprint-scoring-2022","domain_id":"molecular-interactions","task":"Protein–ligand binding energy prediction","model":"PMF + ECFP + PF (LightGBM)","model_version":"","dataset":"Fingerprint-scoring benchmark","dataset_version":"","split":"","metric":"Pearson R","value":"0.79","unit":"unitless","uncertainty":"","protocol":"Binding-energy model using ligand and protein fingerprints with LightGBM.","source_locator":"Table 1, PMF + ECFP + PF / LightGBM row, R column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC9178954/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:29:32Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-b3-051","kind":"result","name":"PMF (LASSO) · Pearson R · Fingerprint-scoring benchmark","description":"","status":"source_checked","facets":{"areas":["molecular-interactions"],"tasks":["Protein–ligand binding energy prediction"]},"source_ids":["fingerprint-scoring-2022"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b3-051"}],"attributes":{"printed_value":"0.67","numeric_value":"0.67","metric":"Pearson R","metric_direction":"unknown","unit":"unitless","uncertainty":null,"source_locator":"Table 1, PMF / LASSO row, R column","review":{"method":"independent_ai_table_review","reviewer":"Codex omics research agent; independent source-table review, not human review","reviewed_at":"2026-09-16T10:44:03.440812+00:00","notes":"Resolved two-row model rowspan: last LightGBM row is PMF+ECFP+PF; first LASSO row is PMF. Pearson R, not RMSE. Source check verifies central value and table context, not experiment reproduction or all metadata.","evidence":"{\"table_xml_id\": \"tbl1\", \"row_cells\": [\"PMF\", \"LASSO\", \"0.67\", \"2.04\"], \"selected_cell_zero_based\": 2, \"selected_cell_xml\": \"<td style=\\\"border:none;\\\" align=\\\"char\\\" char=\\\".\\\" colspan=\\\"1\\\" rowspan=\\\"1\\\">0.67</td>\", \"caption\": \"Pearson Correlation Coefficient (R) and RMSE between the Experimental Values and the Predicted Values by the Newly Developed Scoring Functions\"}","artifact_sha256":"47bd60c6392b801095fdb604de06c0d4bda6f555bae58e9955e491a5abf60576","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC9178954/fullTextXML"},"legacy_id":"lit-b3-051","legacy_row":{"id":"lit-b3-051","paper_id":"fingerprint-scoring-2022","domain_id":"molecular-interactions","task":"Protein–ligand binding energy prediction","model":"PMF (LASSO)","model_version":"","dataset":"Fingerprint-scoring benchmark","dataset_version":"","split":"","metric":"Pearson R","value":"0.67","unit":"unitless","uncertainty":"","protocol":"PMF-only LASSO baseline evaluated by the same authors.","source_locator":"Table 1, PMF / LASSO row, R column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC9178954/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:29:32Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-b4-001","kind":"result","name":"ARSENAL+ChromBPNet · AUROC · Yoruban LCL dsQTLs","description":"","status":"source_checked","facets":{"areas":["dna-genomes"],"tasks":["regulatory-variant scoring"]},"source_ids":["arsenal-regulatory-dna-2026"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b4-001"}],"attributes":{"printed_value":"0.896","numeric_value":"0.896","metric":"AUROC","metric_direction":"unknown","unit":"fraction","uncertainty":"±0.016","source_locator":"Table 1, Yoruban LCL dsQTLs section, ARSENAL+ChromBPNet row, AUROC column","review":{"method":"independent_ai_table_review","reviewer":"Codex primary curator table review","reviewed_at":"2026-09-16T10:41:06Z","notes":"Yoruban LCL dsQTLs; ARSENAL+ChromBPNet AUROC Transcription verified; experimental claims not independently reproduced.","evidence":"0.896 ±0.016","artifact_sha256":"4a264956e47fc633aaff6573aac368dc691dd5de709b27c7421c078608ff542a","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12889687/fullTextXML"},"legacy_id":"lit-b4-001","legacy_row":{"id":"lit-b4-001","paper_id":"arsenal-regulatory-dna-2026","domain_id":"dna-genomes","task":"regulatory-variant scoring","model":"ARSENAL+ChromBPNet","model_version":"","dataset":"Yoruban LCL dsQTLs","dataset_version":"","split":"","metric":"AUROC","value":"0.896","unit":"fraction","uncertainty":"±0.016","protocol":"Supervised ChromBPNet variant scoring with ARSENAL motif-discovery regularization","source_locator":"Table 1, Yoruban LCL dsQTLs section, ARSENAL+ChromBPNet row, AUROC column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12889687/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:37:05Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"lit-b4-002","kind":"result","name":"PlantCAD2 · AUROC · Andropogoneae genome-wide conservation","description":"","status":"source_checked","facets":{"areas":["dna-genomes"],"tasks":["cross-species conservation prediction"]},"source_ids":["plantcad2-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b4-002"}],"attributes":{"printed_value":"0.725","numeric_value":"0.725","metric":"AUROC","metric_direction":"unknown","unit":"fraction","uncertainty":null,"source_locator":"Table 1, Cross-species evolutionary conservation > Conservation within Andropogoneae (Genome-wide) row, PlantCAD2 AUROC entry","review":{"method":"independent_ai_table_review","reviewer":"Codex primary curator table review","reviewed_at":"2026-09-16T10:41:06Z","notes":"First comparison entry is PlantCAD2; AUROC is 0.725 versus comparator 0.691. Transcription verified; experimental claims not independently reproduced.","evidence":"0.725 vs 0.691","artifact_sha256":"4891955edb33af61e36dad51110574aa242e77c2df36429f100ff5a54bd8bd4b","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12425018/fullTextXML"},"legacy_id":"lit-b4-002","legacy_row":{"id":"lit-b4-002","paper_id":"plantcad2-2025","domain_id":"dna-genomes","task":"cross-species conservation prediction","model":"PlantCAD2","model_version":"","dataset":"Andropogoneae genome-wide conservation","dataset_version":"","split":"","metric":"AUROC","value":"0.725","unit":"fraction","uncertainty":"","protocol":"Zero-shot score for conserved versus non-conserved sites from alignments of 35 Andropogoneae genomes","source_locator":"Table 1, Cross-species evolutionary conservation > Conservation within Andropogoneae (Genome-wide) row, PlantCAD2 AUROC entry","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12425018/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:37:05Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-b4-003","kind":"result","name":"Stacking-Auto · accuracy · enhancer independent comparison","description":"","status":"source_checked","facets":{"areas":["dna-genomes"],"tasks":["enhancer prediction"]},"source_ids":["hi-enhancer-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b4-003"}],"attributes":{"printed_value":"80.50","numeric_value":"80.50","metric":"accuracy","metric_direction":"unknown","unit":"percent","uncertainty":null,"source_locator":"Table 2, Ours (Stacking-Auto) row, Accuracy column","review":{"method":"independent_ai_table_review","reviewer":"Codex primary curator table review","reviewed_at":"2026-09-16T10:41:06Z","notes":"Ours row, Accuracy column; original source method is the Stacking-Auto stage. Transcription verified; experimental claims not independently reproduced.","evidence":"80.50","artifact_sha256":"c86488c9f60329b7a3c4370598e7a0a9e4c8c45d1758b87007bfc8242376b009","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12758598/fullTextXML"},"legacy_id":"lit-b4-003","legacy_row":{"id":"lit-b4-003","paper_id":"hi-enhancer-2025","domain_id":"dna-genomes","task":"enhancer prediction","model":"Stacking-Auto","model_version":"","dataset":"enhancer independent comparison","dataset_version":"","split":"","metric":"accuracy","value":"80.50","unit":"percent","uncertainty":"","protocol":"Two-stage Hi-Enhancer system; paper Table 2 method comparison","source_locator":"Table 2, Ours (Stacking-Auto) row, Accuracy column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12758598/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:37:05Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-b4-004","kind":"result","name":"position-aware CNN · AUROC · human enhancer dataset","description":"","status":"source_checked","facets":{"areas":["dna-genomes"],"tasks":["enhancer prediction"]},"source_ids":["enhancer-position-encoding-2024"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b4-004"}],"attributes":{"printed_value":"0.94","numeric_value":"0.94","metric":"AUROC","metric_direction":"unknown","unit":"fraction","uncertainty":null,"source_locator":"Table 2, Human section, CNN row, AUC column","review":{"method":"independent_ai_table_review","reviewer":"Codex primary curator table review","reviewed_at":"2026-09-16T10:41:06Z","notes":"Human dataset row, CNN, AUC column. Transcription verified; experimental claims not independently reproduced.","evidence":"0.94","artifact_sha256":"0183b6a111b1b02344cad35a571a1fd2c56257e406c5be1df69f7902c5d06749","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11167433/fullTextXML"},"legacy_id":"lit-b4-004","legacy_row":{"id":"lit-b4-004","paper_id":"enhancer-position-encoding-2024","domain_id":"dna-genomes","task":"enhancer prediction","model":"position-aware CNN","model_version":"","dataset":"human enhancer dataset","dataset_version":"","split":"","metric":"AUROC","value":"0.94","unit":"fraction","uncertainty":"","protocol":"Nucleotide position-aware feature encoding; average assessment of CNN classifier","source_locator":"Table 2, Human section, CNN row, AUC column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11167433/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:37:05Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-b4-005","kind":"result","name":"ADAR-GPT continual · F1 · liver editing sites","description":"","status":"source_checked","facets":{"areas":["rna-transcriptomes"],"tasks":["A-to-I RNA editing site prediction"]},"source_ids":["adar-gpt-editing-2026"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b4-005"}],"attributes":{"printed_value":"0.763","numeric_value":"0.763","metric":"F1","metric_direction":"unknown","unit":"fraction","uncertainty":null,"source_locator":"Table 2, Adar-GPT (continual) row, F1 column","review":{"method":"independent_ai_table_review","reviewer":"Codex primary curator table review","reviewed_at":"2026-09-16T10:41:06Z","notes":"Adar-GPT continual row; XML inline decimal reordered by parser, original text verified separately. Transcription verified; experimental claims not independently reproduced.","evidence":"0.763","artifact_sha256":"cc8c7eb928f246f1f347a8822f614cd3475381c35eef6d579032ce441580198e","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12798952/fullTextXML"},"legacy_id":"lit-b4-005","legacy_row":{"id":"lit-b4-005","paper_id":"adar-gpt-editing-2026","domain_id":"rna-transcriptomes","task":"A-to-I RNA editing site prediction","model":"ADAR-GPT continual","model_version":"","dataset":"liver editing sites","dataset_version":"","split":"15% validation set","metric":"F1","value":"0.763","unit":"fraction","uncertainty":"","protocol":"Curriculum plus 15% fine-tuning; 201-nt sequence windows; decision threshold 0.5","source_locator":"Table 2, Adar-GPT (continual) row, F1 column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12798952/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:37:05Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-b4-006","kind":"result","name":"R3Design · sequence recovery · Rfam","description":"","status":"source_checked","facets":{"areas":["rna-transcriptomes"],"tasks":["RNA sequence design"]},"source_ids":["r3design-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b4-006"}],"attributes":{"printed_value":"43.27","numeric_value":"43.27","metric":"sequence recovery","metric_direction":"unknown","unit":"percent","uncertainty":null,"source_locator":"Table 3, R3Design row, Recovery (%) > Rfam column","review":{"method":"independent_ai_table_review","reviewer":"Codex primary curator table review","reviewed_at":"2026-09-16T10:41:06Z","notes":"R3Design row, first Recovery column Rfam; 43.27 plus/minus0.56. Transcription verified; experimental claims not independently reproduced.","evidence":"43.27 \\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{upgreek} \\usepackage{mathrsfs} \\setlength{\\oddsidemargin}{-69pt} \\begin{document} $\\pm $\\end{document} 0.56","artifact_sha256":"b49c9ad846e46b11b240aace8e6bfaf953b842df166b69aee4843c02e9349779","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11685104/fullTextXML"},"legacy_id":"lit-b4-006","legacy_row":{"id":"lit-b4-006","paper_id":"r3design-2025","domain_id":"rna-transcriptomes","task":"RNA sequence design","model":"R3Design","model_version":"","dataset":"Rfam","dataset_version":"","split":"external","metric":"sequence recovery","value":"43.27","unit":"percent","uncertainty":"","protocol":"Tertiary-structure-conditioned RNA sequence design; external Rfam assessment","source_locator":"Table 3, R3Design row, Recovery (%) > Rfam column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11685104/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:37:05Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-b4-007","kind":"result","name":"CUPID Data-aug-Avg · AUROC · ncRNA interaction pairs","description":"","status":"source_checked","facets":{"areas":["rna-transcriptomes"],"tasks":["non-coding RNA pairwise interaction prediction"]},"source_ids":["cupid-rna-interactions-2026"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b4-007"}],"attributes":{"printed_value":"0.919","numeric_value":"0.919","metric":"AUROC","metric_direction":"unknown","unit":"fraction","uncertainty":null,"source_locator":"Table 1, CUPID > Data-aug-Avg row, AUROC column","review":{"method":"independent_ai_table_review","reviewer":"Codex primary curator table review","reviewed_at":"2026-09-16T10:41:06Z","notes":"CUPID section Data-aug-Avg row; AUROC not AUPRC. Transcription verified; experimental claims not independently reproduced.","evidence":"0.919","artifact_sha256":"0e6719410b390ee9c4858bb9321042851100fb74df3aa109bf2af2b8aaff7ac1","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12957212/fullTextXML"},"legacy_id":"lit-b4-007","legacy_row":{"id":"lit-b4-007","paper_id":"cupid-rna-interactions-2026","domain_id":"rna-transcriptomes","task":"non-coding RNA pairwise interaction prediction","model":"CUPID Data-aug-Avg","model_version":"","dataset":"ncRNA interaction pairs","dataset_version":"","split":"","metric":"AUROC","value":"0.919","unit":"fraction","uncertainty":"","protocol":"Data augmentation with average pooling for molecule-level ncRNA embeddings","source_locator":"Table 1, CUPID > Data-aug-Avg row, AUROC column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12957212/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:37:05Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-b4-008","kind":"result","name":"ProteinBERT LLM-encoding model · AUROC · mRNA-RBP pairs","description":"","status":"source_checked","facets":{"areas":["rna-transcriptomes"],"tasks":["mRNA-protein interaction prediction"]},"source_ids":["mrna-protein-diversity-2026"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b4-008"}],"attributes":{"printed_value":"71.5","numeric_value":"71.5","metric":"AUROC","metric_direction":"unknown","unit":"percent","uncertainty":null,"source_locator":"Table 2, RBP-aware test set row, auROC (%) column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:57.257Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness.","evidence":"Table 2, RBP-aware test set row, auROC (%) column; cell: 71.5","artifact_sha256":"94f9fe22a5f0e6c8619e4af994eb4f6ded7417efcf0c1240380269280f97f9d1","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC13235059/fullTextXML"},"legacy_id":"lit-b4-008","legacy_row":{"id":"lit-b4-008","paper_id":"mrna-protein-diversity-2026","domain_id":"rna-transcriptomes","task":"mRNA-protein interaction prediction","model":"ProteinBERT LLM-encoding model","model_version":"","dataset":"mRNA-RBP pairs","dataset_version":"","split":"RBP-aware test set","metric":"AUROC","value":"71.5","unit":"percent","uncertainty":"","protocol":"LLM encoding of protein partner; RBP-aware partition tests generalization to unseen protein diversity","source_locator":"Table 2, RBP-aware test set row, auROC (%) column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC13235059/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:37:05Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-b4-009","kind":"result","name":"ESM2 650M · AUROC · human and viral proteins","description":"","status":"source_checked","facets":{"areas":["proteins-complexes"],"tasks":["human-versus-viral protein classification"]},"source_ids":["viral-immune-mimicry-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b4-009"}],"attributes":{"printed_value":"99.67","numeric_value":"99.67","metric":"AUROC","metric_direction":"unknown","unit":"percent","uncertainty":null,"source_locator":"Table 1, ESM2 650M row, AUC (%) column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:57.274Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness.","evidence":"Table 1, ESM2 650M row, AUC (%) column; cell: 99.67","artifact_sha256":"15250af2f75f70e2b6a3725d00bc7e276ed9d2d4a54f0ae9d6eaabf6be13e4a1","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12474240/fullTextXML"},"legacy_id":"lit-b4-009","legacy_row":{"id":"lit-b4-009","paper_id":"viral-immune-mimicry-2025","domain_id":"proteins-complexes","task":"human-versus-viral protein classification","model":"ESM2 650M","model_version":"","dataset":"human and viral proteins","dataset_version":"","split":"","metric":"AUROC","value":"99.67","unit":"percent","uncertainty":"","protocol":"ESM2 650M embedding-based human-virus classifier","source_locator":"Table 1, ESM2 650M row, AUC (%) column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12474240/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:37:05Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-b4-010","kind":"result","name":"ProtT5 embeddings + ensemble classifier · AUROC · Dset_448","description":"","status":"source_checked","facets":{"areas":["proteins-complexes"],"tasks":["protein-protein binding-site prediction"]},"source_ids":["protein-binding-sites-2023"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b4-010"}],"attributes":{"printed_value":"0.810","numeric_value":"0.810","metric":"AUROC","metric_direction":"unknown","unit":"fraction","uncertainty":null,"source_locator":"Table 2, Dset_448 section, ProtT5 row, AUROC column","review":{"method":"independent_ai_table_review","reviewer":"Codex primary curator table review","reviewed_at":"2026-09-16T10:41:06Z","notes":"Dset_448 block; ProtT5 AUROC, downstream ensemble retained in protocol. Transcription verified; experimental claims not independently reproduced.","evidence":"0.810","artifact_sha256":"491711aa7186e74bf33f6d601c4ea6a8f565e770938fe1f91a0cd347b8f06f3f","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC9849350/fullTextXML"},"legacy_id":"lit-b4-010","legacy_row":{"id":"lit-b4-010","paper_id":"protein-binding-sites-2023","domain_id":"proteins-complexes","task":"protein-protein binding-site prediction","model":"ProtT5 embeddings + ensemble classifier","model_version":"","dataset":"Dset_448","dataset_version":"","split":"","metric":"AUROC","value":"0.810","unit":"fraction","uncertainty":"","protocol":"Explainable ensemble binding-site predictor using ProtT5 features","source_locator":"Table 2, Dset_448 section, ProtT5 row, AUROC column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC9849350/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:37:05Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-b4-011","kind":"result","name":"CLAPE-SMB with ESM-2 · AUROC · SJC","description":"","status":"source_checked","facets":{"areas":["proteins-complexes"],"tasks":["protein-small molecule binding-site prediction"]},"source_ids":["clape-smb-2024"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b4-011"}],"attributes":{"printed_value":"0.917","numeric_value":"0.917","metric":"AUROC","metric_direction":"unknown","unit":"fraction","uncertainty":null,"source_locator":"Table 5, ESM-2 / SJC row, AUROC column","review":{"method":"independent_ai_table_review","reviewer":"Codex primary curator table review","reviewed_at":"2026-09-16T10:41:06Z","notes":"ESM-2 on SJC AUROC. Transcription verified; experimental claims not independently reproduced.","evidence":"0.917","artifact_sha256":"215919244c3dd2dfb0b55fce91c211430fd8d4aee4bb28bd03eab9f4feb73e62","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11542454/fullTextXML"},"legacy_id":"lit-b4-011","legacy_row":{"id":"lit-b4-011","paper_id":"clape-smb-2024","domain_id":"proteins-complexes","task":"protein-small molecule binding-site prediction","model":"CLAPE-SMB with ESM-2","model_version":"","dataset":"SJC","dataset_version":"","split":"","metric":"AUROC","value":"0.917","unit":"fraction","uncertainty":"","protocol":"Contrastive CLAPE-SMB binding-site predictor with ESM-2 feature extractor","source_locator":"Table 5, ESM-2 / SJC row, AUROC column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11542454/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:37:05Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-b4-012","kind":"result","name":"Vaxign-DL + ESM · AUPRC · vaccine candidate validation","description":"","status":"source_checked","facets":{"areas":["proteins-complexes"],"tasks":["vaccine-antigen candidate prediction"]},"source_ids":["vaxign-esm-2024"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b4-012"}],"attributes":{"printed_value":"0.92","numeric_value":"0.92","metric":"AUPRC","metric_direction":"unknown","unit":"fraction","uncertainty":"±0.013","source_locator":"Table 2, 4 Layers row, AUPRC column","review":{"method":"independent_ai_table_review","reviewer":"Codex primary curator table review","reviewed_at":"2026-09-16T10:41:06Z","notes":"Source spells 4 Layerss; AUPRC0.92±0.013. Transcription verified; experimental claims not independently reproduced.","evidence":"0.92 ± 0.013","artifact_sha256":"b76fff917addd0e9ff8a2fc843496132ecf832d3ceef248e3edf2dbc78baab5f","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11398487/fullTextXML"},"legacy_id":"lit-b4-012","legacy_row":{"id":"lit-b4-012","paper_id":"vaxign-esm-2024","domain_id":"proteins-complexes","task":"vaccine-antigen candidate prediction","model":"Vaxign-DL + ESM","model_version":"","dataset":"vaccine candidate validation","dataset_version":"","split":"","metric":"AUPRC","value":"0.92","unit":"fraction","uncertainty":"±0.013","protocol":"Combined skip architecture, four layers, ESM-generated sequence features","source_locator":"Table 2, 4 Layers row, AUPRC column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11398487/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:37:05Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"}}}
{"id":"lit-b4-013","kind":"result","name":"scGPT + residual geometry · AUROC · immune tissue","description":"","status":"source_checked","facets":{"areas":["cells-tissues"],"tasks":["gene-regulatory signal prediction"]},"source_ids":["single-cell-residual-geometry-2026"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b4-013"}],"attributes":{"printed_value":"0.677","numeric_value":"0.677","metric":"AUROC","metric_direction":"unknown","unit":"fraction","uncertainty":null,"source_locator":"Table 4, Immune row, scGPT > +geom AUROC column","review":{"method":"independent_ai_table_review","reviewer":"Codex primary curator table review","reviewed_at":"2026-09-16T10:41:06Z","notes":"Immune row, scGPT +geom (second numeric column), not Geneformer or delta. Transcription verified; experimental claims not independently reproduced.","evidence":"0.677","artifact_sha256":"77546faec51cfb5c78b73c5943f940c4e0097d130b1ddde4ab8287b507b36df6","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC13418759/fullTextXML"},"legacy_id":"lit-b4-013","legacy_row":{"id":"lit-b4-013","paper_id":"single-cell-residual-geometry-2026","domain_id":"cells-tissues","task":"gene-regulatory signal prediction","model":"scGPT + residual geometry","model_version":"","dataset":"immune tissue","dataset_version":"","split":"","metric":"AUROC","value":"0.677","unit":"fraction","uncertainty":"","protocol":"Asymmetric extraction, PCA-64 centered cosine geometry added to scGPT baseline","source_locator":"Table 4, Immune row, scGPT > +geom AUROC column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC13418759/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:37:05Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-b4-014","kind":"result","name":"GREmLN · F1 · non-immune cells","description":"","status":"source_checked","facets":{"areas":["cells-tissues"],"tasks":["cell-type annotation"]},"source_ids":["gremln-2026"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b4-014"}],"attributes":{"printed_value":"0.937","numeric_value":"0.937","metric":"F1","metric_direction":"unknown","unit":"fraction","uncertainty":null,"source_locator":"Table 2, Cell type annotation(zero-shot), Non-immune cells, F1 row, GREmLN column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:57.502Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness.","evidence":"Table 2, Cell type annotation(zero-shot), Non-immune cells, F1 row, GREmLN column; cell: 0.937","artifact_sha256":"3a20c4ededb749fc3f1120baf16dcfebe3fcb30418a91c445cfd91a7b5fdf553","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC13060794/fullTextXML"},"legacy_id":"lit-b4-014","legacy_row":{"id":"lit-b4-014","paper_id":"gremln-2026","domain_id":"cells-tissues","task":"cell-type annotation","model":"GREmLN","model_version":"","dataset":"non-immune cells","dataset_version":"","split":"zero-shot","metric":"F1","value":"0.937","unit":"fraction","uncertainty":"","protocol":"Zero-shot cell-type annotation using pre-trained cellular graph foundation model","source_locator":"Table 2, Cell type annotation(zero-shot), Non-immune cells, F1 row, GREmLN column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC13060794/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:37:05Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-b4-015","kind":"result","name":"Cell-DINO ViT-L · F1 · HPA-FoV","description":"","status":"source_checked","facets":{"areas":["cells-tissues"],"tasks":["protein localization classification"]},"source_ids":["cell-dino-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b4-015"}],"attributes":{"printed_value":"65.5","numeric_value":"65.5","metric":"F1","metric_direction":"unknown","unit":"percent","uncertainty":null,"source_locator":"Table 2, HPA-FoV section, Cell-DINO row, PL column","review":{"method":"independent_ai_table_review","reviewer":"Codex primary curator table review","reviewed_at":"2026-09-16T10:41:06Z","notes":"HPA-FoV Cell-DINO PL column (protein localisation), not CL. Transcription verified; experimental claims not independently reproduced.","evidence":"65.5","artifact_sha256":"12a53a78c70b3033c3351cf7afd4da42ebc98bb3281308f07e71e5baffc153a0","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12826486/fullTextXML"},"legacy_id":"lit-b4-015","legacy_row":{"id":"lit-b4-015","paper_id":"cell-dino-2025","domain_id":"cells-tissues","task":"protein localization classification","model":"Cell-DINO ViT-L","model_version":"","dataset":"HPA-FoV","dataset_version":"","split":"","metric":"F1","value":"65.5","unit":"percent","uncertainty":"","protocol":"Self-supervised microscopy embedding pre-trained on HPA-FoV; downstream protein-localization classifier. Dataset-specific pretraining; the paper does not claim a general-purpose foundation model that generalizes beyond these benchmarks.","source_locator":"Table 2, HPA-FoV section, Cell-DINO row, PL column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12826486/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:37:05Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-b4-016","kind":"result","name":"scGen · precision at 50% recall · stimulated immune PBMC","description":"","status":"source_checked","facets":{"areas":["cells-tissues"],"tasks":["differentially expressed gene identification"]},"source_ids":["insilico-perturbation-auprc-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b4-016"}],"attributes":{"printed_value":"0.91","numeric_value":"0.91","metric":"precision at 50% recall","metric_direction":"unknown","unit":"fraction","uncertainty":null,"source_locator":"Table 3, CD14+Mono section, scGen row, Precision at 50% Recall column","review":{"method":"independent_ai_table_review","reviewer":"Codex primary curator table review","reviewed_at":"2026-09-16T10:41:06Z","notes":"CD14+Mono scGen; precision at50%recall. Transcription verified; experimental claims not independently reproduced.","evidence":"0.91","artifact_sha256":"2715709d94f84744afa32cafdcaa72efd206d63af8c60afe7619b2cb90108b6b","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12400816/fullTextXML"},"legacy_id":"lit-b4-016","legacy_row":{"id":"lit-b4-016","paper_id":"insilico-perturbation-auprc-2025","domain_id":"cells-tissues","task":"differentially expressed gene identification","model":"scGen","model_version":"","dataset":"stimulated immune PBMC","dataset_version":"","split":"CD14+Mono","metric":"precision at 50% recall","value":"0.91","unit":"fraction","uncertainty":"","protocol":"In-silico perturbation assessment with precision sampled at fixed 50% recall","source_locator":"Table 3, CD14+Mono section, scGen row, Precision at 50% Recall column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12400816/","evaluation_origin":"independent_paper","reviewed_utc":"2026-09-15T23:37:05Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-b4-017","kind":"result","name":"TCINet + HTRS · F1 · MetaHIT","description":"","status":"source_checked","facets":{"areas":["microbes-communities"],"tasks":["pathogen detection"]},"source_ids":["metagenomic-pathogens-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b4-017"}],"attributes":{"printed_value":"0.84","numeric_value":"0.84","metric":"F1","metric_direction":"unknown","unit":"fraction","uncertainty":null,"source_locator":"Table 3, MetaHIT dataset section, TCINet + HTRS (Ours) row, F1-score column","review":{"method":"independent_ai_table_review","reviewer":"Codex primary curator table review","reviewed_at":"2026-09-16T10:41:06Z","notes":"MetaHIT block TCINet+HTRS F1-score. Transcription verified; experimental claims not independently reproduced.","evidence":"0.84","artifact_sha256":"aa88de1b0f0fd7ba1fedc1074bba9ba6ce199a0b723072a04d531db4585ae77c","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12493982/fullTextXML"},"legacy_id":"lit-b4-017","legacy_row":{"id":"lit-b4-017","paper_id":"metagenomic-pathogens-2025","domain_id":"microbes-communities","task":"pathogen detection","model":"TCINet + HTRS","model_version":"","dataset":"MetaHIT","dataset_version":"","split":"","metric":"F1","value":"0.84","unit":"fraction","uncertainty":"","protocol":"Taxonomy-constrained inference network with hierarchical taxonomy representation","source_locator":"Table 3, MetaHIT dataset section, TCINet + HTRS (Ours) row, F1-score column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12493982/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:37:05Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-b4-018","kind":"result","name":"DETIRE · accuracy · testing viral metagenome dataset","description":"","status":"source_checked","facets":{"areas":["microbes-communities"],"tasks":["viral sequence detection"]},"source_ids":["detire-viral-metagenomes-2023"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b4-018"}],"attributes":{"printed_value":"0.8772","numeric_value":"0.8772","metric":"accuracy","metric_direction":"unknown","unit":"fraction","uncertainty":null,"source_locator":"Table 1, Accuracy row, DETIRE column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:58.392Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness.","evidence":"Table 1, Accuracy row, DETIRE column; cell: 0.8772","artifact_sha256":"9ff7d32758620f7b0b0628425f62abff103ca2e33269ce3763383584bcebfc3c","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC10313334/fullTextXML"},"legacy_id":"lit-b4-018","legacy_row":{"id":"lit-b4-018","paper_id":"detire-viral-metagenomes-2023","domain_id":"microbes-communities","task":"viral sequence detection","model":"DETIRE","model_version":"","dataset":"testing viral metagenome dataset","dataset_version":"","split":"test","metric":"accuracy","value":"0.8772","unit":"fraction","uncertainty":"","protocol":"Hybrid deep learning virus-fragment classifier on paper testing dataset","source_locator":"Table 1, Accuracy row, DETIRE column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC10313334/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:37:05Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-b4-019","kind":"result","name":"PC-mer + LR · accuracy · AMP","description":"","status":"source_checked","facets":{"areas":["microbes-communities"],"tasks":["metagenomic genus classification"]},"source_ids":["pc-mer-2024"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b4-019"}],"attributes":{"printed_value":"96.95","numeric_value":"96.95","metric":"accuracy","metric_direction":"unknown","unit":"percent","uncertainty":null,"source_locator":"Table 3, AMP section, PC-mer + LR k=8 row, Accuracy (%) column","review":{"method":"independent_ai_table_review","reviewer":"Codex primary curator table review","reviewed_at":"2026-09-16T10:41:06Z","notes":"AMP block PC-mer+LR section, k=8, first numeric value after k is Accuracy. Transcription verified; experimental claims not independently reproduced.","evidence":"96.95","artifact_sha256":"0b0a225fa6f5f3ba41dffc7b320c91f47738301cf45835260eecaa03b704a097","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11293629/fullTextXML"},"legacy_id":"lit-b4-019","legacy_row":{"id":"lit-b4-019","paper_id":"pc-mer-2024","domain_id":"microbes-communities","task":"metagenomic genus classification","model":"PC-mer + LR","model_version":"","dataset":"AMP","dataset_version":"","split":"genus-level","metric":"accuracy","value":"96.95","unit":"percent","uncertainty":"","protocol":"k=8 PC-mer feature extraction with logistic regression on AMP genus-classification dataset","source_locator":"Table 3, AMP section, PC-mer + LR k=8 row, Accuracy (%) column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11293629/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:37:05Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-b4-020","kind":"result","name":"MDL4Microbiome · accuracy · CRC microbiome cohort","description":"","status":"source_checked","facets":{"areas":["microbes-communities"],"tasks":["microbiome disease-state classification"]},"source_ids":["mdl4microbiome-2022"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b4-020"}],"attributes":{"printed_value":"0.97","numeric_value":"0.97","metric":"accuracy","metric_direction":"unknown","unit":"fraction","uncertainty":null,"source_locator":"Table 3, CRC row, MDL4Microbiome column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:58.492Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness.","evidence":"Table 3, CRC row, MDL4Microbiome column; cell: 0.97","artifact_sha256":"72330eda245ac97c5de5d47a491bb86d9b8f527a03ae25fa41cae5bfb637143b","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC8763943/fullTextXML"},"legacy_id":"lit-b4-020","legacy_row":{"id":"lit-b4-020","paper_id":"mdl4microbiome-2022","domain_id":"microbes-communities","task":"microbiome disease-state classification","model":"MDL4Microbiome","model_version":"","dataset":"CRC microbiome cohort","dataset_version":"","split":"","metric":"accuracy","value":"0.97","unit":"fraction","uncertainty":"","protocol":"Multimodal deep learning model on colorectal-cancer versus healthy microbiome samples","source_locator":"Table 3, CRC row, MDL4Microbiome column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC8763943/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:37:05Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-b4-021","kind":"result","name":"binding-affinity meta-model · Pearson correlation · CASF-2016","description":"","status":"source_checked","facets":{"areas":["molecular-interactions"],"tasks":["protein-ligand binding affinity prediction"]},"source_ids":["ligand-affinity-meta-model-2024"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b4-021"}],"attributes":{"printed_value":"0.777","numeric_value":"0.777","metric":"Pearson correlation","metric_direction":"unknown","unit":"unitless","uncertainty":null,"source_locator":"Table 4, Meta-models row, CASF-2016 Benchmark > PCC column","review":{"method":"independent_ai_table_review","reviewer":"Codex primary curator table review","reviewed_at":"2026-09-16T10:41:06Z","notes":"Meta-models CASF-2016 PCC. Confirmed XML training-set rowspan inherits preceding row, so0.777 maps to PCC. Transcription verified; experimental claims not independently reproduced.","evidence":"0.777","artifact_sha256":"0be25fe75bc0b2eb8065136555763bbae5964ea3a8fdb8c5de79ff96445f6a29","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11632770/fullTextXML"},"legacy_id":"lit-b4-021","legacy_row":{"id":"lit-b4-021","paper_id":"ligand-affinity-meta-model-2024","domain_id":"molecular-interactions","task":"protein-ligand binding affinity prediction","model":"binding-affinity meta-model","model_version":"","dataset":"CASF-2016","dataset_version":"","split":"core benchmark","metric":"Pearson correlation","value":"0.777","unit":"unitless","uncertainty":"","protocol":"Sequence-or-structure meta-model; predicts ln(Kd/Ki) using docked and deep-learning components","source_locator":"Table 4, Meta-models row, CASF-2016 Benchmark > PCC column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11632770/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:37:05Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-b4-022","kind":"result","name":"DeepInterAware · AUROC · HIV neutralization","description":"","status":"source_checked","facets":{"areas":["molecular-interactions"],"tasks":["antigen-antibody HIV neutralization prediction"]},"source_ids":["deepinteraware-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b4-022"}],"attributes":{"printed_value":"0.826","numeric_value":"0.826","metric":"AUROC","metric_direction":"unknown","unit":"fraction","uncertainty":"±0.017","source_locator":"Table 2, Ab Unseen section, DeepInterAware row, AUROC column","review":{"method":"independent_ai_table_review","reviewer":"Codex primary curator table review","reviewed_at":"2026-09-16T10:41:06Z","notes":"Ab Unseen block DeepInterAware AUROC0.826±0.017, not Ag Unseen. Transcription verified; experimental claims not independently reproduced.","evidence":"0.826 ± 0.017","artifact_sha256":"25d3561934965f754d8712ec02b2052e9a3979e433b88ecebd5db14e930f17a1","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11967782/fullTextXML"},"legacy_id":"lit-b4-022","legacy_row":{"id":"lit-b4-022","paper_id":"deepinteraware-2025","domain_id":"molecular-interactions","task":"antigen-antibody HIV neutralization prediction","model":"DeepInterAware","model_version":"","dataset":"HIV neutralization","dataset_version":"","split":"antibody-unseen","metric":"AUROC","value":"0.826","unit":"fraction","uncertainty":"±0.017","protocol":"Sequence-based interface-aware model, antibody-unseen split","source_locator":"Table 2, Ab Unseen section, DeepInterAware row, AUROC column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11967782/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:37:05Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract"}}}
{"id":"lit-b4-023","kind":"result","name":"TransBind · AUROC · genome-wide TF binding sites","description":"","status":"source_checked","facets":{"areas":["molecular-interactions"],"tasks":["transcription-factor DNA binding-site prediction"]},"source_ids":["transbind-2026"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b4-023"}],"attributes":{"printed_value":"0.9508","numeric_value":"0.9508","metric":"AUROC","metric_direction":"unknown","unit":"fraction","uncertainty":null,"source_locator":"Table 2, TransBind row, AUROC column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:58.585Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness.","evidence":"Table 2, TransBind row, AUROC column; cell: 0.9508","artifact_sha256":"5d777f5925e941b7d087035d5d87e79ef75ae8d6456a770ffe8c527da566fee0","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC13145115/fullTextXML"},"legacy_id":"lit-b4-023","legacy_row":{"id":"lit-b4-023","paper_id":"transbind-2026","domain_id":"molecular-interactions","task":"transcription-factor DNA binding-site prediction","model":"TransBind","model_version":"","dataset":"genome-wide TF binding sites","dataset_version":"","split":"test","metric":"AUROC","value":"0.9508","unit":"fraction","uncertainty":"","protocol":"Integrates protein and DNA embeddings for TFBS prediction on paper test dataset","source_locator":"Table 2, TransBind row, AUROC column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC13145115/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:37:05Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"lit-b4-024","kind":"result","name":"ESM2_AMPS · AUROC · Bernett PPI dataset","description":"","status":"source_checked","facets":{"areas":["molecular-interactions"],"tasks":["protein-protein interaction prediction"]},"source_ids":["esm2-amp-2025"],"links":[{"relation":"evaluation","target_id":"evaluation-lit-b4-024"}],"attributes":{"printed_value":"0.68","numeric_value":"0.68","metric":"AUROC","metric_direction":"unknown","unit":"fraction","uncertainty":null,"source_locator":"Table 4, ESM2_AMPS row, AUROC column","review":{"method":"primary_xml_exact_label_cell_check","reviewer":"rewire deterministic table checker v1","reviewed_at":"2026-09-16T10:33:58.625Z","notes":"Exact row/header labels and numeric cell matched. Check verifies transcription, not experimental correctness.","evidence":"Table 4, ESM2_AMPS row, AUROC column; cell: 0.68","artifact_sha256":"8e7ad6efb72ca28d73037cdf465b0e62f99cd6d0ee4ca9eaf96a4c48da22fd6c","retrieval_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12392411/fullTextXML"},"legacy_id":"lit-b4-024","legacy_row":{"id":"lit-b4-024","paper_id":"esm2-amp-2025","domain_id":"molecular-interactions","task":"protein-protein interaction prediction","model":"ESM2_AMPS","model_version":"","dataset":"Bernett PPI dataset","dataset_version":"","split":"","metric":"AUROC","value":"0.68","unit":"fraction","uncertainty":"","protocol":"ESM2-derived embeddings plus paper interaction predictor","source_locator":"Table 4, ESM2_AMPS row, AUROC column","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12392411/","evaluation_origin":"author_reported","reviewed_utc":"2026-09-15T23:37:05Z"},"missing_metadata":{"model_version":"not_reported_in_legacy_extract","dataset_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","uncertainty":"not_reported_in_legacy_extract"}}}
{"id":"mdl4microbiome-2022","kind":"source","name":"Multimodal deep learning applied to classify healthy and disease states of human microbiome","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["microbes-communities"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC8763943/","version":"PMC archival version PMC8763943.1","retrieved_at":"2026-09-15T23:37:05Z","doi":"10.1038/s41598-022-04773-3","publication_status":"peer_reviewed","year":2022,"artifact_sha256":"72330eda245ac97c5de5d47a491bb86d9b8f527a03ae25fa41cae5bfb637143b","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC8763943/fullTextXML","artifact_retrieved_at":"2026-09-16T10:33:58.492Z","legacy_paper":{"id":"mdl4microbiome-2022","title":"Multimodal deep learning applied to classify healthy and disease states of human microbiome","year":2022,"publication_status":"peer_reviewed","version":"PMC archival version PMC8763943.1","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC8763943/","primary_domain":"microbes-communities","retrieved_utc":"2026-09-15T23:37:05Z","notes":"Primary full text verified via Europe PMC fullTextXML; venue: Scientific Reports; PMC ID: PMC8763943. Study predates most microbial foundation models; useful task baseline only.","doi":"10.1038/s41598-022-04773-3"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"megsite-2025","kind":"source","name":"MegSite: an accurate nucleic acid-binding residue prediction method based on multimodal protein language model","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["proteins-complexes"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12496013/","version":"journal full text in PMC","retrieved_at":"2026-09-15T23:33:26Z","doi":"10.1093/bib/bbaf524","publication_status":"peer_reviewed","year":2025,"artifact_sha256":"10d13122331813243d83b84fe6f9294eac7e7c03cde082ebed276191ac41089c","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12496013/fullTextXML","artifact_retrieved_at":"2026-09-16T10:38:57.558213+00:00","legacy_paper":{"id":"megsite-2025","title":"MegSite: an accurate nucleic acid-binding residue prediction method based on multimodal protein language model","year":2025,"publication_status":"peer_reviewed","version":"journal full text in PMC","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12496013/","primary_domain":"proteins-complexes","retrieved_utc":"2026-09-15T23:33:26Z","doi":"10.1093/bib/bbaf524","notes":"Numeric result checked against Table 2 in primary full-text XML; journal/source: Briefings in Bioinformatics."},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"metadata-correction-6ff08f49b3b45f847dd6","kind":"claim","name":"AlphaFold 3 Server: entity level","description":"Correct the original family classification: this record describes the hosted service, not a separate model family. Historical releases retain the original value.","status":"source_checked","facets":{"areas":["molecular-interactions"],"method_types":["foundation model"]},"source_ids":["evidence-alphafold-server-terms"],"links":[{"relation":"subject","target_id":"catalog-model-alphafold-3-server"}],"attributes":{"field":"attributes.entity_level","previous_value":"family","value":"service","source_locator":"Overview; hosted service identity","review":{"method":"automated_source_review","date":"2026-09-16","note":"Correct the original family classification: this record describes the hosted service, not a separate model family. Historical releases retain the original value."}}}
{"id":"metagenomic-pathogens-2025","kind":"source","name":"Enhancing pathogen identification through AI-assisted metagenomic sequencing","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["microbes-communities"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12493982/","version":"version of record","retrieved_at":"2026-09-15T23:37:05Z","doi":"10.3389/fmicb.2025.1634194","publication_status":"peer_reviewed","year":2025,"artifact_sha256":"aa88de1b0f0fd7ba1fedc1074bba9ba6ce199a0b723072a04d531db4585ae77c","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12493982/fullTextXML","artifact_retrieved_at":"2026-09-16T10:41:06Z","legacy_paper":{"id":"metagenomic-pathogens-2025","title":"Enhancing pathogen identification through AI-assisted metagenomic sequencing","year":2025,"publication_status":"peer_reviewed","version":"version of record","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12493982/","primary_domain":"microbes-communities","retrieved_utc":"2026-09-15T23:37:05Z","notes":"Primary full text verified via Europe PMC fullTextXML; venue: Frontiers in Microbiology; PMC ID: PMC12493982. Primary article has mixed biomedical-text and metagenomics assessments; selected MetaHIT pathogen-detection table only.","doi":"10.3389/fmicb.2025.1634194"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"},"evidence_concerns":[{"source_id":"metagenomic-pathogens-2025","message":"The paper reports a MetaHIT/iHMP sequencing experiment in Section 4.3 and Table 3, but that section does not identify exact samples, split manifests or construction of its taxonomic reference labels. The five-fold procedure in Section 4.2 is described for different text/image datasets and cannot safely be assigned to this result. The printed score is retained; protocol identity remains unresolved and the row is excluded from comparisons.","source_locator":"Section 4.2 Experimental details versus the additional metagenomic experiment in Section 4.3 and Table 3","artifact_sha256":"aa88de1b0f0fd7ba1fedc1074bba9ba6ce199a0b723072a04d531db4585ae77c","reviewed_at":"2026-09-16T20:07:20.757932+00:00","review_method":"automated_primary_source_review"}]}}
{"id":"molas-2026","kind":"source","name":"Molecular embedding-based algorithm selection in protein-ligand docking","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["molecular-interactions"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC13104262/","version":"PMC archival version PMC13104262.1","retrieved_at":"2026-09-15T23:29:32Z","doi":"10.1186/s13321-026-01168-8","publication_status":"peer_reviewed","year":2026,"artifact_sha256":"d556d47e0f7bbdc37eb62374b85ac9092dc7ff438fbae9e42892ac2cc784023f","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC13104262/fullTextXML","artifact_retrieved_at":"2026-09-16T10:44:03.432565+00:00","legacy_paper":{"id":"molas-2026","title":"Molecular embedding-based algorithm selection in protein-ligand docking","year":2026,"publication_status":"peer_reviewed","version":"PMC archival version PMC13104262.1","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC13104262/","primary_domain":"molecular-interactions","retrieved_utc":"2026-09-15T23:29:32Z","notes":"Primary full text verified using Europe PMC XML; venue: Journal of Cheminformatics; PMC ID: PMC13104262.","doi":"10.1186/s13321-026-01168-8"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"mouse-geneformer-2025","kind":"source","name":"Mouse-Geneformer: A deep learning model for mouse single-cell transcriptome and its cross-species utility","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["cells-tissues"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11964219/","version":"version of record","retrieved_at":"2026-09-15T23:29:32Z","doi":"10.1371/journal.pgen.1011420","publication_status":"peer_reviewed","year":2025,"artifact_sha256":"ef6c68f5c9b47c2f89598ddf647f05b72e4155e8b33609ebff23936a84bbd41d","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11964219/fullTextXML","artifact_retrieved_at":"2026-09-16T10:44:03.392488+00:00","legacy_paper":{"id":"mouse-geneformer-2025","title":"Mouse-Geneformer: A deep learning model for mouse single-cell transcriptome and its cross-species utility","year":2025,"publication_status":"peer_reviewed","version":"version of record","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11964219/","primary_domain":"cells-tissues","retrieved_utc":"2026-09-15T23:29:32Z","notes":"Primary full text verified using Europe PMC XML; venue: PLOS Genetics; PMC ID: PMC11964219.","doi":"10.1371/journal.pgen.1011420"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"mpro-pose-affinity-2025","kind":"source","name":"A Comparative Study of Deep Learning and Classical Modeling Approaches for Protein–Ligand Binding Pose and Affinity Prediction in Coronavirus Main Proteases","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["molecular-interactions"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12801289/","version":"version of record","retrieved_at":"2026-09-15T23:25:00Z","doi":"10.1021/acs.jcim.5c02481","publication_status":"peer_reviewed","year":2025,"artifact_sha256":"c356a1c65a0033e5ae18a05d4afab5495856c5b6869328ff49e13547a4801a57","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12801289/fullTextXML","artifact_retrieved_at":"2026-09-16T10:41:16.557756+00:00","legacy_paper":{"id":"mpro-pose-affinity-2025","title":"A Comparative Study of Deep Learning and Classical Modeling Approaches for Protein–Ligand Binding Pose and Affinity Prediction in Coronavirus Main Proteases","year":2025,"publication_status":"peer_reviewed","version":"version of record","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12801289/","primary_domain":"molecular-interactions","retrieved_utc":"2026-09-15T23:25:00Z","notes":"Primary full text via Europe PMC XML; venue: Journal of Chemical Information and Modeling; PMC ID: PMC12801289.","doi":"10.1021/acs.jcim.5c02481"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"mrna-lm-2025","kind":"source","name":"mRNA-LM: full-length integrated SLM for mRNA analysis","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["rna-transcriptomes"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11962594/","version":"journal full text in PMC","retrieved_at":"2026-09-15T23:33:26Z","doi":"10.1093/nar/gkaf044","publication_status":"peer_reviewed","year":2025,"artifact_sha256":"3a23de3c672ec162d13561c483f180a73b550d717256deffdc9099accec205fd","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11962594/fullTextXML","artifact_retrieved_at":"2026-09-16T10:38:57.558214+00:00","legacy_paper":{"id":"mrna-lm-2025","title":"mRNA-LM: full-length integrated SLM for mRNA analysis","year":2025,"publication_status":"peer_reviewed","version":"journal full text in PMC","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11962594/","primary_domain":"rna-transcriptomes","retrieved_utc":"2026-09-15T23:33:26Z","doi":"10.1093/nar/gkaf044","notes":"Numeric result checked against Table 1. in primary full-text XML; journal/source: Nucleic Acids Research."},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"mrna-protein-diversity-2026","kind":"source","name":"Generalizable deep-learning-based mRNA-protein interaction prediction strongly depends on protein diversity","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["rna-transcriptomes"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC13235059/","version":"version of record","retrieved_at":"2026-09-15T23:37:05Z","doi":"10.1186/s13321-026-01197-3","publication_status":"peer_reviewed","year":2026,"artifact_sha256":"94f9fe22a5f0e6c8619e4af994eb4f6ded7417efcf0c1240380269280f97f9d1","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC13235059/fullTextXML","artifact_retrieved_at":"2026-09-16T10:33:57.257Z","legacy_paper":{"id":"mrna-protein-diversity-2026","title":"Generalizable deep-learning-based mRNA-protein interaction prediction strongly depends on protein diversity","year":2026,"publication_status":"peer_reviewed","version":"version of record","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC13235059/","primary_domain":"rna-transcriptomes","retrieved_utc":"2026-09-15T23:37:05Z","notes":"Primary full text verified via Europe PMC fullTextXML; venue: Journal of Cheminformatics; PMC ID: PMC13235059. ProteinBERT encodes the protein side of an mRNA-protein task; score is not an RNA foundation-model result.","doi":"10.1186/s13321-026-01197-3"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"mrnabench-2025","kind":"source","name":"mRNABench: A curated benchmark for mature mRNA property and function prediction","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["rna-transcriptomes"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12265608/","version":"preprint archived 2025-07-08","retrieved_at":"2026-09-15T23:25:00Z","doi":"10.1101/2025.07.05.662870","publication_status":"preprint","year":2025,"artifact_sha256":"79f6264ee883535203c63a313547e7c57baa85585f76b42f8d899eb17fb7e600","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12265608/fullTextXML","artifact_retrieved_at":"2026-09-16T10:41:16.497221+00:00","legacy_paper":{"id":"mrnabench-2025","title":"mRNABench: A curated benchmark for mature mRNA property and function prediction","year":2025,"publication_status":"preprint","version":"preprint archived 2025-07-08","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12265608/","primary_domain":"rna-transcriptomes","retrieved_utc":"2026-09-15T23:25:00Z","notes":"Primary full text via Europe PMC XML; venue: bioRxiv; PMC ID: PMC12265608.","doi":"10.1101/2025.07.05.662870"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"mrnabert-2025","kind":"source","name":"mRNABERT: advancing mRNA sequence design with a universal language model and comprehensive dataset","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["rna-transcriptomes"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12644827/","version":"journal full text in PMC","retrieved_at":"2026-09-15T23:33:26Z","doi":"10.1038/s41467-025-65340-8","publication_status":"peer_reviewed","year":2025,"artifact_sha256":"ff08ba895b7080446c08a930548b48a0041ae990c222ebb07e6ba7dcaf48ad44","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12644827/fullTextXML","artifact_retrieved_at":"2026-09-16T10:38:57.558216+00:00","legacy_paper":{"id":"mrnabert-2025","title":"mRNABERT: advancing mRNA sequence design with a universal language model and comprehensive dataset","year":2025,"publication_status":"peer_reviewed","version":"journal full text in PMC","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12644827/","primary_domain":"rna-transcriptomes","retrieved_utc":"2026-09-15T23:33:26Z","doi":"10.1038/s41467-025-65340-8","notes":"Numeric result checked against Table 2 in primary full-text XML; journal/source: Nature Communications."},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"mulan-2025","kind":"source","name":"MULAN: multimodal protein language model for sequence and structure encoding","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["proteins-complexes"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12452268/","version":"journal full text in PMC","retrieved_at":"2026-09-15T23:33:26Z","doi":"10.1093/bioadv/vbaf117","publication_status":"peer_reviewed","year":2025,"artifact_sha256":"771a9a26ebda6f49ea266540e8dd6e6de0cbaef724de818ca6124a5f9c50d350","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12452268/fullTextXML","artifact_retrieved_at":"2026-09-16T10:38:57.558217+00:00","legacy_paper":{"id":"mulan-2025","title":"MULAN: multimodal protein language model for sequence and structure encoding","year":2025,"publication_status":"peer_reviewed","version":"journal full text in PMC","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12452268/","primary_domain":"proteins-complexes","retrieved_utc":"2026-09-15T23:33:26Z","doi":"10.1093/bioadv/vbaf117","notes":"Numeric result checked against Table 2. in primary full-text XML; journal/source: Bioinformatics Advances."},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"nabas-plus-2025","kind":"source","name":"Advancing metagenomic classification with NABAS+: a novel alignment-based approach","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["microbes-communities"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12231603/","version":"PMC12231603.1","retrieved_at":"2026-09-15T23:25:00Z","doi":"10.1093/nargab/lqaf092","publication_status":"peer_reviewed","year":2025,"artifact_sha256":"49903e751beb86f6744825d2fdb3ea2fbe327b52b1ce68c323bfa8b66dae71ec","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12231603/fullTextXML","artifact_retrieved_at":"2026-09-16T10:41:16.546107+00:00","legacy_paper":{"id":"nabas-plus-2025","title":"Advancing metagenomic classification with NABAS+: a novel alignment-based approach","year":2025,"publication_status":"peer_reviewed","version":"PMC12231603.1","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12231603/","primary_domain":"microbes-communities","retrieved_utc":"2026-09-15T23:25:00Z","notes":"Primary full text via Europe PMC XML; venue: NAR Genomics and Bioinformatics; PMC ID: PMC12231603.","doi":"10.1093/nargab/lqaf092"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"ncd-metagenomics-2026","kind":"source","name":"Normalized compression distance for DNA classification","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["microbes-communities"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12884959/","version":"version of record","retrieved_at":"2026-09-15T23:29:32Z","doi":"10.7717/peerj.20677","publication_status":"peer_reviewed","year":2026,"artifact_sha256":"10e9ba780c45e7787baff9b81ef7b45c014d7fe6c716d6c759e14d89a813dc1c","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12884959/fullTextXML","artifact_retrieved_at":"2026-09-16T10:44:03.414806+00:00","legacy_paper":{"id":"ncd-metagenomics-2026","title":"Normalized compression distance for DNA classification","year":2026,"publication_status":"peer_reviewed","version":"version of record","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12884959/","primary_domain":"microbes-communities","retrieved_utc":"2026-09-15T23:29:32Z","notes":"Primary full text verified using Europe PMC XML; venue: PeerJ; PMC ID: PMC12884959.","doi":"10.7717/peerj.20677"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"nmdn-2025","kind":"source","name":"Normalized Protein–Ligand Distance Likelihood Score for End-to-End Blind Docking and Virtual Screening","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["molecular-interactions"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11815853/","version":"version of record","retrieved_at":"2026-09-15T23:25:00Z","doi":"10.1021/acs.jcim.4c01014","publication_status":"peer_reviewed","year":2025,"artifact_sha256":"194b21478aaedd9a7384cabb8b0040ca5b6a4938f4d275627b86a6b787affc20","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11815853/fullTextXML","artifact_retrieved_at":"2026-09-16T10:41:16.552697+00:00","legacy_paper":{"id":"nmdn-2025","title":"Normalized Protein–Ligand Distance Likelihood Score for End-to-End Blind Docking and Virtual Screening","year":2025,"publication_status":"peer_reviewed","version":"version of record","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11815853/","primary_domain":"molecular-interactions","retrieved_utc":"2026-09-15T23:25:00Z","notes":"Primary full text via Europe PMC XML; venue: Journal of Chemical Information and Modeling; PMC ID: PMC11815853.","doi":"10.1021/acs.jcim.4c01014"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"pc-mer-2024","kind":"source","name":"PC-mer: An Ultra-fast memory-efficient tool for metagenomics profiling and classification","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["microbes-communities"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11293629/","version":"version of record","retrieved_at":"2026-09-15T23:37:05Z","doi":"10.1371/journal.pone.0307279","publication_status":"peer_reviewed","year":2024,"artifact_sha256":"0b0a225fa6f5f3ba41dffc7b320c91f47738301cf45835260eecaa03b704a097","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11293629/fullTextXML","artifact_retrieved_at":"2026-09-16T10:41:06Z","legacy_paper":{"id":"pc-mer-2024","title":"PC-mer: An Ultra-fast memory-efficient tool for metagenomics profiling and classification","year":2024,"publication_status":"peer_reviewed","version":"version of record","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11293629/","primary_domain":"microbes-communities","retrieved_utc":"2026-09-15T23:37:05Z","notes":"Primary full text verified via Europe PMC fullTextXML; venue: PLOS ONE; PMC ID: PMC11293629. Feature-extraction classifier; not a biological foundation model.","doi":"10.1371/journal.pone.0307279"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"phylogpn-2025","kind":"source","name":"A Phylogenetic Approach to Genomic Language Modeling","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["dna-genomes"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11908359/","version":"preprint version in PMC","retrieved_at":"2026-09-15T23:33:26Z","doi":null,"publication_status":"preprint","year":2025,"artifact_sha256":"807f3a26cbfa9b5ce238d92164bd523302c67d1c5794b08273c51cca1acd4224","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11908359/fullTextXML","artifact_retrieved_at":"2026-09-16T10:38:57.558218+00:00","legacy_paper":{"id":"phylogpn-2025","title":"A Phylogenetic Approach to Genomic Language Modeling","year":2025,"publication_status":"preprint","version":"preprint version in PMC","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11908359/","primary_domain":"dna-genomes","retrieved_utc":"2026-09-15T23:33:26Z","notes":"Numeric result checked against Table 1. in primary full-text XML; journal/source: ArXiv."},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"plantcad2-2025","kind":"source","name":"PlantCAD2: A Long-Context DNA Language Model for Cross-Species Functional Annotation in Angiosperms","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["dna-genomes"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12425018/","version":"preprint version in PMC","retrieved_at":"2026-09-15T23:37:05Z","doi":"10.1101/2025.08.27.672609","publication_status":"preprint","year":2025,"artifact_sha256":"4891955edb33af61e36dad51110574aa242e77c2df36429f100ff5a54bd8bd4b","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12425018/fullTextXML","artifact_retrieved_at":"2026-09-16T10:41:06Z","legacy_paper":{"id":"plantcad2-2025","title":"PlantCAD2: A Long-Context DNA Language Model for Cross-Species Functional Annotation in Angiosperms","year":2025,"publication_status":"preprint","version":"preprint version in PMC","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12425018/","primary_domain":"dna-genomes","retrieved_utc":"2026-09-15T23:37:05Z","notes":"Primary full text verified via Europe PMC fullTextXML; venue: bioRxiv; PMC ID: PMC12425018. Preprint; Table 1 pairs PlantCAD2 with an unnamed best benchmark; only PlantCAD2 value recorded.","doi":"10.1101/2025.08.27.672609"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"polya-glm-2025","kind":"source","name":"PolyA-GLM: A comprehensive framework for De novo polyadenylation site prediction using genome language models","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["dna-genomes"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12799945/","version":"journal full text in PMC","retrieved_at":"2026-09-15T23:33:26Z","doi":"10.1016/j.csbj.2025.12.011","publication_status":"peer_reviewed","year":2025,"artifact_sha256":"e9ebd53d88837ad8d457881ffee918d2734dcae87d3c5cd03135947b6cf5dbde","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12799945/fullTextXML","artifact_retrieved_at":"2026-09-16T10:38:57.558220+00:00","legacy_paper":{"id":"polya-glm-2025","title":"PolyA-GLM: A comprehensive framework for De novo polyadenylation site prediction using genome language models","year":2025,"publication_status":"peer_reviewed","version":"journal full text in PMC","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12799945/","primary_domain":"dna-genomes","retrieved_utc":"2026-09-15T23:33:26Z","doi":"10.1016/j.csbj.2025.12.011","notes":"Numeric result checked against Table 1 in primary full-text XML; journal/source: Computational and Structural Biotechnology Journal."},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"prime-2026","kind":"source","name":"PRIME: An evaluation framework for protein representation inference and generalization in viral mutation space","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["proteins-complexes"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC13425921/","version":"version of record","retrieved_at":"2026-09-15T23:25:00Z","doi":"10.1186/s12864-026-12976-5","publication_status":"peer_reviewed","year":2026,"artifact_sha256":"f6aac4c25dd93026f87ce9a2e327c95faf4c3014d7f9ae04bb11f208ce047971","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC13425921/fullTextXML","artifact_retrieved_at":"2026-09-16T10:41:16.525183+00:00","legacy_paper":{"id":"prime-2026","title":"PRIME: An evaluation framework for protein representation inference and generalization in viral mutation space","year":2026,"publication_status":"peer_reviewed","version":"version of record","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC13425921/","primary_domain":"proteins-complexes","retrieved_utc":"2026-09-15T23:25:00Z","notes":"Primary full text via Europe PMC XML; venue: BMC Genomics; PMC ID: PMC13425921.","doi":"10.1186/s12864-026-12976-5"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"profile-association-07f7e706eaf956ccf14c","kind":"claim","name":"AlphaFold 3 Server: uses model AlphaFold 3","description":"Hosted AlphaFold Server uses AlphaFold 3. This is a service-to-model relation, not checkpoint equivalence; results are not rolled up.","status":"source_checked","facets":{"areas":["molecular-interactions"],"method_types":["foundation model"]},"source_ids":["evidence-alphafold-server-terms"],"links":[{"relation":"subject","target_id":"catalog-model-alphafold-3-server"}],"attributes":{"field":"links:uses_model:discovery-model-alphafold-3","target_id":"discovery-model-alphafold-3","source_locator":"Overview and required AlphaFold 3 paper citation","review":{"method":"automated_source_review","date":"2026-09-16","note":"Hosted AlphaFold Server uses AlphaFold 3. This is a service-to-model relation, not checkpoint equivalence; results are not rolled up."}}}
{"id":"profile-association-0bf5ead3625ab59bbc49","kind":"claim","name":"Open Problems: evaluates task Batch integration","description":"Navigation relationship checked against the named primary-source section; not equivalence of protocol versions or independent evidence.","status":"source_checked","facets":{"areas":["single-cell"]},"source_ids":["src-discovery-openproblems-bio-openproblems"],"links":[{"relation":"subject","target_id":"discovery-benchmark-open-problems"}],"attributes":{"field":"links:evaluates_task:catalog-task-cell-batch-integration","target_id":"catalog-task-cell-batch-integration","source_locator":"Official linked benchmark directory, Batch Integration entry: https://openproblems.bio/benchmarks/; platform task membership, not protocol equivalence","review":{"method":"automated_source_review","date":"2026-09-16","note":"Navigation relationship checked against the named primary-source section; not equivalence of protocol versions or independent evidence."}}}
{"id":"profile-association-0d52555df24c3c1f90f6","kind":"claim","name":"RNA-FM: family RNA-FM","description":"Both records point to the same official project; the catalogue record has no independently identified checkpoint. Preserve both IDs and restrict shared explanation to family-level facts.","status":"source_checked","facets":{"areas":["rna-transcriptomes"],"method_types":["foundation model"]},"source_ids":["src-discovery-ml4bio-rna-fm"],"links":[{"relation":"subject","target_id":"catalog-model-rna-fm"}],"attributes":{"field":"links:family:discovery-model-rna-fm","target_id":"discovery-model-rna-fm","source_locator":"Official README project identity and the existing catalogue source URL","review":{"method":"automated_source_review","date":"2026-09-16","note":"Both records point to the same official project; the catalogue record has no independently identified checkpoint. Preserve both IDs and restrict shared explanation to family-level facts."}}}
{"id":"profile-association-0fb2930451b33229a2e0","kind":"claim","name":"GEARS: family GEARS","description":"Both records point to the same official project; the catalogue record has no independently identified checkpoint. Preserve both IDs and restrict shared explanation to family-level facts.","status":"source_checked","facets":{"areas":["cells-tissues"],"method_types":["specialist"]},"source_ids":["src-discovery-snap-stanford-gears"],"links":[{"relation":"subject","target_id":"catalog-model-gears"}],"attributes":{"field":"links:family:discovery-model-gears","target_id":"discovery-model-gears","source_locator":"Official README project identity and the existing catalogue source URL","review":{"method":"automated_source_review","date":"2026-09-16","note":"Both records point to the same official project; the catalogue record has no independently identified checkpoint. Preserve both IDs and restrict shared explanation to family-level facts."}}}
{"id":"profile-association-10c89dc6780906fcd514","kind":"claim","name":"SpliceAI: variant of SpliceAI","description":"The official pinned package version matches the named catalogue version. This does not equate every annotation or ensemble run configuration.","status":"source_checked","facets":{"areas":["dna-genomes"],"method_types":["specialist"]},"source_ids":["src-discovery-illumina-spliceai"],"links":[{"relation":"subject","target_id":"catalog-model-spliceai"}],"attributes":{"field":"links:variant_of:discovery-model-spliceai","target_id":"discovery-model-spliceai","source_locator":"setup.py line 9 at 03f42437aaf56dc5dfd822c4ccee5aec1a705079 declares version 1.3.1; README identifies SpliceAI","review":{"method":"automated_source_review","date":"2026-09-16","note":"The official pinned package version matches the named catalogue version. This does not equate every annotation or ensemble run configuration."}}}
{"id":"profile-association-10ecbaabf760dbba79c8","kind":"claim","name":"GlycanML taxonomy prediction: part of GlycanML","description":"Navigation relationship checked against the named primary-source section; not equivalence of protocol versions or independent evidence.","status":"source_checked","facets":{"areas":["glycomics"]},"source_ids":["src-discovery-glycanml-glycanml"],"links":[{"relation":"subject","target_id":"discovery-benchmark-glycanml-taxonomy-prediction"}],"attributes":{"field":"links:part_of:discovery-benchmark-glycanml","target_id":"discovery-benchmark-glycanml","source_locator":"README: Overview, task list","review":{"method":"automated_source_review","date":"2026-09-16","note":"Navigation relationship checked against the named primary-source section; not equivalence of protocol versions or independent evidence."}}}
{"id":"profile-association-1d86b58d85bc57cbe366","kind":"claim","name":"mRNA-FM: variant of RNA-FM","description":"Verified named released configuration or extension within the official project. This is not alias equivalence and does not move existing result identities.","status":"source_checked","facets":{"areas":["rna-transcriptomes"],"method_types":["foundation model"]},"source_ids":["src-discovery-ml4bio-rna-fm"],"links":[{"relation":"subject","target_id":"catalog-model-mrna-fm"}],"attributes":{"field":"links:variant_of:discovery-model-rna-fm","target_id":"discovery-model-rna-fm","source_locator":"README.md: introduction identifies mRNA-FM as the coding-sequence extension of RNA-FM","review":{"method":"automated_source_review","date":"2026-09-16","note":"Verified named released configuration or extension within the official project. This is not alias equivalence and does not move existing result identities."}}}
{"id":"profile-association-1e3a9f8e48d3e9042bca","kind":"claim","name":"ESMFold: variant of ESMFold","description":"Verified named released configuration or extension within the official project. This is not alias equivalence and does not move existing result identities.","status":"source_checked","facets":{"areas":["proteins-complexes"],"method_types":["foundation model"]},"source_ids":["src-discovery-facebookresearch-esm"],"links":[{"relation":"subject","target_id":"catalog-model-esmfold"}],"attributes":{"field":"links:variant_of:discovery-model-esmfold","target_id":"discovery-model-esmfold","source_locator":"README.md: ESMFold Structure Prediction distinguishes v0 and v1","review":{"method":"automated_source_review","date":"2026-09-16","note":"Verified named released configuration or extension within the official project. This is not alias equivalence and does not move existing result identities."}}}
{"id":"profile-association-2fb39a063eed368e689b","kind":"claim","name":"Boltz-2: variant of Boltz","description":"Verified named released configuration or extension within the official project. This is not alias equivalence and does not move existing result identities.","status":"source_checked","facets":{"areas":["molecular-interactions"],"method_types":["foundation model"]},"source_ids":["src-discovery-jwohlwend-boltz"],"links":[{"relation":"subject","target_id":"catalog-model-boltz-2"}],"attributes":{"field":"links:variant_of:discovery-model-boltz","target_id":"discovery-model-boltz","source_locator":"README.md: Introduction distinguishes Boltz-1 and Boltz-2 in the model family","review":{"method":"automated_source_review","date":"2026-09-16","note":"Verified named released configuration or extension within the official project. This is not alias equivalence and does not move existing result identities."}}}
{"id":"profile-association-3242cc83bee01ceb529e","kind":"claim","name":"GlycanML protein-glycan interaction prediction: part of GlycanML","description":"Navigation relationship checked against the named primary-source section; not equivalence of protocol versions or independent evidence.","status":"source_checked","facets":{"areas":["glycomics"]},"source_ids":["src-discovery-glycanml-glycanml"],"links":[{"relation":"subject","target_id":"discovery-benchmark-glycanml-protein-glycan-interaction-prediction"}],"attributes":{"field":"links:part_of:discovery-benchmark-glycanml","target_id":"discovery-benchmark-glycanml","source_locator":"README: Overview, task list","review":{"method":"automated_source_review","date":"2026-09-16","note":"Navigation relationship checked against the named primary-source section; not equivalence of protocol versions or independent evidence."}}}
{"id":"profile-association-3fc720d75e5875470593","kind":"claim","name":"SpliceAI 1.3.1: uses model SpliceAI","description":"The evaluated MFASS pipeline explicitly uses the named base model. Adaptation, annotation and masking remain properties of the pipeline, not the family.","status":"source_checked","facets":{"areas":["dna-genomes"]},"source_ids":["rewire-mfass-v2-source","src-discovery-illumina-spliceai"],"links":[{"relation":"subject","target_id":"rewire-model-spliceai-1-3-1"}],"attributes":{"field":"links:uses_model:discovery-model-spliceai","target_id":"discovery-model-spliceai","source_locator":"Pinned rewire benchmarks/mfass/README.md and corresponding results JSON, config; official model README","review":{"method":"automated_source_review","date":"2026-09-16","note":"The evaluated MFASS pipeline explicitly uses the named base model. Adaptation, annotation and masking remain properties of the pipeline, not the family."}}}
{"id":"profile-association-49f1b5328fff0bc440af","kind":"claim","name":"MFASS v2: evaluates task MFASS splice-variant prioritisation","description":"Navigation relationship checked against the named primary-source section; not equivalence of protocol versions or independent evidence.","status":"source_checked","facets":{"areas":["dna-genomes"]},"source_ids":["rewire-mfass-v2-source"],"links":[{"relation":"subject","target_id":"rewire-mfass-v2"}],"attributes":{"field":"links:evaluates_task:catalog-task-mfass-splice","target_id":"catalog-task-mfass-splice","source_locator":"Pinned benchmarks/mfass/README.md: Dataset; Cohort reconciliation; Split","review":{"method":"automated_source_review","date":"2026-09-16","note":"Navigation relationship checked against the named primary-source section; not equivalence of protocol versions or independent evidence."}}}
{"id":"profile-association-4aeaae5057842d41e084","kind":"claim","name":"Pangolin · mask=False: uses model Pangolin","description":"The evaluated MFASS pipeline explicitly uses the named base model. Adaptation, annotation and masking remain properties of the pipeline, not the family.","status":"source_checked","facets":{"areas":["dna-genomes"]},"source_ids":["rewire-mfass-v2-source","src-discovery-tkzeng-pangolin"],"links":[{"relation":"subject","target_id":"rewire-model-pangolin-maskfalse"}],"attributes":{"field":"links:uses_model:discovery-model-pangolin","target_id":"discovery-model-pangolin","source_locator":"Pinned rewire benchmarks/mfass/README.md and corresponding results JSON, config; official model README","review":{"method":"automated_source_review","date":"2026-09-16","note":"The evaluated MFASS pipeline explicitly uses the named base model. Adaptation, annotation and masking remain properties of the pipeline, not the family."}}}
{"id":"profile-association-507834867ec7cde103f9","kind":"claim","name":"scGPT: variant of scGPT","description":"Verified named released configuration or extension within the official project. This is not alias equivalence and does not move existing result identities.","status":"source_checked","facets":{"areas":["cells-tissues"],"method_types":["foundation model"]},"source_ids":["src-discovery-bowang-lab-scgpt"],"links":[{"relation":"subject","target_id":"catalog-model-scgpt"}],"attributes":{"field":"links:variant_of:discovery-model-scgpt","target_id":"discovery-model-scgpt","source_locator":"README.md: Pretrained scGPT checkpoints, whole-human row","review":{"method":"automated_source_review","date":"2026-09-16","note":"Verified named released configuration or extension within the official project. This is not alias equivalence and does not move existing result identities."}}}
{"id":"profile-association-524d4e1d2f3954f416b3","kind":"claim","name":"GlycanML glycosylation type prediction: part of GlycanML","description":"Navigation relationship checked against the named primary-source section; not equivalence of protocol versions or independent evidence.","status":"source_checked","facets":{"areas":["glycomics"]},"source_ids":["src-discovery-glycanml-glycanml"],"links":[{"relation":"subject","target_id":"discovery-benchmark-glycanml-glycosylation-type-prediction"}],"attributes":{"field":"links:part_of:discovery-benchmark-glycanml","target_id":"discovery-benchmark-glycanml","source_locator":"README: Overview, task list","review":{"method":"automated_source_review","date":"2026-09-16","note":"Navigation relationship checked against the named primary-source section; not equivalence of protocol versions or independent evidence."}}}
{"id":"profile-association-56fc7ee6772f13885b95","kind":"claim","name":"DNABERT-2 117M · frozen pair embeddings: uses model DNABERT-2","description":"The evaluated MFASS pipeline explicitly uses the named base model. Adaptation, annotation and masking remain properties of the pipeline, not the family.","status":"source_checked","facets":{"areas":["dna-genomes"]},"source_ids":["rewire-mfass-v2-source","src-discovery-magics-lab-dnabert-2"],"links":[{"relation":"subject","target_id":"rewire-model-dnabert2-117m-frozen-pair-logreg"}],"attributes":{"field":"links:uses_model:discovery-model-dnabert-2","target_id":"discovery-model-dnabert-2","source_locator":"Pinned rewire benchmarks/mfass/README.md and corresponding results JSON, config; official model README","review":{"method":"automated_source_review","date":"2026-09-16","note":"The evaluated MFASS pipeline explicitly uses the named base model. Adaptation, annotation and masking remain properties of the pipeline, not the family."}}}
{"id":"profile-association-5dc7c010a32600e0fbd1","kind":"claim","name":"CAMI metagenome assembly: part of CAMI","description":"Navigation relationship checked against the named primary-source section; not equivalence of protocol versions or independent evidence.","status":"source_checked","facets":{"areas":["microbiome"]},"source_ids":["src-discovery-cami"],"links":[{"relation":"subject","target_id":"discovery-benchmark-cami-metagenome-assembly"}],"attributes":{"field":"links:part_of:discovery-benchmark-cami","target_id":"discovery-benchmark-cami","source_locator":"CAMI official home page: Summary / Per category","review":{"method":"automated_source_review","date":"2026-09-16","note":"Navigation relationship checked against the named primary-source section; not equivalence of protocol versions or independent evidence."}}}
{"id":"profile-association-6f91ef59fe79ba4e85e8","kind":"claim","name":"ProteinMPNN: variant of ProteinMPNN","description":"Verified named released configuration or extension within the official project. This is not alias equivalence and does not move existing result identities.","status":"source_checked","facets":{"areas":["proteins-complexes"],"method_types":["specialist"]},"source_ids":["src-discovery-dauparas-proteinmpnn"],"links":[{"relation":"subject","target_id":"catalog-model-proteinmpnn"}],"attributes":{"field":"links:variant_of:discovery-model-proteinmpnn","target_id":"discovery-model-proteinmpnn","source_locator":"README.md: full protein backbone model list includes v_48_020","review":{"method":"automated_source_review","date":"2026-09-16","note":"Verified named released configuration or extension within the official project. This is not alias equivalence and does not move existing result identities."}}}
{"id":"profile-association-7248dbde8b4a47c0358d","kind":"claim","name":"TAPE Contact Prediction: part of TAPE","description":"Navigation relationship checked against the named primary-source section; not equivalence of protocol versions or independent evidence.","status":"source_checked","facets":{"areas":["protein-function"]},"source_ids":["src-discovery-songlab-cal-tape"],"links":[{"relation":"subject","target_id":"discovery-benchmark-tape-contact-prediction"}],"attributes":{"field":"links:part_of:discovery-benchmark-tape","target_id":"discovery-benchmark-tape","source_locator":"README: List of Models and Tasks; Leaderboard","review":{"method":"automated_source_review","date":"2026-09-16","note":"Navigation relationship checked against the named primary-source section; not equivalence of protocol versions or independent evidence."}}}
{"id":"profile-association-7620c8aef622f0317b76","kind":"claim","name":"MassSpecGym Molecule retrieval: part of MassSpecGym","description":"Navigation relationship checked against the named primary-source section; not equivalence of protocol versions or independent evidence.","status":"source_checked","facets":{"areas":["metabolomics"]},"source_ids":["src-discovery-pluskal-lab-massspecgym"],"links":[{"relation":"subject","target_id":"discovery-benchmark-massspecgym-molecule-retrieval"}],"attributes":{"field":"links:part_of:discovery-benchmark-massspecgym","target_id":"discovery-benchmark-massspecgym","source_locator":"README: opening challenge list","review":{"method":"automated_source_review","date":"2026-09-16","note":"Navigation relationship checked against the named primary-source section; not equivalence of protocol versions or independent evidence."}}}
{"id":"profile-association-874580c7cba4c198c076","kind":"claim","name":"scIB: evaluates task Batch integration","description":"Navigation relationship checked against the named primary-source section; not equivalence of protocol versions or independent evidence.","status":"source_checked","facets":{"areas":["single-cell"]},"source_ids":["src-discovery-theislab-scib"],"links":[{"relation":"subject","target_id":"discovery-benchmark-scib"}],"attributes":{"field":"links:evaluates_task:catalog-task-cell-batch-integration","target_id":"catalog-task-cell-batch-integration","source_locator":"README: Metrics, Biological Conservation and Batch Correction","review":{"method":"automated_source_review","date":"2026-09-16","note":"Navigation relationship checked against the named primary-source section; not equivalence of protocol versions or independent evidence."}}}
{"id":"profile-association-923959d8f16ef46a494d","kind":"claim","name":"TAPE Secondary Structure: part of TAPE","description":"Navigation relationship checked against the named primary-source section; not equivalence of protocol versions or independent evidence.","status":"source_checked","facets":{"areas":["protein-function"]},"source_ids":["src-discovery-songlab-cal-tape"],"links":[{"relation":"subject","target_id":"discovery-benchmark-tape-secondary-structure"}],"attributes":{"field":"links:part_of:discovery-benchmark-tape","target_id":"discovery-benchmark-tape","source_locator":"README: List of Models and Tasks; Leaderboard","review":{"method":"automated_source_review","date":"2026-09-16","note":"Navigation relationship checked against the named primary-source section; not equivalence of protocol versions or independent evidence."}}}
{"id":"profile-association-9c3fcc0e37012e7dcbbd","kind":"claim","name":"MassSpecGym De novo molecule generation: part of MassSpecGym","description":"Navigation relationship checked against the named primary-source section; not equivalence of protocol versions or independent evidence.","status":"source_checked","facets":{"areas":["metabolomics"]},"source_ids":["src-discovery-pluskal-lab-massspecgym"],"links":[{"relation":"subject","target_id":"discovery-benchmark-massspecgym-de-novo-molecule-generation"}],"attributes":{"field":"links:part_of:discovery-benchmark-massspecgym","target_id":"discovery-benchmark-massspecgym","source_locator":"README: opening challenge list","review":{"method":"automated_source_review","date":"2026-09-16","note":"Navigation relationship checked against the named primary-source section; not equivalence of protocol versions or independent evidence."}}}
{"id":"profile-association-9ee0416aed06f5b6aa75","kind":"claim","name":"TAPE Remote Homology Detection: part of TAPE","description":"Navigation relationship checked against the named primary-source section; not equivalence of protocol versions or independent evidence.","status":"source_checked","facets":{"areas":["protein-function"]},"source_ids":["src-discovery-songlab-cal-tape"],"links":[{"relation":"subject","target_id":"discovery-benchmark-tape-remote-homology-detection"}],"attributes":{"field":"links:part_of:discovery-benchmark-tape","target_id":"discovery-benchmark-tape","source_locator":"README: List of Models and Tasks; Leaderboard","review":{"method":"automated_source_review","date":"2026-09-16","note":"Navigation relationship checked against the named primary-source section; not equivalence of protocol versions or independent evidence."}}}
{"id":"profile-association-aa25f55c703d596f37e1","kind":"claim","name":"GlycanML immunogenicity prediction: part of GlycanML","description":"Navigation relationship checked against the named primary-source section; not equivalence of protocol versions or independent evidence.","status":"source_checked","facets":{"areas":["glycomics"]},"source_ids":["src-discovery-glycanml-glycanml"],"links":[{"relation":"subject","target_id":"discovery-benchmark-glycanml-immunogenicity-prediction"}],"attributes":{"field":"links:part_of:discovery-benchmark-glycanml","target_id":"discovery-benchmark-glycanml","source_locator":"README: Overview, task list","review":{"method":"automated_source_review","date":"2026-09-16","note":"Navigation relationship checked against the named primary-source section; not equivalence of protocol versions or independent evidence."}}}
{"id":"profile-association-aaccdc5a8a938e70518e","kind":"claim","name":"MFASS v1 (superseded): evaluates task MFASS splice-variant prioritisation","description":"Navigation relationship checked against the named primary-source section; not equivalence of protocol versions or independent evidence.","status":"source_checked","facets":{"areas":["dna-genomes"]},"source_ids":["rewire-mfass-v2-source"],"links":[{"relation":"subject","target_id":"rewire-mfass-v1"}],"attributes":{"field":"links:evaluates_task:catalog-task-mfass-splice","target_id":"catalog-task-mfass-splice","source_locator":"Pinned benchmarks/mfass/README.md: Correction; v1 archived as superseded evidence","review":{"method":"automated_source_review","date":"2026-09-16","note":"Navigation relationship checked against the named primary-source section; not equivalence of protocol versions or independent evidence."}}}
{"id":"profile-association-b77dc7ed3e4c0e21c842","kind":"claim","name":"ESM-2: variant of ESM-2","description":"Verified named released configuration or extension within the official project. This is not alias equivalence and does not move existing result identities.","status":"source_checked","facets":{"areas":["proteins-complexes"],"method_types":["foundation model"]},"source_ids":["src-discovery-facebookresearch-esm"],"links":[{"relation":"subject","target_id":"catalog-model-esm-2"}],"attributes":{"field":"links:variant_of:discovery-model-esm-2","target_id":"discovery-model-esm-2","source_locator":"README.md: Pre-trained Models identifies esm2_t6_8M_UR50D","review":{"method":"automated_source_review","date":"2026-09-16","note":"Verified named released configuration or extension within the official project. This is not alias equivalence and does not move existing result identities."}}}
{"id":"profile-association-b98e7ae409d7f83b469e","kind":"claim","name":"DNABERT-2: variant of DNABERT-2","description":"Verified named released configuration or extension within the official project. This is not alias equivalence and does not move existing result identities.","status":"source_checked","facets":{"areas":["dna-genomes"],"method_types":["foundation model"]},"source_ids":["src-discovery-magics-lab-dnabert-2"],"links":[{"relation":"subject","target_id":"catalog-model-dnabert-2"}],"attributes":{"field":"links:variant_of:discovery-model-dnabert-2","target_id":"discovery-model-dnabert-2","source_locator":"README.md: Model and Data names the 117M checkpoint","review":{"method":"automated_source_review","date":"2026-09-16","note":"Verified named released configuration or extension within the official project. This is not alias equivalence and does not move existing result identities."}}}
{"id":"profile-association-bff7c057c678296de10c","kind":"claim","name":"CAMI genome binning: part of CAMI","description":"Navigation relationship checked against the named primary-source section; not equivalence of protocol versions or independent evidence.","status":"source_checked","facets":{"areas":["microbiome"]},"source_ids":["src-discovery-cami"],"links":[{"relation":"subject","target_id":"discovery-benchmark-cami-genome-binning"}],"attributes":{"field":"links:part_of:discovery-benchmark-cami","target_id":"discovery-benchmark-cami","source_locator":"CAMI official home page: Summary / Per category","review":{"method":"automated_source_review","date":"2026-09-16","note":"Navigation relationship checked against the named primary-source section; not equivalence of protocol versions or independent evidence."}}}
{"id":"profile-association-ceb267609f7dffa0b950","kind":"claim","name":"MassSpecGym Spectrum simulation: part of MassSpecGym","description":"Navigation relationship checked against the named primary-source section; not equivalence of protocol versions or independent evidence.","status":"source_checked","facets":{"areas":["metabolomics"]},"source_ids":["src-discovery-pluskal-lab-massspecgym"],"links":[{"relation":"subject","target_id":"discovery-benchmark-massspecgym-spectrum-simulation"}],"attributes":{"field":"links:part_of:discovery-benchmark-massspecgym","target_id":"discovery-benchmark-massspecgym","source_locator":"README: opening challenge list","review":{"method":"automated_source_review","date":"2026-09-16","note":"Navigation relationship checked against the named primary-source section; not equivalence of protocol versions or independent evidence."}}}
{"id":"profile-association-d69e83003c6ac3369c10","kind":"claim","name":"TAPE Stability: part of TAPE","description":"Navigation relationship checked against the named primary-source section; not equivalence of protocol versions or independent evidence.","status":"source_checked","facets":{"areas":["protein-function"]},"source_ids":["src-discovery-songlab-cal-tape"],"links":[{"relation":"subject","target_id":"discovery-benchmark-tape-stability"}],"attributes":{"field":"links:part_of:discovery-benchmark-tape","target_id":"discovery-benchmark-tape","source_locator":"README: List of Models and Tasks; Leaderboard","review":{"method":"automated_source_review","date":"2026-09-16","note":"Navigation relationship checked against the named primary-source section; not equivalence of protocol versions or independent evidence."}}}
{"id":"profile-association-dcbdd6f96b71ea87cb77","kind":"claim","name":"TAPE Fluorescence: part of TAPE","description":"Navigation relationship checked against the named primary-source section; not equivalence of protocol versions or independent evidence.","status":"source_checked","facets":{"areas":["protein-function"]},"source_ids":["src-discovery-songlab-cal-tape"],"links":[{"relation":"subject","target_id":"discovery-benchmark-tape-fluorescence"}],"attributes":{"field":"links:part_of:discovery-benchmark-tape","target_id":"discovery-benchmark-tape","source_locator":"README: List of Models and Tasks; Leaderboard","review":{"method":"automated_source_review","date":"2026-09-16","note":"Navigation relationship checked against the named primary-source section; not equivalence of protocol versions or independent evidence."}}}
{"id":"profile-association-dd1df650aeabc9d220d1","kind":"claim","name":"CAMI taxonomic profiling: part of CAMI","description":"Navigation relationship checked against the named primary-source section; not equivalence of protocol versions or independent evidence.","status":"source_checked","facets":{"areas":["microbiome"]},"source_ids":["src-discovery-cami"],"links":[{"relation":"subject","target_id":"discovery-benchmark-cami-taxonomic-profiling"}],"attributes":{"field":"links:part_of:discovery-benchmark-cami","target_id":"discovery-benchmark-cami","source_locator":"CAMI official home page: Summary / Per category","review":{"method":"automated_source_review","date":"2026-09-16","note":"Navigation relationship checked against the named primary-source section; not equivalence of protocol versions or independent evidence."}}}
{"id":"profile-association-df9d68d965eef2930ace","kind":"claim","name":"ProteinGym: evaluates task ProteinGym mutation effects","description":"Navigation relationship checked against the named primary-source section; not equivalence of protocol versions or independent evidence.","status":"source_checked","facets":{"areas":["protein-function"]},"source_ids":["src-discovery-oatml-markslab-proteingym"],"links":[{"relation":"subject","target_id":"discovery-benchmark-proteingym"}],"attributes":{"field":"links:evaluates_task:catalog-task-proteingym-effects","target_id":"catalog-task-proteingym-effects","source_locator":"README: Overview; Results","review":{"method":"automated_source_review","date":"2026-09-16","note":"Navigation relationship checked against the named primary-source section; not equivalence of protocol versions or independent evidence."}}}
{"id":"profile-association-e433e3e33937f0e3c53a","kind":"claim","name":"CAMI taxonomic binning: part of CAMI","description":"Navigation relationship checked against the named primary-source section; not equivalence of protocol versions or independent evidence.","status":"source_checked","facets":{"areas":["microbiome"]},"source_ids":["src-discovery-cami"],"links":[{"relation":"subject","target_id":"discovery-benchmark-cami-taxonomic-binning"}],"attributes":{"field":"links:part_of:discovery-benchmark-cami","target_id":"discovery-benchmark-cami","source_locator":"CAMI official home page: Summary / Per category","review":{"method":"automated_source_review","date":"2026-09-16","note":"Navigation relationship checked against the named primary-source section; not equivalence of protocol versions or independent evidence."}}}
{"id":"profile-association-e63a9e1f6beae6543100","kind":"claim","name":"Pangolin: family Pangolin","description":"Both records point to the same official project; the catalogue record has no independently identified checkpoint. Preserve both IDs and restrict shared explanation to family-level facts.","status":"source_checked","facets":{"areas":["dna-genomes"],"method_types":["specialist"]},"source_ids":["src-discovery-tkzeng-pangolin"],"links":[{"relation":"subject","target_id":"catalog-model-pangolin"}],"attributes":{"field":"links:family:discovery-model-pangolin","target_id":"discovery-model-pangolin","source_locator":"Official README project identity and the existing catalogue source URL","review":{"method":"automated_source_review","date":"2026-09-16","note":"Both records point to the same official project; the catalogue record has no independently identified checkpoint. Preserve both IDs and restrict shared explanation to family-level facts."}}}
{"id":"prokbert-2024","kind":"source","name":"ProkBERT family: genomic language models for microbiome applications","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["microbes-communities"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC10810988/","version":"PMC10810988.1","retrieved_at":"2026-09-15T23:25:00Z","doi":"10.3389/fmicb.2023.1331233","publication_status":"peer_reviewed","year":2024,"artifact_sha256":"8610e2a54aa877c8dc565a9cdb6e82099f284c5e0907a52cab18d994ea732436","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC10810988/fullTextXML","artifact_retrieved_at":"2026-09-16T10:33:36.197Z","legacy_paper":{"id":"prokbert-2024","title":"ProkBERT family: genomic language models for microbiome applications","year":2024,"publication_status":"peer_reviewed","version":"PMC10810988.1","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC10810988/","primary_domain":"microbes-communities","retrieved_utc":"2026-09-15T23:25:00Z","notes":"Primary full text via Europe PMC XML; venue: Frontiers in Microbiology; PMC ID: PMC10810988.","doi":"10.3389/fmicb.2023.1331233"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"protein-binding-sites-2023","kind":"source","name":"Learning the protein language of proteome-wide protein-protein binding sites via explainable ensemble deep learning","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["proteins-complexes"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC9849350/","version":"version of record","retrieved_at":"2026-09-15T23:37:05Z","doi":"10.1038/s42003-023-04462-5","publication_status":"peer_reviewed","year":2023,"artifact_sha256":"491711aa7186e74bf33f6d601c4ea6a8f565e770938fe1f91a0cd347b8f06f3f","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC9849350/fullTextXML","artifact_retrieved_at":"2026-09-16T10:41:06Z","legacy_paper":{"id":"protein-binding-sites-2023","title":"Learning the protein language of proteome-wide protein-protein binding sites via explainable ensemble deep learning","year":2023,"publication_status":"peer_reviewed","version":"version of record","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC9849350/","primary_domain":"proteins-complexes","retrieved_utc":"2026-09-15T23:37:05Z","notes":"Primary full text verified via Europe PMC fullTextXML; venue: Communications Biology; PMC ID: PMC9849350. Downstream binding-site classifier; not a native ProtT5 prediction head.","doi":"10.1038/s42003-023-04462-5"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"proteingym-2023","kind":"source","name":"ProteinGym: Large-Scale Benchmarks for Protein Design and Fitness Prediction","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["proteins-complexes"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC10723403/","version":"PMC10723403.1","retrieved_at":"2026-09-15T23:25:00Z","doi":"10.1101/2023.12.07.570727","publication_status":"preprint","year":2023,"artifact_sha256":"4519641f13271bdd09b166e7d93232f22542489bc52a25b5a1628c3df8badce1","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC10723403/fullTextXML","artifact_retrieved_at":"2026-09-16T10:41:16.517323+00:00","legacy_paper":{"id":"proteingym-2023","title":"ProteinGym: Large-Scale Benchmarks for Protein Design and Fitness Prediction","year":2023,"publication_status":"preprint","version":"PMC10723403.1","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC10723403/","primary_domain":"proteins-complexes","retrieved_utc":"2026-09-15T23:25:00Z","notes":"Primary full text via Europe PMC XML; venue: bioRxiv; PMC ID: PMC10723403.","doi":"10.1101/2023.12.07.570727"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"pst-2025","kind":"source","name":"Endowing protein language models with structural knowledge","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["proteins-complexes"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12603367/","version":"PMC12603367.1","retrieved_at":"2026-09-15T23:25:00Z","doi":"10.1093/bioinformatics/btaf582","publication_status":"peer_reviewed","year":2025,"artifact_sha256":"c21ad593de589a7188ca86a8b7ce617e301d48dd939ecc7da03efb22cbe8d7a3","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12603367/fullTextXML","artifact_retrieved_at":"2026-09-16T10:41:16.527633+00:00","legacy_paper":{"id":"pst-2025","title":"Endowing protein language models with structural knowledge","year":2025,"publication_status":"peer_reviewed","version":"PMC12603367.1","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12603367/","primary_domain":"proteins-complexes","retrieved_utc":"2026-09-15T23:25:00Z","notes":"Primary full text via Europe PMC XML; venue: Bioinformatics; PMC ID: PMC12603367.","doi":"10.1093/bioinformatics/btaf582"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"quadruplex-llm-benchmark-2025","kind":"source","name":"Benchmarking DNA large language models on quadruplexes","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["dna-genomes"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11953744/","version":"version of record","retrieved_at":"2026-09-15T23:25:00Z","doi":"10.1016/j.csbj.2025.03.007","publication_status":"peer_reviewed","year":2025,"artifact_sha256":"c3d7c6d068d3c11a9c8255a932197ece3804d94e8b2d4f858bea373a1b6eb32f","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11953744/fullTextXML","artifact_retrieved_at":"2026-09-16T10:33:35.379Z","legacy_paper":{"id":"quadruplex-llm-benchmark-2025","title":"Benchmarking DNA large language models on quadruplexes","year":2025,"publication_status":"peer_reviewed","version":"version of record","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11953744/","primary_domain":"dna-genomes","retrieved_utc":"2026-09-15T23:25:00Z","notes":"Primary full text via Europe PMC XML; venue: Computational and Structural Biotechnology Journal; PMC ID: PMC11953744.","doi":"10.1016/j.csbj.2025.03.007"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"r3design-2025","kind":"source","name":"R3Design: deep tertiary structure-based RNA sequence design and beyond","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["rna-transcriptomes"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11685104/","version":"PMC archival version PMC11685104.1","retrieved_at":"2026-09-15T23:37:05Z","doi":"10.1093/bib/bbae682","publication_status":"peer_reviewed","year":2025,"artifact_sha256":"b49c9ad846e46b11b240aace8e6bfaf953b842df166b69aee4843c02e9349779","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11685104/fullTextXML","artifact_retrieved_at":"2026-09-16T10:41:06Z","legacy_paper":{"id":"r3design-2025","title":"R3Design: deep tertiary structure-based RNA sequence design and beyond","year":2025,"publication_status":"peer_reviewed","version":"PMC archival version PMC11685104.1","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11685104/","primary_domain":"rna-transcriptomes","retrieved_utc":"2026-09-15T23:37:05Z","notes":"Primary full text verified via Europe PMC fullTextXML; venue: Briefings in Bioinformatics; PMC ID: PMC11685104. Architecture and task differ from RNA language-model encoding benchmarks.","doi":"10.1093/bib/bbae682"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-00201f65c32f6d","kind":"dataset","name":"α-synuclein Ligand 47 MD ensemble","description":"","status":"needs_review","facets":{"areas":["molecular-interactions"]},"source_ids":["ensemble-idp-docking-2025"],"links":[],"attributes":{"version":null,"split":null,"missing_metadata":{"version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-034c60a2dabc73","kind":"dataset","name":"genome-wide TF binding sites","description":"","status":"needs_review","facets":{"areas":["molecular-interactions"]},"source_ids":["transbind-2026"],"links":[],"attributes":{"version":null,"split":"test","missing_metadata":{"version":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-065b9fcc8da573","kind":"dataset","name":"CASF-2016 blind docked poses","description":"","status":"needs_review","facets":{"areas":["molecular-interactions"]},"source_ids":["nmdn-2025"],"links":[],"attributes":{"version":null,"split":null,"missing_metadata":{"version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-07d355c146be1f","kind":"dataset","name":"paper PPI test set","description":"","status":"needs_review","facets":{"areas":["proteins-complexes"]},"source_ids":["gsmformer-ppi-2026"],"links":[],"attributes":{"version":null,"split":"test set","missing_metadata":{"version":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-0aab382ca2c063","kind":"dataset","name":"CLA-IND0.6","description":"","status":"needs_review","facets":{"areas":["proteins-complexes"]},"source_ids":["clathrin-plm-2025"],"links":[],"attributes":{"version":null,"split":"independent test","missing_metadata":{"version":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-0b54f42a987b1d","kind":"dataset","name":"testing viral metagenome dataset","description":"","status":"needs_review","facets":{"areas":["microbes-communities"]},"source_ids":["detire-viral-metagenomes-2023"],"links":[],"attributes":{"version":null,"split":"test","missing_metadata":{"version":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-0bba1c9a7ae410","kind":"dataset","name":"genomic benchmark categories","description":"","status":"needs_review","facets":{"areas":["dna-genomes"]},"source_ids":["genomic-tokenizer-selection-2025"],"links":[],"attributes":{"version":null,"split":"paper benchmark summary","missing_metadata":{"version":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-0c3ac7efe99c37","kind":"dataset","name":"GenomeOcean natural/artificial sequence test","description":"","status":"needs_review","facets":{"areas":["microbes-communities"]},"source_ids":["genomeocean-2025"],"links":[],"attributes":{"version":null,"split":null,"missing_metadata":{"version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-0edd8f724db696","kind":"dataset","name":"antibody peptide-mapping training dataset","description":"","status":"needs_review","facets":{"areas":["proteins-complexes"]},"source_ids":["antibody-deamidation-plm-2024"],"links":[],"attributes":{"version":null,"split":"fivefold stratified CV","missing_metadata":{"version":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-158b121281b650","kind":"dataset","name":"Yoruban LCL dsQTLs","description":"","status":"needs_review","facets":{"areas":["dna-genomes"]},"source_ids":["arsenal-regulatory-dna-2026"],"links":[],"attributes":{"version":null,"split":null,"missing_metadata":{"version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-16d01b5ef88e84","kind":"dataset","name":"Fingerprint-scoring benchmark","description":"","status":"needs_review","facets":{"areas":["molecular-interactions"]},"source_ids":["fingerprint-scoring-2022"],"links":[],"attributes":{"version":null,"split":null,"missing_metadata":{"version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-17132fbabd7683","kind":"dataset","name":"CASF-2016","description":"","status":"needs_review","facets":{"areas":["molecular-interactions"]},"source_ids":["ligand-affinity-meta-model-2024"],"links":[],"attributes":{"version":null,"split":"core benchmark","missing_metadata":{"version":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-1744719eef145b","kind":"dataset","name":"human ultra-long mRNAs","description":"","status":"needs_review","facets":{"areas":["rna-transcriptomes"]},"source_ids":["mrnabert-2025"],"links":[],"attributes":{"version":null,"split":"paper evaluation","missing_metadata":{"version":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-18ebde58579c2b","kind":"dataset","name":"DEBFold TestSetβ","description":"","status":"needs_review","facets":{"areas":["rna-transcriptomes"]},"source_ids":["debfold-2024"],"links":[],"attributes":{"version":null,"split":null,"missing_metadata":{"version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-1b4f6ea24c0587","kind":"dataset","name":"LAMBDA genome-wide prophage test","description":"","status":"needs_review","facets":{"areas":["microbes-communities"]},"source_ids":["lambda-prophage-2026"],"links":[],"attributes":{"version":null,"split":null,"missing_metadata":{"version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-1c4c71078ffe01","kind":"dataset","name":"MetaHIT","description":"","status":"needs_review","facets":{"areas":["microbes-communities"]},"source_ids":["metagenomic-pathogens-2025"],"links":[],"attributes":{"version":null,"split":null,"missing_metadata":{"version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-1c7f8ebb1968d9","kind":"dataset","name":"T18","description":"","status":"needs_review","facets":{"areas":["rna-transcriptomes"]},"source_ids":["rlsite-rna-binding-2025"],"links":[],"attributes":{"version":null,"split":"paper evaluation","missing_metadata":{"version":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-235520c84b737f","kind":"dataset","name":"SARS-CoV-2 Mpro ligands","description":"","status":"needs_review","facets":{"areas":["molecular-interactions"]},"source_ids":["mpro-pose-affinity-2025"],"links":[],"attributes":{"version":null,"split":null,"missing_metadata":{"version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-236eaa4e55147f","kind":"dataset","name":"liver editing sites","description":"","status":"needs_review","facets":{"areas":["rna-transcriptomes"]},"source_ids":["adar-gpt-editing-2026"],"links":[],"attributes":{"version":null,"split":"15% validation set","missing_metadata":{"version":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-2ad2fad5e1cd0a","kind":"dataset","name":"hESC cell-type-specific GRN","description":"","status":"needs_review","facets":{"areas":["cells-tissues"]},"source_ids":["scregnet-2025"],"links":[],"attributes":{"version":null,"split":null,"missing_metadata":{"version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-2e87449871ca47","kind":"dataset","name":"mRNA-RBP pairs","description":"","status":"needs_review","facets":{"areas":["rna-transcriptomes"]},"source_ids":["mrna-protein-diversity-2026"],"links":[],"attributes":{"version":null,"split":"RBP-aware test set","missing_metadata":{"version":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-2eaa2a051d45ee","kind":"dataset","name":"variant-effects benchmark","description":"","status":"needs_review","facets":{"areas":["proteins-complexes"]},"source_ids":["structure-informed-plm-2025"],"links":[],"attributes":{"version":null,"split":"paper evaluation","missing_metadata":{"version":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-327cfcdae0c937","kind":"dataset","name":"PDBbind core v2016","description":"","status":"needs_review","facets":{"areas":["molecular-interactions"]},"source_ids":["deelig-2021"],"links":[],"attributes":{"version":"v2016","split":null,"missing_metadata":{"split":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-32ccef507a1dd7","kind":"dataset","name":"ncRNA interaction pairs","description":"","status":"needs_review","facets":{"areas":["rna-transcriptomes"]},"source_ids":["cupid-rna-interactions-2026"],"links":[],"attributes":{"version":null,"split":null,"missing_metadata":{"version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-33a41fe5fc66cf","kind":"dataset","name":"PBMCs-BS","description":"","status":"needs_review","facets":{"areas":["cells-tissues"]},"source_ids":["scalr-2025"],"links":[],"attributes":{"version":null,"split":null,"missing_metadata":{"version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-38151fa548e291","kind":"dataset","name":"HumanPPI","description":"","status":"needs_review","facets":{"areas":["proteins-complexes"]},"source_ids":["mulan-2025"],"links":[],"attributes":{"version":null,"split":"paper evaluation","missing_metadata":{"version":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-3a3e3880a3fed0","kind":"dataset","name":"mRNABench MRL-MPRA","description":"","status":"needs_review","facets":{"areas":["rna-transcriptomes"]},"source_ids":["mrnabench-2025"],"links":[],"attributes":{"version":null,"split":null,"missing_metadata":{"version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-43f24c4dfb7351","kind":"dataset","name":"human and viral proteins","description":"","status":"needs_review","facets":{"areas":["proteins-complexes"]},"source_ids":["viral-immune-mimicry-2025"],"links":[],"attributes":{"version":null,"split":null,"missing_metadata":{"version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-450c1af18cc623","kind":"dataset","name":"Simulated viral metagenome","description":"","status":"needs_review","facets":{"areas":["microbes-communities"]},"source_ids":["lazypipe-2020"],"links":[],"attributes":{"version":null,"split":null,"missing_metadata":{"version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-463197d6a98b99","kind":"dataset","name":"Human 5mC","description":"","status":"needs_review","facets":{"areas":["dna-genomes"]},"source_ids":["dna-foundation-models-2025"],"links":[],"attributes":{"version":null,"split":null,"missing_metadata":{"version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-477a9082515406","kind":"dataset","name":"Human thymus scRNA-seq","description":"","status":"needs_review","facets":{"areas":["cells-tissues"]},"source_ids":["mouse-geneformer-2025"],"links":[],"attributes":{"version":null,"split":null,"missing_metadata":{"version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-488d5de6bb9c1b","kind":"dataset","name":"M.S. single-cell dataset","description":"","status":"needs_review","facets":{"areas":["cells-tissues"]},"source_ids":["single-cell-peft-2024"],"links":[],"attributes":{"version":null,"split":null,"missing_metadata":{"version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-48def1da574597","kind":"dataset","name":"E. coli sigma70 promoter dataset","description":"","status":"needs_review","facets":{"areas":["microbes-communities"]},"source_ids":["prokbert-2024"],"links":[],"attributes":{"version":null,"split":null,"missing_metadata":{"version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-4b6c13924d4256","kind":"dataset","name":"PoseBusters","description":"","status":"needs_review","facets":{"areas":["molecular-interactions"]},"source_ids":["molas-2026"],"links":[],"attributes":{"version":null,"split":null,"missing_metadata":{"version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-50f0bdb7cf9ca4","kind":"dataset","name":"HIV neutralization","description":"","status":"needs_review","facets":{"areas":["molecular-interactions"]},"source_ids":["deepinteraware-2025"],"links":[],"attributes":{"version":null,"split":"antibody-unseen","missing_metadata":{"version":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-5197cca532f89d","kind":"dataset","name":"FUJISAN test sub-dataset","description":"","status":"needs_review","facets":{"areas":["proteins-complexes"]},"source_ids":["fujisan-2024"],"links":[],"attributes":{"version":null,"split":null,"missing_metadata":{"version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-52f00ccaabf0d9","kind":"dataset","name":"mRNA half-life","description":"","status":"needs_review","facets":{"areas":["rna-transcriptomes"]},"source_ids":["mrna-lm-2025"],"links":[],"attributes":{"version":null,"split":"test set across CV splits","missing_metadata":{"version":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-54b9bc432928d6","kind":"dataset","name":"flu-vaccine sequences","description":"","status":"needs_review","facets":{"areas":["rna-transcriptomes"]},"source_ids":["codonbert-vaccines-2024"],"links":[],"attributes":{"version":null,"split":"paper evaluation","missing_metadata":{"version":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-55f200c9481409","kind":"dataset","name":"poly(A) Gene-Gene","description":"","status":"needs_review","facets":{"areas":["dna-genomes"]},"source_ids":["polya-glm-2025"],"links":[],"attributes":{"version":null,"split":"5-fold cross-validation","missing_metadata":{"version":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-561834dfa1682c","kind":"dataset","name":"ICCTax Complete dataset","description":"","status":"needs_review","facets":{"areas":["microbes-communities"]},"source_ids":["icctax-2025"],"links":[],"attributes":{"version":null,"split":null,"missing_metadata":{"version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-571ce000cd74b9","kind":"dataset","name":"AIDA v2 PBMC cohort","description":"","status":"needs_review","facets":{"areas":["cells-tissues"]},"source_ids":["single-cell-aging-probes-2026"],"links":[],"attributes":{"version":"622 donors","split":null,"missing_metadata":{"split":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-59def895fbdbb4","kind":"dataset","name":"non-immune cells","description":"","status":"needs_review","facets":{"areas":["cells-tissues"]},"source_ids":["gremln-2026"],"links":[],"attributes":{"version":null,"split":"zero-shot","missing_metadata":{"version":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-5afaefb87c8a94","kind":"dataset","name":"PLINDER-L95","description":"","status":"needs_review","facets":{"areas":["molecular-interactions"]},"source_ids":["boltz-stereochemistry-2025"],"links":[],"attributes":{"version":null,"split":null,"missing_metadata":{"version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-6212e779949708","kind":"dataset","name":"Liu training dataset","description":"","status":"needs_review","facets":{"areas":["dna-genomes"]},"source_ids":["dnabert2-enhancer-2025"],"links":[],"attributes":{"version":null,"split":"5-fold cross-validation","missing_metadata":{"version":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-6a44f5946cd7ab","kind":"dataset","name":"LiPP lipid–protein complexes","description":"","status":"needs_review","facets":{"areas":["molecular-interactions"]},"source_ids":["lipp-2026"],"links":[],"attributes":{"version":"331 complexes","split":null,"missing_metadata":{"split":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-6e0c28dfde7337","kind":"dataset","name":"CATH superfamily benchmark","description":"","status":"needs_review","facets":{"areas":["proteins-complexes"]},"source_ids":["cathe2-2025"],"links":[],"attributes":{"version":null,"split":"paper evaluation","missing_metadata":{"version":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-701d910b02d25c","kind":"dataset","name":"SJC","description":"","status":"needs_review","facets":{"areas":["proteins-complexes"]},"source_ids":["clape-smb-2024"],"links":[],"attributes":{"version":null,"split":null,"missing_metadata":{"version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-71614d99b3099f","kind":"dataset","name":"Rfam","description":"","status":"needs_review","facets":{"areas":["rna-transcriptomes"]},"source_ids":["r3design-2025"],"links":[],"attributes":{"version":null,"split":"external","missing_metadata":{"version":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-72e837e5b97041","kind":"dataset","name":"HumanGut-all strain-level query","description":"","status":"needs_review","facets":{"areas":["microbes-communities"]},"source_ids":["cammiq-2022"],"links":[],"attributes":{"version":null,"split":null,"missing_metadata":{"version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-739aee3cf8d6f1","kind":"dataset","name":"DNA-129_Test","description":"","status":"needs_review","facets":{"areas":["proteins-complexes"]},"source_ids":["megsite-2025"],"links":[],"attributes":{"version":null,"split":"independent test","missing_metadata":{"version":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-768a7ff5bac414","kind":"dataset","name":"Zymo LOG 10%","description":"","status":"needs_review","facets":{"areas":["microbes-communities"]},"source_ids":["lemur-magnet-2024"],"links":[],"attributes":{"version":null,"split":null,"missing_metadata":{"version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-7bf2cf7d2b2d01","kind":"dataset","name":"stimulated immune PBMC","description":"","status":"needs_review","facets":{"areas":["cells-tissues"]},"source_ids":["insilico-perturbation-auprc-2025"],"links":[],"attributes":{"version":null,"split":"CD14+Mono","missing_metadata":{"version":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-7cec655cd742f3","kind":"dataset","name":"ProteinGym substitution DMS: stability assays","description":"","status":"needs_review","facets":{"areas":["proteins-complexes"]},"source_ids":["proteingym-2023"],"links":[],"attributes":{"version":null,"split":null,"missing_metadata":{"version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-7fc59ce4c0ceaa","kind":"dataset","name":"MosA1 reference → WholeBrainA query","description":"","status":"needs_review","facets":{"areas":["cells-tissues"]},"source_ids":["scatac-llmda-2026"],"links":[],"attributes":{"version":null,"split":null,"missing_metadata":{"version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-8317793f18b026","kind":"dataset","name":"PDB RNA set","description":"","status":"needs_review","facets":{"areas":["rna-transcriptomes"]},"source_ids":["bpfold-2025"],"links":[],"attributes":{"version":"116 RNAs","split":null,"missing_metadata":{"split":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-87e91d9d6e6f4b","kind":"dataset","name":"L1000","description":"","status":"needs_review","facets":{"areas":["cells-tissues"]},"source_ids":["cell2sentence-2024"],"links":[],"attributes":{"version":null,"split":null,"missing_metadata":{"version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-8cad416ddc80dc","kind":"dataset","name":"Real mock community MAGs","description":"","status":"needs_review","facets":{"areas":["microbes-communities"]},"source_ids":["kmetashot-2025"],"links":[],"attributes":{"version":null,"split":null,"missing_metadata":{"version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-8e9488896becd4","kind":"dataset","name":"GUE H-CPD","description":"","status":"needs_review","facets":{"areas":["dna-genomes"]},"source_ids":["eden-genomic-classification-2026"],"links":[],"attributes":{"version":null,"split":"paper evaluation","missing_metadata":{"version":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-8f123f006964ad","kind":"dataset","name":"hPancreas","description":"","status":"needs_review","facets":{"areas":["cells-tissues"]},"source_ids":["scelmo-2025"],"links":[],"attributes":{"version":null,"split":null,"missing_metadata":{"version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-9135087a16af1c","kind":"dataset","name":"Bernett PPI dataset","description":"","status":"needs_review","facets":{"areas":["molecular-interactions"]},"source_ids":["esm2-amp-2025"],"links":[],"attributes":{"version":null,"split":null,"missing_metadata":{"version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-99afd0c86b2954","kind":"dataset","name":"MirTarRAW","description":"","status":"needs_review","facets":{"areas":["rna-transcriptomes"]},"source_ids":["rnaret-2026"],"links":[],"attributes":{"version":null,"split":"held-out test","missing_metadata":{"version":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-9c186c8f4ed3f4","kind":"dataset","name":"ProteinGym substitutions","description":"","status":"needs_review","facets":{"areas":["proteins-complexes"]},"source_ids":["esm2-ofs-fitness-2025"],"links":[],"attributes":{"version":null,"split":"aggregate across assays","missing_metadata":{"version":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-9e9d18bc5bfb8b","kind":"dataset","name":"KEx","description":"","status":"needs_review","facets":{"areas":["dna-genomes"]},"source_ids":["quadruplex-llm-benchmark-2025"],"links":[],"attributes":{"version":null,"split":null,"missing_metadata":{"version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-a03b8e9efde37b","kind":"dataset","name":"human enhancer dataset","description":"","status":"needs_review","facets":{"areas":["dna-genomes"]},"source_ids":["enhancer-position-encoding-2024"],"links":[],"attributes":{"version":null,"split":null,"missing_metadata":{"version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-a1da4a37eb46a5","kind":"dataset","name":"Independent E. coli sigma70 test dataset","description":"","status":"needs_review","facets":{"areas":["microbes-communities"]},"source_ids":["cyaprombert-2022"],"links":[],"attributes":{"version":null,"split":"independent test","missing_metadata":{"version":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-a28180d33f7a23","kind":"dataset","name":"ClinVar 3-prime UTR variants","description":"","status":"needs_review","facets":{"areas":["dna-genomes"]},"source_ids":["phylogpn-2025"],"links":[],"attributes":{"version":null,"split":"paper evaluation","missing_metadata":{"version":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-a8610f2b80cdf0","kind":"dataset","name":"enhancer independent comparison","description":"","status":"needs_review","facets":{"areas":["dna-genomes"]},"source_ids":["hi-enhancer-2025"],"links":[],"attributes":{"version":null,"split":null,"missing_metadata":{"version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-abdfba8cce7486","kind":"dataset","name":"bpRNA-new","description":"","status":"needs_review","facets":{"areas":["rna-transcriptomes"]},"source_ids":["ernie-rna-2025"],"links":[],"attributes":{"version":null,"split":"cross-family test","missing_metadata":{"version":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-b1af840b76b351","kind":"dataset","name":"CoBRA compound-binding test set","description":"","status":"needs_review","facets":{"areas":["rna-transcriptomes"]},"source_ids":["cobra-rna-binding-2026"],"links":[],"attributes":{"version":null,"split":"test set","missing_metadata":{"version":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-b462aa24561fba","kind":"dataset","name":"CAMI II Toy human gastrooral sample19-new","description":"","status":"needs_review","facets":{"areas":["microbes-communities"]},"source_ids":["nabas-plus-2025"],"links":[],"attributes":{"version":null,"split":null,"missing_metadata":{"version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-b6ce37ba678d39","kind":"dataset","name":"DART-Eval cCREs versus matched shuffled controls","description":"","status":"needs_review","facets":{"areas":["dna-genomes"]},"source_ids":["dart-eval-regulatory-2024"],"links":[],"attributes":{"version":null,"split":"paper evaluation","missing_metadata":{"version":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-b6d0ebaca196a6","kind":"dataset","name":"Andropogoneae genome-wide conservation","description":"","status":"needs_review","facets":{"areas":["dna-genomes"]},"source_ids":["plantcad2-2025"],"links":[],"attributes":{"version":null,"split":null,"missing_metadata":{"version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-b7204b005bd476","kind":"dataset","name":"ImmuneBuilder antibody test set","description":"","status":"needs_review","facets":{"areas":["molecular-interactions"]},"source_ids":["ibex-2025"],"links":[],"attributes":{"version":null,"split":null,"missing_metadata":{"version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-b91c871eb7740a","kind":"dataset","name":"vaccine candidate validation","description":"","status":"needs_review","facets":{"areas":["proteins-complexes"]},"source_ids":["vaxign-esm-2024"],"links":[],"attributes":{"version":null,"split":null,"missing_metadata":{"version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-bc127dc9c441fe","kind":"dataset","name":"DNA barcodes of unseen species","description":"","status":"needs_review","facets":{"areas":["dna-genomes"]},"source_ids":["barcodebert-2026"],"links":[],"attributes":{"version":null,"split":"1-NN probe","missing_metadata":{"version":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-bd3d8e7d6cd196","kind":"dataset","name":"human RNA 2OMe sites","description":"","status":"needs_review","facets":{"areas":["rna-transcriptomes"]},"source_ids":["2ome-lm-2025"],"links":[],"attributes":{"version":null,"split":"5-fold cross-validation","missing_metadata":{"version":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-bd3f98e2eeb5d3","kind":"dataset","name":"CRC microbiome cohort","description":"","status":"needs_review","facets":{"areas":["microbes-communities"]},"source_ids":["mdl4microbiome-2022"],"links":[],"attributes":{"version":null,"split":null,"missing_metadata":{"version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-bd9255afb783d6","kind":"dataset","name":"ProteinShake VEP datasets","description":"","status":"needs_review","facets":{"areas":["proteins-complexes"]},"source_ids":["pst-2025"],"links":[],"attributes":{"version":null,"split":null,"missing_metadata":{"version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-beb4f5da29da0a","kind":"dataset","name":"CAMI II Sample_0 10,000-read subsample","description":"","status":"needs_review","facets":{"areas":["microbes-communities"]},"source_ids":["ncd-metagenomics-2026"],"links":[],"attributes":{"version":"10,000 reads","split":null,"missing_metadata":{"split":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-becc215358afd0","kind":"dataset","name":"PRIME mutated RBD","description":"","status":"needs_review","facets":{"areas":["proteins-complexes"]},"source_ids":["prime-2026"],"links":[],"attributes":{"version":null,"split":"position-stratified","missing_metadata":{"version":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-cd51026cdb6a7a","kind":"dataset","name":"20 medium/high-complexity viral simulations","description":"","status":"needs_review","facets":{"areas":["microbes-communities"]},"source_ids":["viral-contig-simulation-2021"],"links":[],"attributes":{"version":null,"split":null,"missing_metadata":{"version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-d28955d5872903","kind":"dataset","name":"AMP","description":"","status":"needs_review","facets":{"areas":["microbes-communities"]},"source_ids":["pc-mer-2024"],"links":[],"attributes":{"version":null,"split":"genus-level","missing_metadata":{"version":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-d356eac961cb69","kind":"dataset","name":"HPA-FoV","description":"","status":"needs_review","facets":{"areas":["cells-tissues"]},"source_ids":["cell-dino-2025"],"links":[],"attributes":{"version":null,"split":null,"missing_metadata":{"version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-d9fdd8dc7a0184","kind":"dataset","name":"immune tissue","description":"","status":"needs_review","facets":{"areas":["cells-tissues"]},"source_ids":["single-cell-residual-geometry-2026"],"links":[],"attributes":{"version":null,"split":null,"missing_metadata":{"version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-dba1707164d296","kind":"dataset","name":"TRX","description":"","status":"needs_review","facets":{"areas":["proteins-complexes"]},"source_ids":["spin-protein-function-2026"],"links":[],"attributes":{"version":null,"split":"test set","missing_metadata":{"version":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-e0f34dcaa1ba3b","kind":"dataset","name":"23 independent prokaryotic promoter test sets","description":"","status":"needs_review","facets":{"areas":["microbes-communities"]},"source_ids":["ipromp-2025"],"links":[],"attributes":{"version":"23 test sets","split":"independent test","missing_metadata":{"accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-e45a5a140888ee","kind":"dataset","name":"Boltz-1 structure test set","description":"","status":"needs_review","facets":{"areas":["molecular-interactions"]},"source_ids":["boltz1-2025"],"links":[],"attributes":{"version":null,"split":null,"missing_metadata":{"version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-eaa2965545c87b","kind":"dataset","name":"Aorta single-cell dataset","description":"","status":"needs_review","facets":{"areas":["cells-tissues"]},"source_ids":["genept-2024"],"links":[],"attributes":{"version":null,"split":null,"missing_metadata":{"version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-ebc3f5fda43972","kind":"dataset","name":"extremely long-sequence species classification","description":"","status":"needs_review","facets":{"areas":["rna-transcriptomes"]},"source_ids":["birna-bert-2025"],"links":[],"attributes":{"version":null,"split":"paper evaluation","missing_metadata":{"version":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-ee26acbd6e8cf7","kind":"dataset","name":"Antibody–antigen GEP test set","description":"","status":"needs_review","facets":{"areas":["molecular-interactions"]},"source_ids":["antibody-flexibility-2025"],"links":[],"attributes":{"version":null,"split":null,"missing_metadata":{"version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-f08b1a60aebeeb","kind":"dataset","name":"Dset_448","description":"","status":"needs_review","facets":{"areas":["proteins-complexes"]},"source_ids":["protein-binding-sites-2023"],"links":[],"attributes":{"version":null,"split":null,"missing_metadata":{"version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-f0bf60a62ad7c3","kind":"dataset","name":"Genomic Benchmarks Mouse Enhancers","description":"","status":"needs_review","facets":{"areas":["dna-genomes"]},"source_ids":["enbed-2024"],"links":[],"attributes":{"version":null,"split":null,"missing_metadata":{"version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-f18fcc23dfa798","kind":"dataset","name":"CASF-2016","description":"","status":"needs_review","facets":{"areas":["molecular-interactions"]},"source_ids":["akscore-2020"],"links":[],"attributes":{"version":null,"split":null,"missing_metadata":{"version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-f2e729f333a333","kind":"dataset","name":"RNA8F","description":"","status":"needs_review","facets":{"areas":["rna-transcriptomes"]},"source_ids":["tu-fold-2025"],"links":[],"attributes":{"version":null,"split":null,"missing_metadata":{"version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-f6922a9744ba27","kind":"dataset","name":"gene fusion breakpoint DNA sequences","description":"","status":"needs_review","facets":{"areas":["dna-genomes"]},"source_ids":["fusion-breakpoint-foundation-models-2026"],"links":[],"attributes":{"version":null,"split":"full test set","missing_metadata":{"version":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-f7210686a78474","kind":"dataset","name":"scXDR transfer scenario 2","description":"","status":"needs_review","facets":{"areas":["cells-tissues"]},"source_ids":["scxdr-2026"],"links":[],"attributes":{"version":null,"split":null,"missing_metadata":{"version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-dataset-fcb5752d916d5d","kind":"dataset","name":"DNALongBench ETGP","description":"","status":"needs_review","facets":{"areas":["dna-genomes"]},"source_ids":["dnalongbench-2025"],"links":[],"attributes":{"version":null,"split":null,"missing_metadata":{"version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract","accession":"not_reported_in_legacy_extract"}}}
{"id":"reported-model-0068c3eff1bf7b","kind":"model","name":"TOPBP (Complex)","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["molecular-interactions"]},"source_ids":["deelig-2021"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"TOPBP (Complex)","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"TOPBP (Complex) is a complex-structure binding-affinity comparator listed in the DEELIG study.","summary_source_ids":["deelig-2021"],"summary_source_locator":"Discussion (paragraph 7); Materials and Methods/Additional case studies of specific protein families (paragraph 3)","sections":[{"title":"How the evaluated method works","body":"This record preserves the complex-input comparator identity printed in the source table. The table does not establish which upstream TopBP implementation/checkpoint generated this exact row.","source_ids":["deelig-2021"],"source_locator":"Materials and Methods/Additional case studies of specific protein families (paragraph 2); Materials and Methods/Additional case studies of specific protein families (paragraph 3)"},{"title":"What was evaluated","body":"The linked evaluation record identifies TOPBP (Complex): Protein–ligand binding affinity prediction. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["deelig-2021"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b3-043"}],"facts":[{"label":"Model type","value":"Study-specific predictive method; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["deelig-2021"],"source_locator":"Materials and Methods/Additional case studies of specific protein families (paragraph 2); Materials and Methods/Additional case studies of specific protein families (paragraph 3)"},{"label":"Architecture / procedure","value":"This record preserves the complex-input comparator identity printed in the source table. The table does not establish which upstream TopBP implementation/checkpoint generated this exact row.","status":"source_checked","source_ids":["deelig-2021"],"source_locator":"Materials and Methods/Additional case studies of specific protein families (paragraph 2); Materials and Methods/Additional case studies of specific protein families (paragraph 3)"},{"label":"Biological inputs","value":"Protein–ligand complex structure","status":"source_checked","source_ids":["deelig-2021"],"source_locator":"Materials and Methods/Data set refinement (paragraph 3); Materials and Methods/Data set refinement (paragraph 1)"},{"label":"Outputs","value":"Predicted binding affinity","status":"source_checked","source_ids":["deelig-2021"],"source_locator":"Materials and Methods/Novel data set: raw data (paragraph 3); Materials and Methods/Novel data set: raw data (paragraph 2)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["deelig-2021","evidence-reported-deelig-2021-readme-md"],"source_locator":"Materials and Methods/Novel data set: raw data; Materials and Methods/Data set refinement; Materials and Methods/Feature extraction; Materials and Methods/Feature extraction/Protein-pocket features; Materials and Methods/Feature extraction/Ligand features; Materials and Methods/Feature extraction/Grid formation; Materials and Methods/Strategies; Materials and Methods/Strategies/Atomic model/Preprocessing; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"TOPBP (Complex) is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["deelig-2021"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"The Deelig comparison table identifies TOPBP (Complex) but does not resolve an exact implementation, fitted artifact or training configuration.","status":"unreported","source_ids":["deelig-2021"],"source_locator":"The Deelig model-comparison table; TOPBP (Complex) row"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["deelig-2021","evidence-reported-deelig-2021-readme-md"],"source_locator":"Materials and Methods/Novel data set: raw data; Materials and Methods/Data set refinement; Materials and Methods/Feature extraction; Materials and Methods/Feature extraction/Protein-pocket features; Materials and Methods/Feature extraction/Ligand features; Materials and Methods/Feature extraction/Grid formation; Materials and Methods/Strategies; Materials and Methods/Strategies/Atomic model/Preprocessing; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/asadahmedtech/DEELIG/blob/3a3993fc903c40f1ce904111c8e085c79fb45df6/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-deelig-2021-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","status":"unreported","source_ids":["evidence-reported-deelig-2021-readme-md"],"source_locator":"README.md and repository-root licence-file search"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-deelig-2021-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Offers a complex-input reference against a method intended to work without a docked pose as user input.","source_ids":["deelig-2021"],"source_locator":"Introduction (paragraph 3); Abstract (paragraph 1)"}],"limitations":[{"text":"The printed abbreviated identity has not been resolved to a unique upstream version; detailed architecture and licensing must not be inferred from the name alone.","source_ids":["deelig-2021"],"source_locator":"Materials and Methods/Feature extraction/Ligand features (paragraph 3); Materials and Methods/Strategies (paragraph 1)"}],"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","TOPBP (Complex) is the comparison-table label; that label does not specify an immutable weight revision.","The Deelig comparison table identifies TOPBP (Complex) but does not resolve an exact implementation, fitted artifact or training configuration.","A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-028e4bb9baa074","kind":"model","name":"Single best solver","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["molecular-interactions"]},"source_ids":["molas-2026"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"Single best solver","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"The single-best-solver baseline selects one docking algorithm for every instance in a fixed benchmark.","summary_source_ids":["molas-2026"],"summary_source_locator":"Introduction (paragraph 2); Introduction (paragraph 3)","sections":[{"title":"How the evaluated method works","body":"Choose the algorithm with the best aggregate performance in the protocol’s selection data, then use that same solver on all evaluated cases.","source_ids":["molas-2026"],"source_locator":"Materials and methods/Scoring function/Physicochemical plausibility (paragraph 2); Materials and methods/Datasets and preprocessing/Extra benchmarks (paragraph 3)"},{"title":"What was evaluated","body":"The linked evaluation record identifies Single best solver: Physically valid protein–ligand pose selection. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["molas-2026"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b3-045"}],"facts":[{"label":"Model type","value":"Study-specific predictive method; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["molas-2026"],"source_locator":"Materials and methods/Scoring function/Physicochemical plausibility (paragraph 2); Materials and methods/Datasets and preprocessing/Extra benchmarks (paragraph 3)"},{"label":"Architecture / procedure","value":"Choose the algorithm with the best aggregate performance in the protocol’s selection data, then use that same solver on all evaluated cases.","status":"source_checked","source_ids":["molas-2026"],"source_locator":"Materials and methods/Scoring function/Physicochemical plausibility (paragraph 2); Materials and methods/Datasets and preprocessing/Extra benchmarks (paragraph 3)"},{"label":"Biological inputs","value":"A fixed candidate-solver set and the benchmark’s solver-selection data","status":"source_checked","source_ids":["molas-2026"],"source_locator":"Introduction (paragraph 3); Introduction (paragraph 2)"},{"label":"Outputs","value":"One constant algorithm choice","status":"source_checked","source_ids":["molas-2026"],"source_locator":"Materials and methods/Scoring function/Physicochemical plausibility (paragraph 2); Materials and methods/Experimental setup/Reporting protocol (paragraph 2)"},{"label":"Parameters","value":"Not applicable: this is a selection reference, not a neural model.","status":"inapplicable","source_ids":["molas-2026"],"source_locator":"Materials and methods/Datasets and preprocessing/Dataset (paragraph 2); Results/Benchmark performance/In-domain learning (paragraph 2)"},{"label":"Known versions / configuration","value":"Single best solver is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["molas-2026"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"Select one solver using aggregate training-set performance, then reuse it across test instances; the policy does not train a molecular representation.","status":"source_checked","source_ids":["molas-2026"],"source_locator":"Materials and methods/Datasets and preprocessing/Extra benchmarks (paragraph 7); Materials and methods/Datasets and preprocessing/Dataset (paragraph 1)"},{"label":"Context limits","value":"Not applicable to a sequence-token limit; applicability follows the selected docking solver and input instance.","status":"inapplicable","source_ids":["molas-2026"],"source_locator":"Introduction (paragraph 2); Discussion/When is MolAS useful in practice? (paragraph 1)"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/BradWangW/MolAS/blob/a6c417216ddb992f7dc513d511d0429aece4bd61/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-molas-2026-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"MIT (study repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-molas-2026-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"Not applicable.","status":"inapplicable","source_ids":["molas-2026"],"source_locator":"Back/Availability notes (paragraph 1); Appendix/Appendix/Supplemental results/Top-k overlap across post-processing settings (paragraph 1)"}],"strengths":[{"text":"Measures whether per-instance selection improves over one consistently chosen solver.","source_ids":["molas-2026"],"source_locator":"Introduction (paragraph 2); Materials and methods/Experimental setup/Reporting protocol (paragraph 2)"}],"limitations":[{"text":"The identity of the winning solver depends on the benchmark and protocol; it is not a universal best docking method.","source_ids":["molas-2026"],"source_locator":"Materials and methods/Experimental setup/Reporting protocol (paragraph 1); Introduction (paragraph 3)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["A fixed candidate-solver set and the benchmark’s solver-selection data","Single best solver","One constant algorithm choice"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["molas-2026"],"source_locator":"Materials and methods/Scoring function/Physicochemical plausibility (paragraph 2); Materials and methods/Datasets and preprocessing/Extra benchmarks (paragraph 3)"},"coverage":"limited","gaps":["Single best solver is the comparison-table label; that label does not specify an immutable weight revision.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-03080a5289c07e","kind":"model","name":"Prompt","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["microbes-communities"]},"source_ids":["ipromp-2025"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"Prompt","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Prompt is the voting-based prokaryotic promoter comparator in the iPro-MP study.","summary_source_ids":["ipromp-2025"],"summary_source_locator":"Results/iPro-MP outperforms existing tools in multi-species promoter prediction (paragraph 1); Discussion (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"An existing voting-based prediction method is evaluated on the same 23 independent species test sets as the study’s other comparators.","source_ids":["ipromp-2025"],"source_locator":"Methods/Data collection and preprocessing (paragraph 2); Discussion (paragraph 2)"},{"title":"What was evaluated","body":"The linked evaluation record identifies Prompt: Multi-species prokaryotic promoter detection. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["ipromp-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b3-034"}],"facts":[{"label":"Model type","value":"Study-specific predictive method; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["ipromp-2025"],"source_locator":"Methods/Data collection and preprocessing (paragraph 2); Discussion (paragraph 2)"},{"label":"Architecture / procedure","value":"An existing voting-based prediction method is evaluated on the same 23 independent species test sets as the study’s other comparators.","status":"source_checked","source_ids":["ipromp-2025"],"source_locator":"Methods/Data collection and preprocessing (paragraph 2); Discussion (paragraph 2)"},{"label":"Biological inputs","value":"Prokaryotic DNA promoter windows","status":"source_checked","source_ids":["ipromp-2025"],"source_locator":"Results/iPro-MP reveals the species-specificity at the sequential level (paragraph 1); Results/iPro-MP outperforms existing tools in multi-species promoter prediction (paragraph 1)"},{"label":"Outputs","value":"Promoter classifications","status":"source_checked","source_ids":["ipromp-2025"],"source_locator":"Results/iPro-MP reveals the species-specificity at the sequential level (paragraph 1); Results/iPro-MP outperforms existing tools in multi-species promoter prediction (paragraph 1)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["ipromp-2025","evidence-reported-ipromp-2025-readme-md"],"source_locator":"Methods; Methods/Data collection and preprocessing; Methods/DNABERT model; Methods/Fine-tuning of DNABERT model; Methods/Performance evaluation; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"Prompt is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["ipromp-2025"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"The original Prompt method is described as supporting 16 prokaryotes; that scope differs from the 23-species evaluation.","status":"source_checked","source_ids":["ipromp-2025"],"source_locator":"Background (paragraph 3); Discussion (paragraph 4)"},{"label":"Context limits","value":"The comparison uses 81-bp promoter windows covering −60 to +20 relative to the transcription start site.","status":"source_checked","source_ids":["ipromp-2025"],"source_locator":"Methods / Data collection and preprocessing"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/Jackie-Suv/iPro-MP/blob/4266b521bc6617db939c5871cb1b6850dff63fdb/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-ipromp-2025-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","status":"unreported","source_ids":["evidence-reported-ipromp-2025-readme-md"],"source_locator":"README.md and repository-root licence-file search"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-ipromp-2025-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Provides a multi-species specialist reference.","source_ids":["ipromp-2025"],"source_locator":"Discussion (paragraph 1); Results/iPro-MP reveals the species-specificity at the sequential level (paragraph 5)"}],"limitations":[{"text":"The iPro-MP paper identifies the method but does not pin every component model or deployed webserver revision.","source_ids":["ipromp-2025"],"source_locator":"Methods (paragraph 1); Results/iPro-MP exhibits excellent performance and robustness in multi-species promoter prediction (paragraph 4)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Prokaryotic DNA promoter windows","Prompt","Promoter classifications"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["ipromp-2025"],"source_locator":"Methods/Data collection and preprocessing (paragraph 2); Discussion (paragraph 2)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","Prompt is the comparison-table label; that label does not specify an immutable weight revision.","No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-035a3ab36a3a6a","kind":"model","name":"structure-informed pLM","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["proteins-complexes"]},"source_ids":["structure-informed-plm-2025"],"links":[],"attributes":{"entity_level":"method","version":"not stated in table","reported_name":"structure-informed pLM","historical_missing_metadata":{"checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"This structure-informed protein language model uses structural prediction objectives during training to improve variant-effect scoring.","summary_source_ids":["evidence-reported-structure-informed-current-html"],"summary_source_locator":"Structure-Informed Protein Language Models; Cross-Modal Masked Learning (Denoising) Framework; Table 4","sections":[{"title":"How the evaluated method works","body":"Cross-modal masked learning combines amino-acid recovery with secondary-structure, relative-solvent-accessibility and contact-map prediction heads. The linked AA+SS+RSA+CM configuration uses all four objectives.","source_ids":["evidence-reported-structure-informed-current-html"],"source_locator":"Structure-Informed Protein Language Models; Cross-Modal Masked Learning (Denoising) Framework; Table 4"},{"title":"What was evaluated","body":"The linked evaluation record identifies structure-informed pLM: protein variant-effect classification. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["evidence-reported-structure-informed-current-html"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-b2-structure-informed-plm-2025"}],"facts":[{"label":"Model type","value":"Study-specific predictive method; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-structure-informed-current-html"],"source_locator":"Structure-Informed Protein Language Models; Cross-Modal Masked Learning (Denoising) Framework; Table 4"},{"label":"Architecture / procedure","value":"Cross-modal masked learning combines amino-acid recovery with secondary-structure, relative-solvent-accessibility and contact-map prediction heads. The linked AA+SS+RSA+CM configuration uses all four objectives.","status":"source_checked","source_ids":["evidence-reported-structure-informed-current-html"],"source_locator":"Structure-Informed Protein Language Models; Cross-Modal Masked Learning (Denoising) Framework; Table 4"},{"label":"Biological inputs","value":"Protein sequence at inference; sequence/structure pairs supply additional training supervision","status":"source_checked","source_ids":["evidence-reported-structure-informed-current-html"],"source_locator":"Structure-Informed Protein Language Models; Cross-Modal Masked Learning (Denoising) Framework; Table 4"},{"label":"Outputs","value":"Sequence-based variant-effect scores","status":"source_checked","source_ids":["evidence-reported-structure-informed-current-html"],"source_locator":"Structure-Informed Protein Language Models; Cross-Modal Masked Learning (Denoising) Framework; Table 4"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["evidence-reported-structure-informed-current-html","evidence-reported-structure-informed-plm-2025-readme-md"],"source_locator":"Complete primary text and named comparison table; inspected for aggregate parameter count (component sizes are not added without an exact configuration); readMe.md at pinned repository revision"},{"label":"Known versions / configuration","value":"structure-informed pLM is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["evidence-reported-structure-informed-current-html"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"Family datasets combine sequence information with experimental PDB and predicted AlphaFold structures.","status":"source_checked","source_ids":["evidence-reported-structure-informed-current-html"],"source_locator":"Structure-Informed Protein Language Models; Cross-Modal Masked Learning (Denoising) Framework; Table 4"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["evidence-reported-structure-informed-current-html","evidence-reported-structure-informed-plm-2025-readme-md"],"source_locator":"Complete primary text and named comparison table; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); readMe.md at pinned repository revision"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/Shen-Lab/Structure-informed_PLM/blob/2307b101f9bf08223729a68f52b8a6fb21f18991/readMe.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-structure-informed-plm-2025-readme-md"],"source_locator":"readMe.md; installation, model download and usage instructions"},{"label":"Code licence","value":"MIT (study repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-structure-informed-plm-2025-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-structure-informed-plm-2025-readme-md"],"source_locator":"readMe.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Structural context regularises training without requiring a variant structure at inference.","source_ids":["evidence-reported-structure-informed-current-html"],"source_locator":"Structure-Informed Protein Language Models; Cross-Modal Masked Learning (Denoising) Framework; Table 4"}],"limitations":[{"text":"The training structural information and family-specific adaptation are part of the evaluated method; it is not equivalent to an unadapted sequence-only encoder.","source_ids":["evidence-reported-structure-informed-current-html"],"source_locator":"Structure-Informed Protein Language Models; Cross-Modal Masked Learning (Denoising) Framework; Table 4"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Protein sequence at inference","structure-informed pLM","Sequence-based variant-effect scores"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["evidence-reported-structure-informed-current-html"],"source_locator":"Structure-Informed Protein Language Models; Cross-Modal Masked Learning (Denoising) Framework; Table 4"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","structure-informed pLM is the comparison-table label; that label does not specify an immutable weight revision.","A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-05103f72325fe5","kind":"model","name":"BarcodeBERT (4–4-4)","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["dna-genomes"]},"source_ids":["barcodebert-2026"],"links":[],"attributes":{"entity_level":"method","version":"4–4–4","reported_name":"BarcodeBERT (4–4-4)","historical_missing_metadata":{"checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"BarcodeBERT learns DNA-barcode representations for taxonomic identification; this record is the four-layer, four-head, 4-mer configuration.","summary_source_ids":["barcodebert-2026"],"summary_source_locator":"3 Methods/3.2 Proposed method: BarcodeBERT (paragraph 1); 6 Discussion (paragraph 2)","sections":[{"title":"How the evaluated method works","body":"Non-overlapping 4-mer tokens pass through four transformer layers. Masked-token pretraining with random offsets learns barcode representations, which are average-pooled for a cosine-similarity nearest-neighbour genus probe.","source_ids":["barcodebert-2026"],"source_locator":"3 Methods/3.2 Proposed method: BarcodeBERT (paragraph 4); 3 Methods/3.2 Proposed method: BarcodeBERT (paragraph 2)"},{"title":"What was evaluated","body":"The linked evaluation record identifies BarcodeBERT (4–4-4): unseen-species genus classification. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["barcodebert-2026"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-b2-barcodebert-2026"}],"facts":[{"label":"Model type","value":"Transformer representation pipeline; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["barcodebert-2026"],"source_locator":"3 Methods/3.2 Proposed method: BarcodeBERT (paragraph 4); 3 Methods/3.2 Proposed method: BarcodeBERT (paragraph 2)"},{"label":"Architecture / procedure","value":"Non-overlapping 4-mer tokens pass through four transformer layers. Masked-token pretraining with random offsets learns barcode representations, which are average-pooled for a cosine-similarity nearest-neighbour genus probe.","status":"source_checked","source_ids":["barcodebert-2026"],"source_locator":"3 Methods/3.2 Proposed method: BarcodeBERT (paragraph 4); 3 Methods/3.2 Proposed method: BarcodeBERT (paragraph 2)"},{"label":"Biological inputs","value":"COI DNA barcode sequences","status":"source_checked","source_ids":["barcodebert-2026"],"source_locator":"4 Experiments/4.1 Experimental setup (paragraph 1); 7 Conclusions (paragraph 2)"},{"label":"Outputs","value":"Barcode embeddings and genus assignments through the evaluated 1-nearest-neighbour probe","status":"source_checked","source_ids":["barcodebert-2026"],"source_locator":"6 Discussion (paragraph 4); 4 Experiments/4.1 Experimental setup/4.1.2 Linear probing (paragraph 1)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["barcodebert-2026","evidence-reported-barcodebert-2026-readme-md"],"source_locator":"3 Methods; 3 Methods/3.1 Dataset; 3 Methods/3.1 Dataset/3.1.1 Data pre-processing; 3 Methods/3.1 Dataset/3.1.2 Data partitioning; 3 Methods/3.2 Proposed method: BarcodeBERT; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"4–4–4","status":"source_checked","source_ids":["barcodebert-2026"],"source_locator":"Table vbag054-T5 (paragraph 1); Table vbag054-T4 (paragraph 1)"},{"label":"Training data / fitting","value":"The raw BOLD reference contains about 1.5 million Canadian invertebrate barcodes. Filtering yields 965,289 sequences; the pretraining partition contains 893,744, separate from seen-species and unseen-species evaluation partitions.","status":"source_checked","source_ids":["barcodebert-2026"],"source_locator":"3 Methods/3.1 Dataset/3.1.2 Data partitioning (paragraph 1); 3 Methods/3.1 Dataset (paragraph 1)"},{"label":"Context limits","value":"Sequences are padded or truncated to 660 nucleotides.","status":"source_checked","source_ids":["barcodebert-2026"],"source_locator":"3 Methods/3.2 Proposed method: BarcodeBERT (paragraph 3); 3 Methods/3.1 Dataset/3.1.1 Data pre-processing (paragraph 1)"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/bioscan-ml/BarcodeBERT/blob/00e492374eb748ed0f034a3a5981ab4eeffd92cc/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-barcodebert-2026-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"MIT (study repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-barcodebert-2026-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-barcodebert-2026-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Barcode-specific pretraining is evaluated on species absent from the training partition.","source_ids":["barcodebert-2026"],"source_locator":"4 Experiments/4.1 Experimental setup/4.1.1 Fine-tuning (paragraph 1); 4 Experiments/4.1 Experimental setup/4.1.2 Linear probing (paragraph 1)"}],"limitations":[{"text":"The unseen-species genus probe retains known genera. Its accuracy is not evidence for identifying entirely new genera.","source_ids":["barcodebert-2026"],"source_locator":"4 Experiments/4.1 Experimental setup/4.1.3 1-NN probing (paragraph 1); 5 Results/5.4 Substitution token rate (paragraph 1)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["COI DNA barcode sequences","BarcodeBERT (4–4-4)","Barcode embeddings and genus assignments through the evaluated 1-nearest-neighbour probe"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["barcodebert-2026"],"source_locator":"3 Methods/3.2 Proposed method: BarcodeBERT (paragraph 4); 3 Methods/3.2 Proposed method: BarcodeBERT (paragraph 2)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-06816ce9073144","kind":"model","name":"Geneformer","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["cells-tissues"]},"source_ids":["cell2sentence-2024"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"Geneformer","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Geneformer is an expression-pretrained comparator in the Cell2Sentence label-prediction study.","summary_source_ids":["cell2sentence-2024"],"summary_source_locator":"Discussion (paragraph 1); Methods (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"Rank-encoded gene-expression input passes through a pretrained transformer; the study evaluates its representation on label-prediction tasks.","source_ids":["cell2sentence-2024"],"source_locator":"Methods/Data transformation (paragraph 7); Methods/Data transformation (paragraph 6)"},{"title":"Underlying method and version boundaries","body":"Geneformer ranks genes by expression scaled against its pretraining corpus, then uses a transformer encoder with a masked-gene objective. V1 and V2 have different corpora, vocabularies, sizes and context limits, so a historical paper name is not replaced with today’s default checkpoint.","source_ids":["evidence-reported-base-geneformer-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies Geneformer: Combinatorial cell-label classification. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["cell2sentence-2024"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-028"}],"facts":[{"label":"Model type","value":"Single-cell transformer; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-geneformer-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"Rank-encoded gene-expression input passes through a pretrained transformer; the study evaluates its representation on label-prediction tasks.","status":"source_checked","source_ids":["cell2sentence-2024"],"source_locator":"Methods/Data transformation (paragraph 7); Methods/Data transformation (paragraph 6)"},{"label":"Biological inputs","value":"Rank-encoded transcriptomic profiles","status":"source_checked","source_ids":["cell2sentence-2024"],"source_locator":"Methods/Data transformation (paragraph 7); Methods/Data transformation (paragraph 6)"},{"label":"Outputs","value":"Cell/sample representations and label predictions","status":"source_checked","source_ids":["cell2sentence-2024"],"source_locator":"Experiments/Fine-Tuning Datasets (paragraph 5); Inference Details (paragraph 1)"},{"label":"Parameters","value":"46,107,089 parameters for the model checkpoint used in the Cell2Sentence L1000 experiment; this is not a claim about every checkpoint in the family.","status":"source_checked","source_ids":["cell2sentence-2024"],"source_locator":"Table 5, Comparison of required compute on the L1000 dataset; # Parameters column and caption"},{"label":"Known versions / configuration","value":"Geneformer is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["cell2sentence-2024"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"The reported comparator transfers its pretrained single-cell representation to the study’s label-prediction tasks, including out-of-distribution L1000 and GTEx bulk data and stimulated human PBMCs. Its precise pretraining checkpoint is not pinned by the comparison passage.","status":"source_checked","source_ids":["cell2sentence-2024"],"source_locator":"Experiments / Experiment 2: cell label prediction / Methodology"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["cell2sentence-2024","evidence-reported-base-geneformer-readme-md"],"source_locator":"Background and Related Work/Multimodal Training and Cross-Modality Encoding; Methods; Methods/Data transformation; Methods/Transformation robustness; Methods/Tasks; Experiments/Experiment 2: cell label prediction/Methodology:; Experiments/Experiment 3: abstract summary generation/Methodology:; Method Details; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://huggingface.co/ctheodoris/Geneformer/blob/1f7fbae4e469a5f4f1af8c111a529cfe1b3829f5/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-geneformer-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","status":"unreported","source_ids":["evidence-reported-base-geneformer-readme-md"],"source_locator":"README.md and repository-root licence-file search"},{"label":"Weights licence","value":"Apache 2.0 is declared in the official Geneformer model-card metadata; the exact historical configuration still needs its checkpoint identity.","status":"source_checked","source_ids":["evidence-reported-base-geneformer-readme-md"],"source_locator":"README.md front matter, license field; model-version list"}],"strengths":[{"text":"Supplies a single-cell-pretrained reference for evaluating the C2S representation.","source_ids":["cell2sentence-2024"],"source_locator":"Methods/Data transformation (paragraph 8); Generated abstracts/Examples of generated abstracts. (paragraph 1)"}],"limitations":[{"text":"Bulk L1000 and GTEx evaluations are distribution shifts from single-cell pretraining; the C2S method’s architecture must not be assigned to this comparator.","source_ids":["cell2sentence-2024"],"source_locator":"Experiments/Experiment 2: cell label prediction/Methodology: (paragraph 1); Experimental Details/Evaluation Datasets (paragraph 1)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Rank-encoded transcriptomic profiles","Geneformer","Cell/sample representations and label predictions"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["cell2sentence-2024"],"source_locator":"Methods/Data transformation (paragraph 7); Methods/Data transformation (paragraph 6)"},"coverage":"limited","gaps":["Geneformer is the comparison-table label; that label does not specify an immutable weight revision.","A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-0829aff5471d4b","kind":"model","name":"Stacking-Auto","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["dna-genomes"]},"source_ids":["hi-enhancer-2025"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"Stacking-Auto","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Stacking-Auto is the sequence-based enhancer-location stage of the Hi-Enhancer framework.","summary_source_ids":["hi-enhancer-2025"],"summary_source_locator":"1 Introduction (paragraph 6); 4 Discussion (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"DNABERT-2 extracts DNA sequence features and an AutoGluon stacking ensemble predicts enhancer labels. This is separate from the first-stage Blending-KAN predictor that uses epigenetic signals.","source_ids":["hi-enhancer-2025"],"source_locator":"Abstract (paragraph 1); 4 Discussion (paragraph 1)"},{"title":"What was evaluated","body":"The linked evaluation record identifies Stacking-Auto: enhancer prediction. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["hi-enhancer-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b4-003"}],"facts":[{"label":"Model type","value":"Study-specific predictive method; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["hi-enhancer-2025"],"source_locator":"Abstract (paragraph 1); 4 Discussion (paragraph 1)"},{"label":"Architecture / procedure","value":"DNABERT-2 extracts DNA sequence features and an AutoGluon stacking ensemble predicts enhancer labels. This is separate from the first-stage Blending-KAN predictor that uses epigenetic signals.","status":"source_checked","source_ids":["hi-enhancer-2025"],"source_locator":"Abstract (paragraph 1); 4 Discussion (paragraph 1)"},{"label":"Biological inputs","value":"DNA sequences encoded with DNABERT-2","status":"source_checked","source_ids":["hi-enhancer-2025"],"source_locator":"2 Materials and methods/2.2 Localization of the boundaries of enhancers/2.2.1 Stacking-Auto model (paragraph 1); 4 Discussion/4.2 A novel strategy for enhancer localization (paragraph 1)"},{"label":"Outputs","value":"Enhancer predictions","status":"source_checked","source_ids":["hi-enhancer-2025"],"source_locator":"1 Introduction (paragraph 3); 3 Results/3.3 Performance of Blending-KAN on cross-cell line prediction/3.3.4 Best performance of the five kinds of signal combinations (paragraph 2)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["hi-enhancer-2025","evidence-reported-hi-enhancer-2025-readme-txt"],"source_locator":"2 Materials and methods/2.1 Blending-KAN for predicting enhancer regions/2.1.1 Preprocessing datasets; 2 Materials and methods/2.1 Blending-KAN for predicting enhancer regions/2.1.2 Blending-KAN model; 2 Materials and methods/2.1 Blending-KAN for predicting enhancer regions/2.1.2 Blending-KAN model/2.1.2.1 Layer 1 of Blending-KAN; 2 Materials and methods/2.1 Blending-KAN for predicting enhancer regions/2.1.2 Blending-KAN model/2.1.2.2 Layer 2 of Blending-KAN; 2 Materials and methods/2.2 Localization of the boundaries of enhancers; 2 Materials and methods/2.2 Localization of the boundaries of enhancers/2.2.1 Stacking-Auto model; 2 Materials and methods/2.2 Localization of the boundaries of enhancers/2.2.2 Locating enhancers; 3 Results/3.4 Comparison of Stacking-Auto with existing methods; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.txt at pinned repository revision"},{"label":"Known versions / configuration","value":"Stacking-Auto is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["hi-enhancer-2025"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"Stacking-Auto is fitted on the iEnhancer-2L benchmark dataset using DNABERT-2 embeddings and ten-fold cross-validated LightGBM meta-features. This differs from the HCT116/A549 data used for Blending-KAN.","status":"source_checked","source_ids":["hi-enhancer-2025"],"source_locator":"3 Results/3.3 Performance of Blending-KAN on cross-cell line prediction (paragraph 1); 2 Materials and methods/2.2 Localization of the boundaries of enhancers/2.2.1 Stacking-Auto model (paragraph 2)"},{"label":"Context limits","value":"Enhancer localisation evaluates 77 sliding subsequences within each 4,000-bp region; the paper points to Supplementary Text S8 for exact window details.","status":"source_checked","source_ids":["hi-enhancer-2025"],"source_locator":"4 Discussion (paragraph 1); 2 Materials and methods/2.2 Localization of the boundaries of enhancers/2.2.2 Locating enhancers (paragraph 1)"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/emanlee/Hi-Enhancer/blob/435bb1cc9ec2909d6ee551bb53c56f6a7bdde8fd/README.txt. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-hi-enhancer-2025-readme-txt"],"source_locator":"README.txt; installation, model download and usage instructions"},{"label":"Code licence","value":"No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","status":"unreported","source_ids":["evidence-reported-hi-enhancer-2025-readme-txt"],"source_locator":"README.txt and repository-root licence-file search"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-hi-enhancer-2025-readme-txt"],"source_locator":"README.txt; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Combines pretrained sequence features with multiple downstream learners.","source_ids":["hi-enhancer-2025"],"source_locator":"2 Materials and methods/2.2 Localization of the boundaries of enhancers/2.2.1 Stacking-Auto model (paragraph 2); 1 Introduction (paragraph 2)"}],"limitations":[{"text":"The much higher epigenetic-signal-stage performance belongs to Blending-KAN and must not be presented as this sequence-based model’s result.","source_ids":["hi-enhancer-2025"],"source_locator":"3 Results/3.3 Performance of Blending-KAN on cross-cell line prediction/3.3.4 Best performance of the five kinds of signal combinations (paragraph 2); 3 Results/3.1 Performance of Blending-KAN on various signal combinations (paragraph 2)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["DNA sequences encoded with DNABERT-2","Stacking-Auto","Enhancer predictions"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["hi-enhancer-2025"],"source_locator":"Abstract (paragraph 1); 4 Discussion (paragraph 1)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","Stacking-Auto is the comparison-table label; that label does not specify an immutable weight revision.","No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-0d147487bf97be","kind":"model","name":"Promotech","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["microbes-communities"]},"source_ids":["prokbert-2024"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"Promotech","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Promotech is an existing bacterial promoter predictor in the ProkBERT evaluation.","summary_source_ids":["prokbert-2024"],"summary_source_locator":"2 Materials and methods/2.3 Application I: bacterial promoter prediction/2.3.1 Dataset overview/2.3.1.2 Dataset construction for multispecies train, test and validation sets (paragraph 6); 3 Results and discussion/3.3 ProkBERT performs accurately and robustly in promoter sequence recognition (paragraph 11)","sections":[{"title":"How the evaluated method works","body":"The official method uses a random forest on binary-encoded promoter sequences; the study evaluates it as a comparator, separate from ProkBERT training.","source_ids":["evidence-reported-base-promotech-readme-md"],"source_locator":"README.md; introduction and model-installation instructions"},{"title":"Underlying method and version boundaries","body":"Promotech’s selected model uses binary-encoded 40-bp promoter windows and a random forest. Its released workflow also scans genomes with a sliding window; the ProkBERT table does not pin the historical fitted model file.","source_ids":["evidence-reported-base-promotech-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies Promotech: E. coli sigma70 promoter prediction. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["prokbert-2024"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-034"}],"facts":[{"label":"Model type","value":"Random-forest promoter classifier; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-promotech-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"The official method uses a random forest on binary-encoded promoter sequences; the study evaluates it as a comparator, separate from ProkBERT training.","status":"source_checked","source_ids":["evidence-reported-base-promotech-readme-md"],"source_locator":"README.md; introduction and model-installation instructions"},{"label":"Biological inputs","value":"Bacterial DNA promoter windows","status":"source_checked","source_ids":["prokbert-2024"],"source_locator":"2 Materials and methods/2.3 Application I: bacterial promoter prediction/2.3.1 Dataset overview/2.3.1.2 Dataset construction for multispecies train, test and validation sets (paragraph 6); 2 Materials and methods/2.3 Application I: bacterial promoter prediction (paragraph 1)"},{"label":"Outputs","value":"Promoter classifications","status":"source_checked","source_ids":["prokbert-2024"],"source_locator":"2 Materials and methods/2.3 Application I: bacterial promoter prediction/2.3.1 Dataset overview/2.3.1.2 Dataset construction for multispecies train, test and validation sets (paragraph 6); 2 Materials and methods/2.5 Applied metrics (paragraph 1)"},{"label":"Parameters","value":"Not applicable to a neural parameter count: the selected predictor is a random forest.","status":"inapplicable","source_ids":["evidence-reported-base-promotech-readme-md"],"source_locator":"README.md; introduction and model-installation instructions"},{"label":"Known versions / configuration","value":"Promotech is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["prokbert-2024"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"The original selected model is trained across nine bacterial species and validated on four held-out species. The ProkBERT comparison does not identify a new fit or immutable model-file digest.","status":"source_checked","source_ids":["evidence-reported-base-promotech-readme-md"],"source_locator":"README.md; introduction and model-installation instructions"},{"label":"Context limits","value":"The official predictor accepts 40-bp windows and scans whole bacterial genomes with a sliding window.","status":"source_checked","source_ids":["evidence-reported-base-promotech-readme-md"],"source_locator":"README.md; introduction and model-installation instructions"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://github.com/BioinformaticsLabAtMUN/Promotech/blob/56251ad9b883ef831b4753fc623d5ec970fe65e0/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-promotech-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"GNU GPL version 3 (upstream repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-base-promotech-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-base-promotech-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Adds an established specialist baseline to the transformer comparison.","source_ids":["prokbert-2024"],"source_locator":"2 Materials and methods/2.4 Application II: phage sequence analysis/2.4.1 Phage dataset description (paragraph 5); 2 Materials and methods (paragraph 1)"}],"limitations":[{"text":"The ProkBERT comparison does not uniquely pin every implementation and fitted-model artifact; these cannot be filled from the comparator name alone.","source_ids":["prokbert-2024"],"source_locator":"1 Introduction (paragraph 8); 4 Conclusion (paragraph 6)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Bacterial DNA promoter windows","Promotech","Promoter classifications"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["prokbert-2024"],"source_locator":"2 Materials and methods (paragraph 1); 3 Results and discussion/3.3 ProkBERT performs accurately and robustly in promoter sequence recognition (paragraph 11)"},"coverage":"limited","gaps":["Promotech is the comparison-table label; that label does not specify an immutable weight revision.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-0eb4b0535b58e3","kind":"model","name":"MolAS","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["molecular-interactions"]},"source_ids":["molas-2026"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"MolAS","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"MolAS predicts which docking algorithm is likely to work best for a particular protein–ligand instance.","summary_source_ids":["molas-2026"],"summary_source_locator":"Materials and methods/Scoring function/Physicochemical plausibility (paragraph 2); Discussion/When is MolAS useful in practice? (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"Pretrained protein and ligand embeddings pass through attentional pooling and a shallow residual decoder that estimates per-algorithm performance.","source_ids":["molas-2026"],"source_locator":"Introduction (paragraph 5); Materials and methods/Model architecture (paragraph 1)"},{"title":"What was evaluated","body":"The linked evaluation record identifies MolAS: Physically valid protein–ligand pose selection. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["molas-2026"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b3-044"}],"facts":[{"label":"Model type","value":"Study-specific predictive method; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["molas-2026"],"source_locator":"Introduction (paragraph 5); Materials and methods/Model architecture (paragraph 1)"},{"label":"Architecture / procedure","value":"Pretrained protein and ligand embeddings pass through attentional pooling and a shallow residual decoder that estimates per-algorithm performance.","status":"source_checked","source_ids":["molas-2026"],"source_locator":"Introduction (paragraph 5); Materials and methods/Model architecture (paragraph 1)"},{"label":"Biological inputs","value":"Protein/ligand embeddings for an instance and a fixed candidate-solver set","status":"source_checked","source_ids":["molas-2026"],"source_locator":"Discussion/When is MolAS useful in practice? (paragraph 1); Discussion/What does MolAS reveal about docking algorithm selection? (paragraph 4)"},{"label":"Outputs","value":"Predicted solver performance and a selected docking algorithm","status":"source_checked","source_ids":["molas-2026"],"source_locator":"Materials and methods/Scoring function/Physicochemical plausibility (paragraph 2); Introduction (paragraph 2)"},{"label":"Parameters","value":"Approximately 638,000 parameters for the MolAS selector. The pretrained protein and ligand feature extractors are separate components.","status":"source_checked","source_ids":["molas-2026"],"source_locator":"Results / Comparison to MC-GNNAS-Dock"},{"label":"Known versions / configuration","value":"MolAS is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["molas-2026"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"Hundreds to a few thousand labelled complexes depending on the benchmark.","status":"source_checked","source_ids":["molas-2026"],"source_locator":"Introduction (paragraph 6); Materials and methods/Datasets and preprocessing/Extra benchmarks (paragraph 6)"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["molas-2026","evidence-reported-molas-2026-readme-md"],"source_locator":"Materials and methods/Pipeline overview; Materials and methods/Datasets and preprocessing/Dataset; Materials and methods/Datasets and preprocessing/Extra benchmarks; Materials and methods/Datasets and preprocessing/Preprocessing; Materials and methods/Model architecture; Materials and methods/Scoring function/Geometric accuracy; Materials and methods/Scoring function/Physicochemical plausibility; Materials and methods/Training objective; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/BradWangW/MolAS/blob/a6c417216ddb992f7dc513d511d0429aece4bd61/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-molas-2026-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"MIT (study repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-molas-2026-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-molas-2026-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Evaluates algorithm selection against the single-best and virtual-best solver references.","source_ids":["molas-2026"],"source_locator":"Introduction (paragraph 2); Materials and methods/Scoring function/Physicochemical plausibility (paragraph 2)"}],"limitations":[{"text":"Solver rankings can change under protocol mismatch; the paper positions MolAS as an in-domain selector for fixed pipelines.","source_ids":["molas-2026"],"source_locator":"Discussion/What does MolAS reveal about docking algorithm selection? (paragraph 3); Discussion/When is MolAS useful in practice? (paragraph 1)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Protein/ligand embeddings for an instance and a fixed candidate-solver set","MolAS","Predicted solver performance and a selected docking algorithm"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["molas-2026"],"source_locator":"Introduction (paragraph 5); Materials and methods/Model architecture (paragraph 1)"},"coverage":"limited","gaps":["MolAS is the comparison-table label; that label does not specify an immutable weight revision.","A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-10d85f2a035720","kind":"model","name":"Human-Geneformer","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["cells-tissues"]},"source_ids":["mouse-geneformer-2025"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"Human-Geneformer","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Human-Geneformer is the human-pretrained reference used to evaluate the value of mouse-specific pretraining.","summary_source_ids":["mouse-geneformer-2025"],"summary_source_locator":"Materials and methods/Construction of the mouse-Geneformer/Pretraining of the mouse-Geneformer. (paragraph 1); Materials and methods/Construction of the mouse-Geneformer/Architecture. (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"Rank Value Encoding feeds the original Geneformer transformer encoder before adaptation to the study’s downstream tasks.","source_ids":["mouse-geneformer-2025"],"source_locator":"Materials and methods/Construction of the mouse-Geneformer/Architecture. (paragraph 1); Materials and methods/Construction of the mouse-Geneformer/Pretraining of the mouse-Geneformer. (paragraph 1)"},{"title":"Underlying method and version boundaries","body":"Geneformer ranks genes by expression scaled against its pretraining corpus, then uses a transformer encoder with a masked-gene objective. V1 and V2 have different corpora, vocabularies, sizes and context limits, so a historical paper name is not replaced with today’s default checkpoint.","source_ids":["evidence-reported-base-geneformer-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies Human-Geneformer: Human thymus cell-type classification. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["mouse-geneformer-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b3-004"}],"facts":[{"label":"Model type","value":"Single-cell transformer; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-geneformer-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"Rank Value Encoding feeds the original Geneformer transformer encoder before adaptation to the study’s downstream tasks.","status":"source_checked","source_ids":["mouse-geneformer-2025"],"source_locator":"Materials and methods/Construction of the mouse-Geneformer/Architecture. (paragraph 1); Materials and methods/Construction of the mouse-Geneformer/Pretraining of the mouse-Geneformer. (paragraph 1)"},{"label":"Biological inputs","value":"Expression profiles mapped to the human model’s gene representation","status":"source_checked","source_ids":["mouse-geneformer-2025"],"source_locator":"Materials and methods/In silico perturbation of human data using mouse-Geneformer (paragraph 1); Results/Cross-species application of mouse-Geneformer through orthologous gene name conversion/Human cell type classification using mouse-Geneformer. (paragraph 1)"},{"label":"Outputs","value":"Gene/cell embeddings and downstream cell-type predictions","status":"source_checked","source_ids":["mouse-geneformer-2025"],"source_locator":"Results/Cross-species application of mouse-Geneformer through orthologous gene name conversion/Human cell type classification using mouse-Geneformer. (paragraph 1); Discussion (paragraph 1)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["mouse-geneformer-2025","evidence-reported-base-geneformer-readme-md"],"source_locator":"Materials and methods/Construction of the mouse-Geneformer/Architecture.; Materials and methods/Construction of the mouse-Geneformer/Construction of mouse-Genecoupus-20M.; Materials and methods/Construction of the mouse-Geneformer/Pretraining of the mouse-Geneformer.; Materials and methods/Fine-turning of the mouse-Geneformer; Materials and methods/Cell type classification using the mouse-Geneformer and conventional methods; Materials and methods/Cell type classification using mouse-Geneformer with and without prior learning; Materials and methods/In silico perturbation experiments; Materials and methods/Gene name conversion between mouse and human for cross-species application; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"Human-Geneformer is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["mouse-geneformer-2025"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"Approximately 30 million human single-cell transcriptomes in the original comparator described by this paper.","status":"source_checked","source_ids":["mouse-geneformer-2025"],"source_locator":"Materials and methods/Construction of the mouse-Geneformer/Pretraining of the mouse-Geneformer. (paragraph 1); Results/Development of mouse-Geneformer (paragraph 1)"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["mouse-geneformer-2025","evidence-reported-base-geneformer-readme-md"],"source_locator":"Materials and methods/Construction of the mouse-Geneformer/Architecture.; Materials and methods/Construction of the mouse-Geneformer/Construction of mouse-Genecoupus-20M.; Materials and methods/Construction of the mouse-Geneformer/Pretraining of the mouse-Geneformer.; Materials and methods/Fine-turning of the mouse-Geneformer; Materials and methods/Cell type classification using the mouse-Geneformer and conventional methods; Materials and methods/Cell type classification using mouse-Geneformer with and without prior learning; Materials and methods/In silico perturbation experiments; Materials and methods/Gene name conversion between mouse and human for cross-species application; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://huggingface.co/ctheodoris/Geneformer/blob/1f7fbae4e469a5f4f1af8c111a529cfe1b3829f5/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-geneformer-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","status":"unreported","source_ids":["evidence-reported-base-geneformer-readme-md"],"source_locator":"README.md and repository-root licence-file search"},{"label":"Weights licence","value":"Apache 2.0 is declared in the official Geneformer model-card metadata; the exact historical configuration still needs its checkpoint identity.","status":"source_checked","source_ids":["evidence-reported-base-geneformer-readme-md"],"source_locator":"README.md front matter, license field; model-version list"}],"strengths":[{"text":"Provides a cross-species transfer control for the mouse-specific model.","source_ids":["mouse-geneformer-2025"],"source_locator":"Results/Cross-species application of mouse-Geneformer through orthologous gene name conversion (paragraph 2); Materials and methods/Gene name conversion between mouse and human for cross-species application (paragraph 1)"}],"limitations":[{"text":"Gene mapping and species differences affect transfer; this historical comparator must not be silently replaced with current Geneformer V2.","source_ids":["mouse-geneformer-2025"],"source_locator":"Results/Cross-species application of mouse-Geneformer through orthologous gene name conversion (paragraph 2); Results/Cross-species application of mouse-Geneformer through orthologous gene name conversion (paragraph 1)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Expression profiles mapped to the human model’s gene representation","Human-Geneformer","Gene/cell embeddings and downstream cell-type predictions"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["mouse-geneformer-2025"],"source_locator":"Materials and methods/Construction of the mouse-Geneformer/Architecture. (paragraph 1); Materials and methods/Construction of the mouse-Geneformer/Pretraining of the mouse-Geneformer. (paragraph 1)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","Human-Geneformer is the comparison-table label; that label does not specify an immutable weight revision.","A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-13bd2a6c2d8178","kind":"model","name":"mRNABERT","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["rna-transcriptomes"]},"source_ids":["mrnabert-2025"],"links":[],"attributes":{"entity_level":"method","version":"3066-nt input","reported_name":"mRNABERT","historical_missing_metadata":{"checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"mRNABERT models complete mRNA sequences using dual tokenisation and protein-linked contrastive learning.","summary_source_ids":["mrnabert-2025"],"summary_source_locator":"Abstract (paragraph 1); Results/Overview of mRNABERT and benchmarks (paragraph 3)","sections":[{"title":"How the evaluated method works","body":"The framework combines an mRNA language model with a cross-modality contrastive objective that incorporates information from corresponding protein sequences.","source_ids":["mrnabert-2025"],"source_locator":"Abstract (paragraph 1); Methods/Model training (paragraph 3)"},{"title":"What was evaluated","body":"The linked evaluation record identifies mRNABERT: translation-efficiency prediction. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["mrnabert-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-b2-mrnabert-2025"}],"facts":[{"label":"Model type","value":"Study-specific predictive method; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["mrnabert-2025"],"source_locator":"Abstract (paragraph 1); Methods/Model training (paragraph 3)"},{"label":"Architecture / procedure","value":"The framework combines an mRNA language model with a cross-modality contrastive objective that incorporates information from corresponding protein sequences.","status":"source_checked","source_ids":["mrnabert-2025"],"source_locator":"Abstract (paragraph 1); Methods/Model training (paragraph 3)"},{"label":"Biological inputs","value":"mRNA sequences; protein correspondence supplies additional information during training","status":"source_checked","source_ids":["mrnabert-2025"],"source_locator":"Results/Applying mRNABERT to protein engineering tasks (paragraph 5); Results/Evaluating mRNABERT on CDS prediction tasks (paragraph 1)"},{"label":"Outputs","value":"mRNA representations and task-specific design/property predictions","status":"source_checked","source_ids":["mrnabert-2025"],"source_locator":"Results/Predicting splice sites and alternative polyadenylation (paragraph 5); Discussion (paragraph 4)"},{"label":"Parameters","value":"Twelve transformer layers with hidden width 768 are specified. The inspected architecture section does not give a complete parameter total.","status":"unreported","source_ids":["mrnabert-2025"],"source_locator":"Methods/Model architecture (paragraph 1); Methods/Model evaluation (paragraph 2)"},{"label":"Known versions / configuration","value":"3066-nt input","status":"source_checked","source_ids":["mrnabert-2025"],"source_locator":"Results/Evaluating the applicability using full-length mRNA sequences (paragraph 5); Table Tab2 (paragraph 1)"},{"label":"Training data / fitting","value":"Approximately 36 million mature mRNAs were collected from NCBI nt, MG-RAST, GWH and MGnify, then filtered to approximately 18 million unique sequences for the curated corpus.","status":"source_checked","source_ids":["mrnabert-2025"],"source_locator":"Methods/Training datasets/Database construction and data collection (paragraph 2); Methods/Training datasets/Pre-processing the data (paragraph 2)"},{"label":"Context limits","value":"This catalogue row is explicitly the 3,066-nt input configuration.","status":"source_checked","source_ids":["mrnabert-2025"],"source_locator":"Results/Overview of mRNABERT and benchmarks (paragraph 2); Results/Capturing multi-dimensional biological information of mRNA (paragraph 3)"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/yyly6/mRNABERT/blob/893ccc920bb9be02a4677d14d96b03126da17689/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-mrnabert-2025-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"Apache 2.0 (study repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-mrnabert-2025-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-mrnabert-2025-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"The evaluation includes UTR, CDS, RNA-binding-site and full-length mRNA-property tasks.","source_ids":["mrnabert-2025"],"source_locator":"Abstract (paragraph 1); Results/Overview of mRNABERT and benchmarks (paragraph 4)"}],"limitations":[{"text":"The 3,066-nt evaluated configuration must not be conflated with an unconstrained full-transcript model; performance remains task- and adaptation-specific.","source_ids":["mrnabert-2025"],"source_locator":"Results/Overview of mRNABERT and benchmarks (paragraph 4); Results/Applying mRNABERT to protein engineering tasks (paragraph 4)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["mRNA sequences","mRNABERT","mRNA representations and task-specific design/property predictions"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["mrnabert-2025"],"source_locator":"Abstract (paragraph 1); Methods/Model training (paragraph 3)"},"coverage":"limited","gaps":["Twelve transformer layers with hidden width 768 are specified. The inspected architecture section does not give a complete parameter total.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-148b613975b6eb","kind":"model","name":"Best frozen single-cell foundation model","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["cells-tissues"]},"source_ids":["single-cell-aging-probes-2026"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"Best frozen single-cell foundation model","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"This study compares frozen single-cell foundation-model probes with a gene-expression PCA baseline for donor age.","summary_source_ids":["single-cell-aging-probes-2026"],"summary_source_locator":"Results/Foundation models detect age, but not better than a gene-expression PCA baseline (paragraph 2); Results/Foundation models detect age, but not better than a gene-expression PCA baseline (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"Frozen scGPT and Geneformer representations feed age probes. The conventional baseline uses 50 principal components of gene expression with regression under the same donor-aware splits.","source_ids":["single-cell-aging-probes-2026"],"source_locator":"Results/Foundation models detect age, but not better than a gene-expression PCA baseline (paragraph 1); Introduction/Why use frozen foundation models, given that simpler baselines predict equally well? (paragraph 1)"},{"title":"What was evaluated","body":"The linked evaluation record identifies Best frozen single-cell foundation model: Donor-aware age-class prediction. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["single-cell-aging-probes-2026"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b3-013"}],"facts":[{"label":"Model type","value":"Study-specific predictive method; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["single-cell-aging-probes-2026"],"source_locator":"Results/Foundation models detect age, but not better than a gene-expression PCA baseline (paragraph 1); Introduction/Why use frozen foundation models, given that simpler baselines predict equally well? (paragraph 1)"},{"label":"Architecture / procedure","value":"Frozen scGPT and Geneformer representations feed age probes. The conventional baseline uses 50 principal components of gene expression with regression under the same donor-aware splits.","status":"source_checked","source_ids":["single-cell-aging-probes-2026"],"source_locator":"Results/Foundation models detect age, but not better than a gene-expression PCA baseline (paragraph 1); Introduction/Why use frozen foundation models, given that simpler baselines predict equally well? (paragraph 1)"},{"label":"Biological inputs","value":"PBMC single-cell expression data with donor identities and age labels","status":"source_checked","source_ids":["single-cell-aging-probes-2026"],"source_locator":"Methods/Statistical safeguards (paragraph 1); Methods/Datasets (paragraph 1)"},{"label":"Outputs","value":"Donor-age predictions and representation analyses","status":"source_checked","source_ids":["single-cell-aging-probes-2026"],"source_locator":"Discussion/Limitations (paragraph 1); Methods/Statistical safeguards (paragraph 1)"},{"label":"Parameters","value":"This aggregate label does not identify a unique parameter count. The paper separately describes Geneformer V1-10M, V2-104M and V2-316M in its scaling analysis; those variants are not collapsed into one model identity.","status":"unreported","source_ids":["single-cell-aging-probes-2026"],"source_locator":"Discussion / Robustness and control analyses; cross-model summary label versus Cross-Geneformer-size benchmark"},{"label":"Known versions / configuration","value":"Best frozen single-cell foundation model is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["single-cell-aging-probes-2026"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"Five PBMC cohorts containing approximately 4–5 million cells from about 2,000 donors; foundation models remain frozen.","status":"source_checked","source_ids":["single-cell-aging-probes-2026"],"source_locator":"Methods/Datasets (paragraph 1); Methods/Datasets (paragraph 2)"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["single-cell-aging-probes-2026","evidence-reported-single-cell-aging-probes-2026-readme-md"],"source_locator":"Methods/Pipeline overview; Methods/Datasets; Methods/Models, baselines, and null calibrations; Methods/Interpretability blocks; Methods/Statistical safeguards; Methods/Exploratory methylation extension; Results/Randomized-weights ablations isolate the contribution of pretraining (asymmetric across models); inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/Biodyn-AI/longevity-mechinterp/blob/5a61464632a3c3e8bebd396eb1ab17bce1dc2493/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-single-cell-aging-probes-2026-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","status":"unreported","source_ids":["evidence-reported-single-cell-aging-probes-2026-readme-md"],"source_locator":"README.md and repository-root licence-file search"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-single-cell-aging-probes-2026-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Donor-aware splitting and cell-composition matching probe important confounding mechanisms.","source_ids":["single-cell-aging-probes-2026"],"source_locator":"Methods/Statistical safeguards (paragraph 1); Methods/Interpretability blocks (paragraph 1)"}],"limitations":[{"text":"The “best frozen model” label is a selection across candidate probes, not a uniquely specified model checkpoint; PCA matches or exceeds its age-prediction performance in the reported cohorts.","source_ids":["single-cell-aging-probes-2026"],"source_locator":"Results/Foundation models detect age, but not better than a gene-expression PCA baseline (paragraph 1); Introduction/Why use frozen foundation models, given that simpler baselines predict equally well? (paragraph 1)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["PBMC single-cell expression data with donor identities and age labels","Best frozen single-cell foundation model","Donor-age predictions and representation analyses"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["single-cell-aging-probes-2026"],"source_locator":"Results/Foundation models detect age, but not better than a gene-expression PCA baseline (paragraph 1); Introduction/Why use frozen foundation models, given that simpler baselines predict equally well? (paragraph 1)"},"coverage":"limited","gaps":["This aggregate label does not identify a unique parameter count. The paper separately describes Geneformer V1-10M, V2-104M and V2-316M in its scaling analysis; those variants are not collapsed into one model identity.","Best frozen single-cell foundation model is the comparison-table label; that label does not specify an immutable weight revision.","A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-1587ab674d30a2","kind":"model","name":"AutoDock Vina holo","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["molecular-interactions"]},"source_ids":["ensemble-idp-docking-2025"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"AutoDock Vina holo","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"This holo-ensemble docking configuration evaluates small-molecule binding to α-synuclein conformations.","summary_source_ids":["ensemble-idp-docking-2025"],"summary_source_locator":"Discussion (paragraph 7); Introduction (paragraph 6)","sections":[{"title":"How the evaluated method works","body":"For each holo α-synuclein conformation, Vina docks at each residue with rigid side chains and ligand torsional freedom. The best-scoring pose is retained, then scores are min–max normalised within the method/ensemble setting.","source_ids":["ensemble-idp-docking-2025"],"source_locator":"Methods/AutoDock Vina ensemble docking. (paragraph 1); Methods/DiffDock Ensemble docking. (paragraph 1)"},{"title":"Underlying method and version boundaries","body":"AutoDock Vina performs molecular docking using an empirical scoring function and an optimisation procedure. The scoring-function choice, search space and supplied receptor structure define the evaluation setting.","source_ids":["evidence-reported-base-vina-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies AutoDock Vina holo: Intrinsically disordered protein ensemble docking. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["ensemble-idp-docking-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b3-046"}],"facts":[{"label":"Model type","value":"Empirical docking algorithm; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-vina-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"For each holo α-synuclein conformation, Vina docks at each residue with rigid side chains and ligand torsional freedom. The best-scoring pose is retained, then scores are min–max normalised within the method/ensemble setting.","status":"source_checked","source_ids":["ensemble-idp-docking-2025"],"source_locator":"Methods/AutoDock Vina ensemble docking. (paragraph 1); Methods/DiffDock Ensemble docking. (paragraph 1)"},{"label":"Biological inputs","value":"Ligands and holo protein conformations from the study’s molecular-dynamics ensemble","status":"source_checked","source_ids":["ensemble-idp-docking-2025"],"source_locator":"Introduction (paragraph 4); Methods/Docked-pose RMSD calculations. (paragraph 2)"},{"label":"Outputs","value":"Docked poses and normalised relative docking scores","status":"source_checked","source_ids":["ensemble-idp-docking-2025"],"source_locator":"Results (paragraph 3); Results/Ensemble docking accurately predicts the relative affinities of small molecules to α-synuclein. (paragraph 1)"},{"label":"Parameters","value":"Not applicable to a neural parameter count: this is an empirical scoring and conformational search procedure.","status":"inapplicable","source_ids":["ensemble-idp-docking-2025"],"source_locator":"Results/Ensemble docking protocols for intrinsically disordered proteins. (paragraph 3); Discussion (paragraph 5)"},{"label":"Known versions / configuration","value":"AutoDock Vina holo is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["ensemble-idp-docking-2025"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"No task-specific neural fitting is described for this docking comparator; the scoring function and receptor/ligand preparation define the procedure.","status":"source_checked","source_ids":["ensemble-idp-docking-2025"],"source_locator":"Results/Ensemble docking accurately reproduces IDP ligand binding modes observed in experimentally validated long timescale MD simulations. (paragraph 1); Methods/Comparing docking scores. (paragraph 2)"},{"label":"Context limits","value":"Not applicable to a sequence-token limit: receptor coordinates, ligand conformers and the defined search region determine the input.","status":"inapplicable","source_ids":["ensemble-idp-docking-2025"],"source_locator":"Results/Ensemble docking protocols for intrinsically disordered proteins. (paragraph 4); Results/Quantifying the similarity of ligand binding poses obtained from MD simulations and ensemble docking. (paragraph 5)"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://github.com/ccsb-scripps/AutoDock-Vina/blob/3c65c0b3e6c2c1d183f6a175ecb65e3c5ba91645/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-vina-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"Apache 2.0 (upstream repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-base-vina-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"Not applicable: no pretrained neural checkpoint is used.","status":"inapplicable","source_ids":["ensemble-idp-docking-2025"],"source_locator":"Data & Code Availability (paragraph 1); Discussion (paragraph 5)"}],"strengths":[{"text":"Represents an intrinsically disordered target as an ensemble rather than a single rigid structure.","source_ids":["ensemble-idp-docking-2025"],"source_locator":"Abstract (paragraph 1); Introduction (paragraph 1)"}],"limitations":[{"text":"The normalised scores have method-specific meanings and are not directly calibrated binding free energies.","source_ids":["ensemble-idp-docking-2025"],"source_locator":"Methods/Comparing docking scores. (paragraph 1); Results/Ensemble docking accurately reproduces IDP ligand binding modes observed in experimentally validated long timescale MD simulations. (paragraph 2)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Ligands and holo protein conformations from the study’s molecular-dynamics ensemble","AutoDock Vina holo","Docked poses and normalised relative docking scores"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["ensemble-idp-docking-2025"],"source_locator":"Methods/AutoDock Vina ensemble docking. (paragraph 1); Methods/DiffDock Ensemble docking. (paragraph 1)"},"coverage":"limited","gaps":["AutoDock Vina holo is the comparison-table label; that label does not specify an immutable weight revision.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-177f32ce8189a0","kind":"model","name":"Gene-expression PCA","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["cells-tissues"]},"source_ids":["single-cell-aging-probes-2026"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"Gene-expression PCA","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"The gene-expression PCA baseline predicts donor age from 50 principal components.","summary_source_ids":["single-cell-aging-probes-2026"],"summary_source_locator":"Results/Foundation models detect age, but not better than a gene-expression PCA baseline (paragraph 2); Discussion/Robustness and control analyses (paragraph 7)","sections":[{"title":"How the evaluated method works","body":"PCA fits a linear low-dimensional representation of gene expression; a regression probe uses the same donor-aware partitions as the foundation-model comparisons.","source_ids":["single-cell-aging-probes-2026"],"source_locator":"Methods/Models, baselines, and null calibrations (paragraph 2); Results/Foundation models detect age, but not better than a gene-expression PCA baseline (paragraph 1)"},{"title":"What was evaluated","body":"The linked evaluation record identifies Gene-expression PCA: Donor-aware age-class prediction. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["single-cell-aging-probes-2026"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b3-014"}],"facts":[{"label":"Model type","value":"Linear predictive procedure; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["single-cell-aging-probes-2026"],"source_locator":"Methods/Models, baselines, and null calibrations (paragraph 2); Results/Foundation models detect age, but not better than a gene-expression PCA baseline (paragraph 1)"},{"label":"Architecture / procedure","value":"PCA fits a linear low-dimensional representation of gene expression; a regression probe uses the same donor-aware partitions as the foundation-model comparisons.","status":"source_checked","source_ids":["single-cell-aging-probes-2026"],"source_locator":"Methods/Models, baselines, and null calibrations (paragraph 2); Results/Foundation models detect age, but not better than a gene-expression PCA baseline (paragraph 1)"},{"label":"Biological inputs","value":"Single-cell gene expression aggregated/handled under the study’s donor-aware protocol","status":"source_checked","source_ids":["single-cell-aging-probes-2026"],"source_locator":"Discussion/Robustness and control analyses (paragraph 7); Methods/Statistical safeguards (paragraph 1)"},{"label":"Outputs","value":"Age predictions","status":"source_checked","source_ids":["single-cell-aging-probes-2026"],"source_locator":"Conclusion (paragraph 1); Discussion/Limitations (paragraph 7)"},{"label":"Parameters","value":"50 principal components; this is not a pretrained neural parameter count.","status":"source_checked","source_ids":["single-cell-aging-probes-2026"],"source_locator":"Results/Randomized-weights ablations isolate the contribution of pretraining (asymmetric across models) (paragraph 2); Discussion/Robustness and control analyses (paragraph 7)"},{"label":"Known versions / configuration","value":"Gene-expression PCA is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["single-cell-aging-probes-2026"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"Fit within the same donor-aware training splits used for the competing probes.","status":"source_checked","source_ids":["single-cell-aging-probes-2026"],"source_locator":"Methods/Datasets (paragraph 1); Methods/Models, baselines, and null calibrations (paragraph 2)"},{"label":"Context limits","value":"The baseline uses 50 gene-expression principal components; a sequence-token context is inapplicable.","status":"source_checked","source_ids":["single-cell-aging-probes-2026"],"source_locator":"Results/Foundation models detect age, but not better than a gene-expression PCA baseline (paragraph 2); Table Tab2 (paragraph 1)"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/Biodyn-AI/longevity-mechinterp/blob/5a61464632a3c3e8bebd396eb1ab17bce1dc2493/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-single-cell-aging-probes-2026-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","status":"unreported","source_ids":["evidence-reported-single-cell-aging-probes-2026-readme-md"],"source_locator":"README.md and repository-root licence-file search"},{"label":"Weights licence","value":"Not applicable to pretrained model weights; PCA and the regression probe are fitted experiment artifacts.","status":"inapplicable","source_ids":["single-cell-aging-probes-2026"],"source_locator":"Results/Randomized-weights ablations isolate the contribution of pretraining (asymmetric across models) (paragraph 2); Results/Randomized-weights ablations isolate the contribution of pretraining (asymmetric across models) (paragraph 3)"}],"strengths":[{"text":"Matches or exceeds the best frozen foundation-model probe across the five reported PBMC cohorts.","source_ids":["single-cell-aging-probes-2026"],"source_locator":"Results/Foundation models detect age, but not better than a gene-expression PCA baseline (paragraph 2); Results/Foundation models detect age, but not better than a gene-expression PCA baseline (paragraph 1)"}],"limitations":[{"text":"The baseline’s performance is a comparator under this experiment, not a biological ceiling on age prediction.","source_ids":["single-cell-aging-probes-2026"],"source_locator":"Results/The strongest aging signal: how it changes with sample size and under composition matching (paragraph 1); Discussion/Robustness and control analyses (paragraph 7)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Single-cell gene expression aggregated/handled under the study’s donor-aware protocol","Gene-expression PCA","Age predictions"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["single-cell-aging-probes-2026"],"source_locator":"Methods/Models, baselines, and null calibrations (paragraph 2); Results/Foundation models detect age, but not better than a gene-expression PCA baseline (paragraph 1)"},"coverage":"limited","gaps":["Gene-expression PCA is the comparison-table label; that label does not specify an immutable weight revision.","No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-1a67087ac262c5","kind":"model","name":"Nucleotide Transformer + NN (middle)","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["dna-genomes"]},"source_ids":["fusion-breakpoint-foundation-models-2026"],"links":[],"attributes":{"entity_level":"method","version":"not stated in table","reported_name":"Nucleotide Transformer + NN (middle)","historical_missing_metadata":{"checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"This fusion-breakpoint classifier uses frozen Nucleotide Transformer features with a separately trained neural-network head.","summary_source_ids":["fusion-breakpoint-foundation-models-2026"],"summary_source_locator":"Discussion (paragraph 2); Results/Classification performance (paragraph 2)","sections":[{"title":"How the evaluated method works","body":"DNA windows around candidate fusion breakpoints are encoded by a frozen genomic foundation model. Only the lightweight downstream classifier is trained; the middle-labelled representation is the exact configuration of this row.","source_ids":["fusion-breakpoint-foundation-models-2026"],"source_locator":"Methods/Evaluation metrics (paragraph 2); Methods (paragraph 1)"},{"title":"What was evaluated","body":"The linked evaluation record identifies Nucleotide Transformer + NN (middle): gene fusion breakpoint classification. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["fusion-breakpoint-foundation-models-2026"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-b2-fusion-breakpoint-foundation-models-2026"}],"facts":[{"label":"Model type","value":"Study-specific predictive method; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["fusion-breakpoint-foundation-models-2026"],"source_locator":"Methods/Evaluation metrics (paragraph 2); Methods (paragraph 1)"},{"label":"Architecture / procedure","value":"DNA windows around candidate fusion breakpoints are encoded by a frozen genomic foundation model. Only the lightweight downstream classifier is trained; the middle-labelled representation is the exact configuration of this row.","status":"source_checked","source_ids":["fusion-breakpoint-foundation-models-2026"],"source_locator":"Methods/Evaluation metrics (paragraph 2); Methods (paragraph 1)"},{"label":"Biological inputs","value":"DNA sequence around fusion breakpoints","status":"source_checked","source_ids":["fusion-breakpoint-foundation-models-2026"],"source_locator":"Methods/Evaluation metrics (paragraph 2); Discussion/Limitations and future directions (paragraph 3)"},{"label":"Outputs","value":"Fusion-breakpoint classification","status":"source_checked","source_ids":["fusion-breakpoint-foundation-models-2026"],"source_locator":"Methods/Evaluation metrics (paragraph 2); Methods/DNABERT2 (BERT) (paragraph 3)"},{"label":"Parameters","value":"500 M-parameter Nucleotide Transformer backbone, plus a separately fitted neural classifier.","status":"source_checked","source_ids":["fusion-breakpoint-foundation-models-2026"],"source_locator":"Table1; NT column, Parameters row"},{"label":"Known versions / configuration","value":"Nucleotide Transformer + NN (middle) is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["fusion-breakpoint-foundation-models-2026"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"Approximately 52,000 sequences from the curated FusionAI dataset; foundation-model weights remain frozen.","status":"source_checked","source_ids":["fusion-breakpoint-foundation-models-2026"],"source_locator":"Methods/Evaluation metrics (paragraph 2); Discussion/Limitations and future directions (paragraph 1)"},{"label":"Context limits","value":"10-kbp DNA windows in the study; model-specific embedding selection remains part of the evaluated protocol.","status":"source_checked","source_ids":["fusion-breakpoint-foundation-models-2026"],"source_locator":"Results/Classification performance (paragraph 1); Results/Computational efficiency (paragraph 1)"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/kbi-fbmi/articles--2026fusionEmbBenchmark/blob/085a6d7d2f899b0f62d764f35d1248b2eda567da/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-fusion-breakpoint-foundation-models-2026-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"MIT (study repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-fusion-breakpoint-foundation-models-2026-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-fusion-breakpoint-foundation-models-2026-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Separates representation extraction from the training of a small task-specific classifier.","source_ids":["fusion-breakpoint-foundation-models-2026"],"source_locator":"Conclusion (paragraph 3); Discussion (paragraph 6)"}],"limitations":[{"text":"A successful frozen-embedding classifier does not establish zero-shot performance or clinical validity of a fusion caller.","source_ids":["fusion-breakpoint-foundation-models-2026"],"source_locator":"Results/Classification performance (paragraph 3); Results/Visual assessment of embedding quality (paragraph 2)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["DNA sequence around fusion breakpoints","Nucleotide Transformer + NN (middle)","Fusion-breakpoint classification"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["fusion-breakpoint-foundation-models-2026"],"source_locator":"Methods/Evaluation metrics (paragraph 2); Methods (paragraph 1)"},"coverage":"limited","gaps":["Nucleotide Transformer + NN (middle) is the comparison-table label; that label does not specify an immutable weight revision.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-1b5fa066945d3d","kind":"model","name":"R3Design","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["rna-transcriptomes"]},"source_ids":["r3design-2025"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"R3Design","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"R3Design generates RNA sequences conditioned on a tertiary backbone.","summary_source_ids":["r3design-2025"],"summary_source_locator":"Methods/The R3Design framework (paragraph 4); Methods/The R3Design framework (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"Three PiGNN layers, adapted from PiFold, encode a graph of the RNA backbone and spatial nucleotide relationships before sequence prediction.","source_ids":["r3design-2025"],"source_locator":"Methods/The R3Design framework (paragraph 4); Methods/The R3Design framework (paragraph 2)"},{"title":"What was evaluated","body":"The linked evaluation record identifies R3Design: RNA sequence design. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["r3design-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b4-006"}],"facts":[{"label":"Model type","value":"Graph-based predictive method; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["r3design-2025"],"source_locator":"Methods/The R3Design framework (paragraph 4); Methods/The R3Design framework (paragraph 2)"},{"label":"Architecture / procedure","value":"Three PiGNN layers, adapted from PiFold, encode a graph of the RNA backbone and spatial nucleotide relationships before sequence prediction.","status":"source_checked","source_ids":["r3design-2025"],"source_locator":"Methods/The R3Design framework (paragraph 4); Methods/The R3Design framework (paragraph 2)"},{"label":"Biological inputs","value":"RNA tertiary-structure backbone graphs","status":"source_checked","source_ids":["r3design-2025"],"source_locator":"Methods/The R3Design framework (paragraph 1); Methods/The R3Design framework (paragraph 5)"},{"label":"Outputs","value":"RNA nucleotide sequences designed for the supplied backbone","status":"source_checked","source_ids":["r3design-2025"],"source_locator":"Methods/Graph-based RNA tertiary structure modeling (paragraph 9); Methods/Graph-based RNA tertiary structure modeling (paragraph 7)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["r3design-2025","evidence-reported-r3design-2025-readme-md"],"source_locator":"Methods/Preliminaries; Methods/Graph-based RNA tertiary structure modeling; Methods/The R3Design framework; inspected for aggregate parameter count (component sizes are not added without an exact configuration); readme.md at pinned repository revision"},{"label":"Known versions / configuration","value":"R3Design is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["r3design-2025"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"2,218 representative RNA structures from PDB/RNASolo, filtered to resolution below 4Å and at most 500 nucleotides. A structural-similarity split assigns 1,774 training,223 validation and 221 test structures.","status":"source_checked","source_ids":["r3design-2025"],"source_locator":"Results/R3Design designs RNA sequences with high sequence-level fidelity (paragraph 1); Results/The designed sequence can fold into desirable tertiary structure (paragraph 2)"},{"label":"Context limits","value":"The curated training/evaluation set excludes RNAs longer than 500 nucleotides; this is a dataset restriction, not a proved inference maximum.","status":"source_checked","source_ids":["r3design-2025"],"source_locator":"Results/R3Design designs RNA sequences with high sequence-level fidelity (paragraph 1); Table TB2 (paragraph 1)"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/A4Bio/R3Design/blob/c05dc4b35100949011c77c08f61a9e80280987ce/readme.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-r3design-2025-readme-md"],"source_locator":"readme.md; installation, model download and usage instructions"},{"label":"Code licence","value":"No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","status":"unreported","source_ids":["evidence-reported-r3design-2025-readme-md"],"source_locator":"readme.md and repository-root licence-file search"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-r3design-2025-readme-md"],"source_locator":"readme.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Evaluates native-backbone sequence recovery and checks generated sequences with structure-prediction models.","source_ids":["r3design-2025"],"source_locator":"Methods/The R3Design framework (paragraph 5); Methods/The R3Design framework (paragraph 1)"}],"limitations":[{"text":"Computational refolding is not experimental validation that a designed RNA adopts the intended structure or function.","source_ids":["r3design-2025"],"source_locator":"Methods/The R3Design framework (paragraph 5); Results/The designed sequence can fold into desirable tertiary structure (paragraph 1)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["RNA tertiary-structure backbone graphs","R3Design","RNA nucleotide sequences designed for the supplied backbone"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["r3design-2025"],"source_locator":"Methods/The R3Design framework (paragraph 4); Methods/The R3Design framework (paragraph 2)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","R3Design is the comparison-table label; that label does not specify an immutable weight revision.","No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-1be5c4b7c52a41","kind":"model","name":"scVI + scANVI","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["cells-tissues"]},"source_ids":["scalr-2025"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"scVI + scANVI","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"The scVI plus scANVI workflow is a probabilistic single-cell comparator in the scaLR study.","summary_source_ids":["scalr-2025"],"summary_source_locator":"Methods/scaLR performance comparison with different pipelines (paragraph 6); Results/Comparison of scaLR extracted top 3,500 features as input for different pipelines (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"Variational autoencoders learn low-dimensional cell representations; the scANVI stage uses annotation information in the combined workflow.","source_ids":["scalr-2025"],"source_locator":"Methods/scaLR performance comparison with different pipelines (paragraph 6); Methods/scaLR performance comparison with different pipelines (paragraph 3)"},{"title":"Underlying method and version boundaries","body":"scvi-tools contains probabilistic models for single-cell analyses. scVI, scANVI and study-specific downstream heads are distinct procedures even when distributed through the same software package.","source_ids":["evidence-reported-base-scvi-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies scVI + scANVI: PBMC cell-type classification. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["scalr-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b3-016"}],"facts":[{"label":"Model type","value":"Probabilistic single-cell model; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-scvi-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"Variational autoencoders learn low-dimensional cell representations; the scANVI stage uses annotation information in the combined workflow.","status":"source_checked","source_ids":["scalr-2025"],"source_locator":"Methods/scaLR performance comparison with different pipelines (paragraph 6); Methods/scaLR performance comparison with different pipelines (paragraph 3)"},{"label":"Biological inputs","value":"Single-cell expression profiles and available cell labels","status":"source_checked","source_ids":["scalr-2025"],"source_locator":"Introduction (paragraph 1); Introduction (paragraph 4)"},{"label":"Outputs","value":"Latent cell representations and cell annotations","status":"source_checked","source_ids":["scalr-2025"],"source_locator":"Methods/scaLR performance comparison with different pipelines (paragraph 6); Results/Performance and analysis cost comparison (paragraph 1)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["scalr-2025","evidence-reported-base-scvi-readme-md"],"source_locator":"Methods/Single-cell RNA-Seq data selection and download; Methods/The scaLR platform; Methods/The scaLR platform/Data processing; Methods/The scaLR platform/Feature extraction; Methods/The scaLR platform/Training; Methods/The scaLR platform/Evaluation and downstream analysis; Methods/scaLR performance comparison with different pipelines; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"Two network layers,30 latent dimensions, scVI 100 epochs and scANVI 25 epochs (Table 2 footnote b).","status":"source_checked","source_ids":["scalr-2025"],"source_locator":"Table2 footnote b"},{"label":"Training data / fitting","value":"Fit the VAE and annotation model on the PBMCs-BS expression data: scVI 100 epochs and scANVI 25 epochs.","status":"source_checked","source_ids":["scalr-2025"],"source_locator":"Table2 footnote b"},{"label":"Context limits","value":"Table 2 uses all gene features and a 30-dimensional latent space; Table 3 separately uses the top 3,500 features.","status":"source_checked","source_ids":["scalr-2025"],"source_locator":"Tables2–3; feature and latent-dimension settings"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://github.com/scverse/scvi-tools/blob/73b28e44223621470e582a81a102c107bb22678b/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-scvi-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"BSD 3-Clause (upstream repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-base-scvi-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-base-scvi-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Provides an established probabilistic integration/annotation reference.","source_ids":["scalr-2025"],"source_locator":"Introduction (paragraph 1); Introduction (paragraph 2)"}],"limitations":[{"text":"This is a two-model workflow, not scVI alone; label availability and preprocessing must match the evaluated protocol.","source_ids":["scalr-2025"],"source_locator":"Results/Comparison of top-K differential gene expression features between full and test set samples (paragraph 2); Methods/scaLR performance comparison with different pipelines (paragraph 6)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Single-cell expression profiles and available cell labels","scVI + scANVI","Latent cell representations and cell annotations"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["scalr-2025"],"source_locator":"Methods/scaLR performance comparison with different pipelines (paragraph 6); Methods/scaLR performance comparison with different pipelines (paragraph 3)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-1d2aa9880a1c77","kind":"model","name":"ADAR-GPT continual","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["rna-transcriptomes"]},"source_ids":["adar-gpt-editing-2026"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"ADAR-GPT continual","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"ADAR-GPT is a supervised RNA-editing classifier obtained by continually fine-tuning GPT-4o-mini on marked RNA sequence windows.","summary_source_ids":["adar-gpt-editing-2026"],"summary_source_locator":"Results and Analysis (paragraph 1); Materials and Methods/Fine-Tuning Protocol. (paragraph 4)","sections":[{"title":"How the evaluated method works","body":"The continual configuration first trains on progressively ordered editing-threshold data and then refines on sites meeting the 15% editing threshold. This is a task-specific fine-tuned pipeline, not an evaluation of the unadapted general-purpose model.","source_ids":["adar-gpt-editing-2026"],"source_locator":"Methodology/Model Training Approaches—Static vs. Continual Fine-Tuning. (paragraph 4); Contribution. (paragraph 1)"},{"title":"What was evaluated","body":"The linked evaluation record identifies ADAR-GPT continual: A-to-I RNA editing site prediction. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["adar-gpt-editing-2026"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b4-005"}],"facts":[{"label":"Model type","value":"Study-specific predictive method; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["adar-gpt-editing-2026"],"source_locator":"Methodology/Model Training Approaches—Static vs. Continual Fine-Tuning. (paragraph 4); Contribution. (paragraph 1)"},{"label":"Architecture / procedure","value":"The continual configuration first trains on progressively ordered editing-threshold data and then refines on sites meeting the 15% editing threshold. This is a task-specific fine-tuned pipeline, not an evaluation of the unadapted general-purpose model.","status":"source_checked","source_ids":["adar-gpt-editing-2026"],"source_locator":"Methodology/Model Training Approaches—Static vs. Continual Fine-Tuning. (paragraph 4); Contribution. (paragraph 1)"},{"label":"Biological inputs","value":"201-nt RNA windows with the target adenosine explicitly marked","status":"source_checked","source_ids":["adar-gpt-editing-2026"],"source_locator":"Materials and Methods/Input Representation. (paragraph 1); Methodology/Data Collection and Preprocessing—Liver GTEx Dataset. (paragraph 5)"},{"label":"Outputs","value":"A-to-I editing-site classification at the study’s editing threshold","status":"source_checked","source_ids":["adar-gpt-editing-2026"],"source_locator":"Materials and Methods/Dataset Design and Labeling. (paragraph 1); Contribution. (paragraph 1)"},{"label":"Parameters","value":"The paper identifies GPT-4o-mini but does not disclose the backbone parameter count.","status":"unreported","source_ids":["adar-gpt-editing-2026"],"source_locator":"Results and Analysis (paragraph 1); Materials and Methods/Fine-Tuning Protocol. (paragraph 4)"},{"label":"Known versions / configuration","value":"ADAR-GPT continual is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["adar-gpt-editing-2026"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"GTEx liver data from 131 samples; curriculum thresholds 1%, 5%, 10% and 15%, followed by refinement on 15% sites.","status":"source_checked","source_ids":["adar-gpt-editing-2026"],"source_locator":"Methodology/Model Training Approaches—Static vs. Continual Fine-Tuning. (paragraph 4); Methodology/Data Collection and Preprocessing—Liver GTEx Dataset. (paragraph 3)"},{"label":"Context limits","value":"201 nucleotides: 100 upstream, central adenosine and 100 downstream","status":"source_checked","source_ids":["adar-gpt-editing-2026"],"source_locator":"Methodology/Data Collection and Preprocessing—Liver GTEx Dataset. (paragraph 5); Table t03 (paragraph 1)"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/Scientific-Computing-Lab/ADAR-GPT/blob/c0fd23679922d91a45520455d4ca0202a5ca609f/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-adar-gpt-editing-2026-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"MIT (study repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-adar-gpt-editing-2026-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-adar-gpt-editing-2026-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Explicitly controls the target position and uses a threshold curriculum to adapt the classifier to the final editing definition.","source_ids":["adar-gpt-editing-2026"],"source_locator":"Previous Work (paragraph 5); Abstract (paragraph 2)"}],"limitations":[{"text":"Evaluation is restricted to the study’s GTEx liver setting; broader tissue transfer and independent experimental validation remain separate questions.","source_ids":["adar-gpt-editing-2026"],"source_locator":"Abstract (paragraph 1); Discussion and Conclusions (paragraph 5)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["201-nt RNA windows with the target adenosine explicitly marked","ADAR-GPT continual","A-to-I editing-site classification at the study’s editing threshold"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["adar-gpt-editing-2026"],"source_locator":"Methodology/Model Training Approaches—Static vs. Continual Fine-Tuning. (paragraph 4); Contribution. (paragraph 1)"},"coverage":"limited","gaps":["The paper identifies GPT-4o-mini but does not disclose the backbone parameter count.","ADAR-GPT continual is the comparison-table label; that label does not specify an immutable weight revision.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-1e51ccbfd2de61","kind":"model","name":"Vina","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["molecular-interactions"]},"source_ids":["nmdn-2025"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"Vina","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"AutoDock Vina is a conventional docking/scoring comparator in the DiffDock-NMDN study.","summary_source_ids":["nmdn-2025"],"summary_source_locator":"Data Sets/Evaluation Metrics/CASF-2016 (paragraph 3); DiffDock-NMDN\nBlind Docking and Virtual Screening\nProtocol (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"A docking search and empirical scoring function estimate ligand poses and scores from receptor/ligand structures.","source_ids":["nmdn-2025"],"source_locator":"Methods/Normalized Mixture Density Network Module/Training Loss for the NMDN Modules (paragraph 3); Conclusions (paragraph 1)"},{"title":"Underlying method and version boundaries","body":"AutoDock Vina performs molecular docking using an empirical scoring function and an optimisation procedure. The scoring-function choice, search space and supplied receptor structure define the evaluation setting.","source_ids":["evidence-reported-base-vina-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies Vina: Protein–ligand virtual screening. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["nmdn-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-044"}],"facts":[{"label":"Model type","value":"Empirical docking algorithm; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-vina-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"A docking search and empirical scoring function estimate ligand poses and scores from receptor/ligand structures.","status":"source_checked","source_ids":["nmdn-2025"],"source_locator":"Methods/Normalized Mixture Density Network Module/Training Loss for the NMDN Modules (paragraph 3); Conclusions (paragraph 1)"},{"label":"Biological inputs","value":"Receptor and ligand structural information under the study’s search-box protocol","status":"source_checked","source_ids":["nmdn-2025"],"source_locator":"Methods/Interaction\nModule/Calculation of Protein–Ligand Pair\nContributions (paragraph 8); Methods/Interaction\nModule/Calculation of Protein–Ligand Pair\nContributions (paragraph 6)"},{"label":"Outputs","value":"Docked poses and empirical docking scores","status":"source_checked","source_ids":["nmdn-2025"],"source_locator":"Results and Discussion/DiffDock-NMDN Blind Docking Protocol (paragraph 4); Data Sets/Evaluation Metrics/CASF-2016 (paragraph 2)"},{"label":"Parameters","value":"Not applicable to a neural parameter count: this is an empirical scoring and conformational search procedure.","status":"inapplicable","source_ids":["nmdn-2025"],"source_locator":"Introduction (paragraph 2); Introduction (paragraph 1)"},{"label":"Known versions / configuration","value":"Vina is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["nmdn-2025"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"No task-specific neural fitting is described for this docking comparator; the scoring function and receptor/ligand preparation define the procedure.","status":"source_checked","source_ids":["nmdn-2025"],"source_locator":"Methods/Normalized Mixture Density Network Module/Training Loss for the NMDN Modules (paragraph 3); Methods/Normalized Mixture Density Network Module/Training Loss for the NMDN Modules (paragraph 5)"},{"label":"Context limits","value":"Not applicable to a sequence-token limit: receptor coordinates, ligand conformers and the defined search region determine the input.","status":"inapplicable","source_ids":["nmdn-2025"],"source_locator":"Methods/Normalized Mixture Density Network Module/Inference (paragraph 1); Methods/Normalized Mixture Density Network Module/Inference (paragraph 4)"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://github.com/ccsb-scripps/AutoDock-Vina/blob/3c65c0b3e6c2c1d183f6a175ecb65e3c5ba91645/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-vina-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"Apache 2.0 (upstream repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-base-vina-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"Not applicable: no pretrained neural checkpoint is used.","status":"inapplicable","source_ids":["nmdn-2025"],"source_locator":"Methods/Normalized Mixture Density Network Module/Protein–ligand NMDN Module Architecture (paragraph 3); Methods/Normalized Mixture Density Network Module/Inference (paragraph 6)"}],"strengths":[{"text":"Provides a non-foundation-model docking reference.","source_ids":["nmdn-2025"],"source_locator":"Methods/Normalized Mixture Density Network Module/Inference (paragraph 2); Ablation\nStudy/Normalization Terms and Selection of Reference\nDistance (paragraph 1)"}],"limitations":[{"text":"Docking scores are approximations and must not be interpreted as experimentally measured affinities.","source_ids":["nmdn-2025"],"source_locator":"Methods/Interaction\nModule/Enhancing the Virtual Screening Capability\nof pKd Score (paragraph 2); Data Sets/Evaluation Metrics/CASF-2016 (paragraph 2)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Receptor and ligand structural information under the study’s search-box protocol","Vina","Docked poses and empirical docking scores"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["nmdn-2025"],"source_locator":"Methods/Normalized Mixture Density Network Module/Training Loss for the NMDN Modules (paragraph 3); Conclusions (paragraph 1)"},"coverage":"limited","gaps":["Vina is the comparison-table label; that label does not specify an immutable weight revision.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-23cb15b93c00ff","kind":"model","name":"iPro70-FMWin","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["microbes-communities"]},"source_ids":["cyaprombert-2022"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"iPro70-FMWin","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"iPro70-FMWin is an established σ70 promoter predictor used as a webserver comparator in the CyaPromBERT paper.","summary_source_ids":["cyaprombert-2022"],"summary_source_locator":"Results and discussion/Evaluating model performance compared to existing promoter prediction models using independent datasets from E. coli\n (paragraph 3); Results and discussion/Evaluating model performance compared to existing promoter prediction models using independent datasets from E. coli\n (paragraph 2)","sections":[{"title":"How the evaluated method works","body":"The original method extracts multi-window sequence features and uses AdaBoost-based feature selection, retaining 27 common features from 22,595 candidates before classifier evaluation. The CyaPromBERT study calls the existing server and uses its returned probabilities.","source_ids":["evidence-reported-ipro70-original"],"source_locator":"Materials and methods / Benchmark dataset; Feature selection"},{"title":"What was evaluated","body":"The linked evaluation record identifies iPro70-FMWin: E. coli sigma70 promoter prediction. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["cyaprombert-2022"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-036"}],"facts":[{"label":"Model type","value":"Study-specific predictive method; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["cyaprombert-2022"],"source_locator":"Materials and methods/Model training (paragraph 3); Results and discussion/Evaluating model performance compared to existing promoter prediction models using independent datasets from E. coli\n (paragraph 2)"},{"label":"Architecture / procedure","value":"The original method extracts multi-window sequence features and uses AdaBoost-based feature selection, retaining 27 common features from 22,595 candidates before classifier evaluation. The CyaPromBERT study calls the existing server and uses its returned probabilities.","status":"source_checked","source_ids":["evidence-reported-ipro70-original"],"source_locator":"Materials and methods / Benchmark dataset; Feature selection"},{"label":"Biological inputs","value":"E. coli σ70 promoter/non-promoter DNA windows","status":"source_checked","source_ids":["cyaprombert-2022"],"source_locator":"Materials and methods/Datasets (paragraph 2); Results and discussion/Evaluating model performance compared to existing promoter prediction models using independent datasets from E. coli\n (paragraph 1)"},{"label":"Outputs","value":"Promoter probabilities and class predictions","status":"source_checked","source_ids":["cyaprombert-2022"],"source_locator":"Materials and methods/Datasets (paragraph 2); Results and discussion/Evaluating model performance compared to existing promoter prediction models using independent datasets from E. coli\n (paragraph 1)"},{"label":"Parameters","value":"A neural parameter count is inapplicable. The original feature-selection procedure retains 27 features; the later server’s fitted classifier artifact is not pinned.","status":"inapplicable","source_ids":["evidence-reported-ipro70-original"],"source_locator":"Materials and methods / Feature selection"},{"label":"Known versions / configuration","value":"iPro70-FMWin is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["cyaprombert-2022"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"The original method paper uses 2,141 E. coli K-12 sequences from a RegulonDB 9.0-derived collection, including 741 σ70 promoters, with cross-validation. This is the underlying method’s training description; the later server build is not pinned.","status":"source_checked","source_ids":["evidence-reported-ipro70-original"],"source_locator":"Materials and methods / Benchmark dataset; Feature selection"},{"label":"Context limits","value":"81-bp windows spanning 60 bases upstream and 20 downstream of the transcription start site.","status":"source_checked","source_ids":["evidence-reported-ipro70-original"],"source_locator":"Materials and methods / Benchmark dataset"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/hanepira/TSSnote-CyaPromBert/blob/e86f5449e2e2af3fead1b418ba721f38feb61318/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-cyaprombert-2022-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","status":"unreported","source_ids":["evidence-reported-cyaprombert-2022-readme-md"],"source_locator":"README.md and repository-root licence-file search"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-cyaprombert-2022-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Provides a task-specialist comparator on the same independent E. coli dataset.","source_ids":["cyaprombert-2022"],"source_locator":"Results and discussion/Evaluating model performance compared to existing promoter prediction models using independent datasets from E. coli\n (paragraph 1); Materials and methods/Datasets (paragraph 2)"}],"limitations":[{"text":"The benchmark does not fully specify a reproducible server build or weight revision; webserver results are not a BERT-model configuration.","source_ids":["cyaprombert-2022"],"source_locator":"Results and discussion/Evaluating model performance compared to existing promoter prediction models using independent datasets from E. coli\n (paragraph 3); Results and discussion/Evaluating model performance compared to existing promoter prediction models using independent datasets from E. coli\n (paragraph 1)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["E. coli σ70 promoter/non-promoter DNA windows","iPro70-FMWin","Promoter probabilities and class predictions"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["cyaprombert-2022"],"source_locator":"Materials and methods/Model training (paragraph 3); Results and discussion/Evaluating model performance compared to existing promoter prediction models using independent datasets from E. coli\n (paragraph 2)"},"coverage":"limited","gaps":["iPro70-FMWin is the comparison-table label; that label does not specify an immutable weight revision.","No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-27dca28a87cf3c","kind":"model","name":"VIBRANT","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["microbes-communities"]},"source_ids":["viral-contig-simulation-2021"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"VIBRANT","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"This viral-contig detector is compared on simulated metagenomes with varying taxonomic composition and complexity.","summary_source_ids":["viral-contig-simulation-2021"],"summary_source_locator":"Conclusion (paragraph 1); Methods/Building simulated Illumina metagenomes (paragraph 3)","sections":[{"title":"How the evaluated method works","body":"The benchmark runs existing viral-identification tools on assembled contigs and evaluates both standalone viruses and integrated prophages.","source_ids":["viral-contig-simulation-2021"],"source_locator":"Methods/Tools used in simulation study (paragraph 1); Methods/Tool performance scoring (paragraph 1)"},{"title":"What was evaluated","body":"The linked evaluation record identifies VIBRANT: Simulated prophage-contig detection. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["viral-contig-simulation-2021"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b3-025"}],"facts":[{"label":"Model type","value":"Study-specific predictive method; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["viral-contig-simulation-2021"],"source_locator":"Methods/Tools used in simulation study (paragraph 1); Methods/Tool performance scoring (paragraph 1)"},{"label":"Architecture / procedure","value":"The benchmark runs existing viral-identification tools on assembled contigs and evaluates both standalone viruses and integrated prophages.","status":"source_checked","source_ids":["viral-contig-simulation-2021"],"source_locator":"Methods/Tools used in simulation study (paragraph 1); Methods/Tool performance scoring (paragraph 1)"},{"label":"Biological inputs","value":"Metagenomic contigs","status":"source_checked","source_ids":["viral-contig-simulation-2021"],"source_locator":"Methods/Building simulated Illumina metagenomes (paragraph 3); Results/Tool performance by contig length (paragraph 1)"},{"label":"Outputs","value":"Viral-contig or prophage predictions","status":"source_checked","source_ids":["viral-contig-simulation-2021"],"source_locator":"Results/Prophage identification performance (paragraph 1); Discussion (paragraph 1)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["viral-contig-simulation-2021","evidence-reported-viral-contig-simulation-2021-readme-md"],"source_locator":"Methods/Phybrid, a hybrid gene content and nucleotide feature set for viral classification; Methods/Gene content feature set creation; Methods/Model and hyperparameter selection; Methods/Building simulated Illumina metagenomes; Methods/Integrated prophage identification; Methods/Tools used in simulation study; Methods/Tool performance scoring; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"VIBRANT is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["viral-contig-simulation-2021"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"The study evaluates released software on simulated assembled contigs, rather than fitting this comparator on the study’s labels. The exact upstream training/reference inventory is not restated.","status":"source_checked","source_ids":["viral-contig-simulation-2021"],"source_locator":"Methods / Tools used in simulation study"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["viral-contig-simulation-2021","evidence-reported-viral-contig-simulation-2021-readme-md"],"source_locator":"Methods/Phybrid, a hybrid gene content and nucleotide feature set for viral classification; Methods/Gene content feature set creation; Methods/Model and hyperparameter selection; Methods/Building simulated Illumina metagenomes; Methods/Integrated prophage identification; Methods/Tools used in simulation study; Methods/Tool performance scoring; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/Strong-Lab/Viral_Classification_in_Metagenomics/blob/f583cbff6b022ce3a7e3870003e22e14769566fa/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-viral-contig-simulation-2021-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"MIT (study repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-viral-contig-simulation-2021-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-viral-contig-simulation-2021-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Uses several simulation conditions and separately examines integrated viral sequences.","source_ids":["viral-contig-simulation-2021"],"source_locator":"Discussion (paragraph 4); Results/Overall tool performance (paragraph 1)"}],"limitations":[{"text":"Sensitivity and false-positive behaviour depend on contig length, community composition and reference coverage.","source_ids":["viral-contig-simulation-2021"],"source_locator":"Methods/Building simulated Illumina metagenomes (paragraph 4); Results/Tool performance by contig length (paragraph 1)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Metagenomic contigs","VIBRANT","Viral-contig or prophage predictions"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["viral-contig-simulation-2021"],"source_locator":"Methods/Tools used in simulation study (paragraph 1); Methods/Tool performance scoring (paragraph 1)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","VIBRANT is the comparison-table label; that label does not specify an immutable weight revision.","A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-28413ae1766316","kind":"model","name":"DNABERT-2","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["dna-genomes"]},"source_ids":["dart-eval-regulatory-2024"],"links":[],"attributes":{"entity_level":"method","version":"not stated in table","reported_name":"DNABERT-2","historical_missing_metadata":{"checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"DART-Eval tests DNABERT-2 on regulatory DNA under separately defined zero-shot, probed and fine-tuned protocols.","summary_source_ids":["dart-eval-regulatory-2024"],"summary_source_locator":"Table 2 (models); Section 3.2 zero-shot analysis; Table 3 regulatory-element identification; Appendix B evaluation procedures","sections":[{"title":"How the evaluated method works","body":"The masked DNA transformer uses byte-pair tokenisation. The linked regulatory-element row uses the paper’s zero-shot procedure; trained probing or fine-tuning rows are distinct evaluations.","source_ids":["dart-eval-regulatory-2024"],"source_locator":"Table 2 (models); Section 3.2 zero-shot analysis; Table 3 regulatory-element identification; Appendix B evaluation procedures"},{"title":"Underlying method and version boundaries","body":"DNABERT-2 replaces overlapping k-mer tokens with byte-pair encoding and uses ALiBi positional biases. The official 117M model produces 768-dimensional token representations; downstream classifiers and pooling choices are separate configuration details.","source_ids":["evidence-reported-base-dnabert2-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies DNABERT-2: regulatory element identification. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["dart-eval-regulatory-2024"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-b2-dart-eval-regulatory-2024"}],"facts":[{"label":"Model type","value":"DNA sequence transformer; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-dnabert2-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"The masked DNA transformer uses byte-pair tokenisation. The linked regulatory-element row uses the paper’s zero-shot procedure; trained probing or fine-tuning rows are distinct evaluations.","status":"source_checked","source_ids":["dart-eval-regulatory-2024"],"source_locator":"Table 2 (models); Section 3.2 zero-shot analysis; Table 3 regulatory-element identification; Appendix B evaluation procedures"},{"label":"Biological inputs","value":"Regulatory DNA and matched controls defined by the DART-Eval task","status":"source_checked","source_ids":["dart-eval-regulatory-2024"],"source_locator":"Table 2 (models); Section 3.2 zero-shot analysis; Table 3 regulatory-element identification; Appendix B evaluation procedures"},{"label":"Outputs","value":"Regulatory-element scores or task-specific predictions, according to the evaluation protocol","status":"source_checked","source_ids":["dart-eval-regulatory-2024"],"source_locator":"Table 2 (models); Section 3.2 zero-shot analysis; Table 3 regulatory-element identification; Appendix B evaluation procedures"},{"label":"Parameters","value":"117 million parameters","status":"source_checked","source_ids":["dart-eval-regulatory-2024"],"source_locator":"Table 2 (models); Section 3.2 zero-shot analysis; Table 3 regulatory-element identification; Appendix B evaluation procedures"},{"label":"Known versions / configuration","value":"DNABERT-2 is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["dart-eval-regulatory-2024"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"The benchmark identifies multispecies genomic pretraining; task adaptation is reported separately.","status":"source_checked","source_ids":["dart-eval-regulatory-2024"],"source_locator":"Table 2 (models); Section 3.2 zero-shot analysis; Table 3 regulatory-element identification; Appendix B evaluation procedures"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["dart-eval-regulatory-2024","evidence-reported-base-dnabert2-readme-md"],"source_locator":"Complete primary text and named comparison table; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://github.com/MAGICS-LAB/DNABERT_2/blob/f25bed9ee20db966dff39e5c1571249d04e36404/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-dnabert2-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"Apache 2.0 (upstream repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-base-dnabert2-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-base-dnabert2-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Separates zero-shot representation evidence from the effect of training a downstream predictor.","source_ids":["dart-eval-regulatory-2024"],"source_locator":"Table 2 (models); Section 3.2 zero-shot analysis; Table 3 regulatory-element identification; Appendix B evaluation procedures"}],"limitations":[{"text":"Regulatory-element discrimination does not establish quantitative cell-specific accessibility or regulatory-variant accuracy.","source_ids":["dart-eval-regulatory-2024"],"source_locator":"Table 2 (models); Section 3.2 zero-shot analysis; Table 3 regulatory-element identification; Appendix B evaluation procedures"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Regulatory DNA and matched controls defined by the DART-Eval task","DNABERT-2","Regulatory-element scores or task-specific predictions, according to the evaluation protocol"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["dart-eval-regulatory-2024"],"source_locator":"Table 2 (models); Section 3.2 zero-shot analysis; Table 3 regulatory-element identification; Appendix B evaluation procedures"},"coverage":"limited","gaps":["DNABERT-2 is the comparison-table label; that label does not specify an immutable weight revision.","A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-2894d253c5e8a8","kind":"model","name":"CUPID Data-aug-Avg","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["rna-transcriptomes"]},"source_ids":["cupid-rna-interactions-2026"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"CUPID Data-aug-Avg","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"CUPID predicts pairwise interactions between non-coding RNAs from their sequences.","summary_source_ids":["cupid-rna-interactions-2026"],"summary_source_locator":"Abstract (paragraph 1); Methods/Negative examples generation (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"GenerRNA sequence embeddings feed a feed-forward interaction classifier. The Data-aug-Avg configuration uses the paper’s augmented training pairs and averaging variant.","source_ids":["cupid-rna-interactions-2026"],"source_locator":"Methods/Model architecture/The classification unit (paragraph 1); Methods/Model architecture/The overall CUPID architecture (paragraph 1)"},{"title":"What was evaluated","body":"The linked evaluation record identifies CUPID Data-aug-Avg: non-coding RNA pairwise interaction prediction. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["cupid-rna-interactions-2026"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b4-007"}],"facts":[{"label":"Model type","value":"Study-specific predictive method; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["cupid-rna-interactions-2026"],"source_locator":"Methods/Model architecture/The classification unit (paragraph 1); Methods/Model architecture/The overall CUPID architecture (paragraph 1)"},{"label":"Architecture / procedure","value":"GenerRNA sequence embeddings feed a feed-forward interaction classifier. The Data-aug-Avg configuration uses the paper’s augmented training pairs and averaging variant.","status":"source_checked","source_ids":["cupid-rna-interactions-2026"],"source_locator":"Methods/Model architecture/The classification unit (paragraph 1); Methods/Model architecture/The overall CUPID architecture (paragraph 1)"},{"label":"Biological inputs","value":"Pairs of non-coding RNA sequences","status":"source_checked","source_ids":["cupid-rna-interactions-2026"],"source_locator":"Methods/Negative examples generation (paragraph 4); Methods/Dataset (paragraph 1)"},{"label":"Outputs","value":"Predicted ncRNA–ncRNA interaction labels or scores","status":"source_checked","source_ids":["cupid-rna-interactions-2026"],"source_locator":"Methods/Dataset (paragraph 8); Methods/Dataset (paragraph 1)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["cupid-rna-interactions-2026","evidence-reported-cupid-rna-interactions-2026-readme-md"],"source_locator":"Methods/Dataset; Methods/Data augmentation; Methods/Negative examples generation; Methods/Model architecture/The overall CUPID architecture; Methods/Model architecture/Pooling techniques; Methods/Model architecture/The classification unit; Methods/Model architecture/Mini-batch balancing; Methods/Experimental evaluation/Data preparation and splitting; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"CUPID Data-aug-Avg is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["cupid-rna-interactions-2026"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"The study’s filtered ncRNA-interaction dataset; augmentation produces four training instances per original pair.","status":"source_checked","source_ids":["cupid-rna-interactions-2026"],"source_locator":"Methods/Dataset (paragraph 1); Methods/Dataset (paragraph 8)"},{"label":"Context limits","value":"Input filtering is tied to GenerRNA token capacity; the paper describes approximately 4,096 nucleotides, with BPE compression affecting the correspondence.","status":"source_checked","source_ids":["cupid-rna-interactions-2026"],"source_locator":"Methods/Dataset (paragraph 4); Methods/Model architecture/The overall CUPID architecture (paragraph 2)"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/AnacletoLAB/ncRNA-CUPID/blob/f663c10d2f6c33f8513614badbbc673e654a64d8/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-cupid-rna-interactions-2026-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"Apache 2.0 (study repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-cupid-rna-interactions-2026-license-txt"],"source_locator":"LICENSE.txt; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-cupid-rna-interactions-2026-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Uses sequence representations across several ncRNA interaction types without requiring explicit thermodynamic folding features.","source_ids":["cupid-rna-interactions-2026"],"source_locator":"Methods/Dataset (paragraph 8); Discussion (paragraph 1)"}],"limitations":[{"text":"RNA pairs exceeding the study’s length filter are excluded; this restricts the distribution on which the reported result is established.","source_ids":["cupid-rna-interactions-2026"],"source_locator":"Methods/Dataset (paragraph 5); Methods/Dataset (paragraph 4)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Pairs of non-coding RNA sequences","CUPID Data-aug-Avg","Predicted ncRNA–ncRNA interaction labels or scores"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["cupid-rna-interactions-2026"],"source_locator":"Methods/Model architecture/The classification unit (paragraph 1); Methods/Model architecture/The overall CUPID architecture (paragraph 1)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","CUPID Data-aug-Avg is the comparison-table label; that label does not specify an immutable weight revision.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-2ae5fb0c147618","kind":"model","name":"Ibex","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["molecular-interactions"]},"source_ids":["ibex-2025"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"Ibex","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Ibex predicts antibody, nanobody and T-cell-receptor structures with an explicit bound/unbound conformation input.","summary_source_ids":["ibex-2025"],"summary_source_locator":"Introduction (paragraph 5); Methods/Data (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"An AlphaFold2-derived network has 16 structure-module blocks. An apo/holo conformation token is supplied with the variable-region sequence and propagated through residual connections.","source_ids":["ibex-2025"],"source_locator":"The Ibex model (paragraph 1); Methods/Model (paragraph 2)"},{"title":"What was evaluated","body":"The linked evaluation record identifies Ibex: Antibody loop structure prediction. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["ibex-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b3-040"}],"facts":[{"label":"Model type","value":"Study-specific predictive method; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["ibex-2025"],"source_locator":"The Ibex model (paragraph 1); Methods/Model (paragraph 2)"},{"label":"Architecture / procedure","value":"An AlphaFold2-derived network has 16 structure-module blocks. An apo/holo conformation token is supplied with the variable-region sequence and propagated through residual connections.","status":"source_checked","source_ids":["ibex-2025"],"source_locator":"The Ibex model (paragraph 1); Methods/Model (paragraph 2)"},{"label":"Biological inputs","value":"Immunoglobulin-family variable-region sequences and an apo/holo conformation token","status":"source_checked","source_ids":["ibex-2025"],"source_locator":"Methods/Data (paragraph 1); The Ibex model (paragraph 1)"},{"label":"Outputs","value":"Predicted three-dimensional structures in the requested conformational state","status":"source_checked","source_ids":["ibex-2025"],"source_locator":"Methods/Training (paragraph 1); Results (paragraph 1)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["ibex-2025","evidence-reported-ibex-2025-readme-md"],"source_locator":"Methods/Model; Methods/Data; Methods/Training; Methods/Private dataset generation and characterization; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"Ibex is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["ibex-2025"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"Labelled apo and holo structural pairs described in the paper.","status":"source_checked","source_ids":["ibex-2025"],"source_locator":"Methods/Training (paragraph 3); Methods/Training (paragraph 5)"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["ibex-2025","evidence-reported-ibex-2025-readme-md"],"source_locator":"Methods/Model; Methods/Data; Methods/Training; Methods/Private dataset generation and characterization; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/prescient-design/ibex/blob/2e785563806a3b63600eeaa2107d1254ddf5d196/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-ibex-2025-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"Apache 2.0 (study repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-ibex-2025-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-ibex-2025-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Treats bound and unbound conformations as distinct training conditions rather than conflating them.","source_ids":["ibex-2025"],"source_locator":"Results/Analysis of paired apo/holo structures (paragraph 1); Methods/Data (paragraph 2)"}],"limitations":[{"text":"Specifying a holo token does not supply the actual binding partner or prove the predicted conformation matches a particular complex.","source_ids":["ibex-2025"],"source_locator":"Methods/Data (paragraph 1); Discussion (paragraph 2)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Immunoglobulin-family variable-region sequences and an apo/holo conformation token","Ibex","Predicted three-dimensional structures in the requested conformational state"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["ibex-2025"],"source_locator":"The Ibex model (paragraph 1); Methods/Model (paragraph 2)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","Ibex is the comparison-table label; that label does not specify an immutable weight revision.","A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-2cb8118b4c77c0","kind":"model","name":"DNABERT-2","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["dna-genomes"]},"source_ids":["eden-genomic-classification-2026"],"links":[],"attributes":{"entity_level":"method","version":"not stated in table","reported_name":"DNABERT-2","historical_missing_metadata":{"checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"This DNABERT-2 row is a comparator in the EDEN genomic-classification paper.","summary_source_ids":["eden-genomic-classification-2026"],"summary_source_locator":"Experimental setup, results, and discussion/Comparison with state-of-the-art methods (paragraph 12); Experimental setup, results, and discussion/Comparison with state-of-the-art methods (paragraph 2)","sections":[{"title":"How the evaluated method works","body":"DNABERT-2 is a pretrained DNA transformer with a downstream classification procedure; the EDEN paper imports comparator values from their respective original studies.","source_ids":["eden-genomic-classification-2026"],"source_locator":"Experimental setup, results, and discussion/Comparison with state-of-the-art methods (paragraph 2); Experimental setup, results, and discussion/Comparison with state-of-the-art methods (paragraph 11)"},{"title":"Underlying method and version boundaries","body":"DNABERT-2 replaces overlapping k-mer tokens with byte-pair encoding and uses ALiBi positional biases. The official 117M model produces 768-dimensional token representations; downstream classifiers and pooling choices are separate configuration details.","source_ids":["evidence-reported-base-dnabert2-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies DNABERT-2: human core-promoter classification. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["eden-genomic-classification-2026"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-b2-eden-genomic-classification-2026"}],"facts":[{"label":"Model type","value":"DNA sequence transformer; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-dnabert2-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"DNABERT-2 is a pretrained DNA transformer with a downstream classification procedure; the EDEN paper imports comparator values from their respective original studies.","status":"source_checked","source_ids":["eden-genomic-classification-2026"],"source_locator":"Experimental setup, results, and discussion/Comparison with state-of-the-art methods (paragraph 2); Experimental setup, results, and discussion/Comparison with state-of-the-art methods (paragraph 11)"},{"label":"Biological inputs","value":"DNA sequences from genomic classification benchmarks","status":"source_checked","source_ids":["eden-genomic-classification-2026"],"source_locator":"Related work (paragraph 1); Proposed method: EDEN framework/Biological interpretation of EDN/Relationship to known biological patterns (paragraph 1)"},{"label":"Outputs","value":"Task-specific genomic class predictions","status":"source_checked","source_ids":["eden-genomic-classification-2026"],"source_locator":"Proposed method: EDEN framework/Multiscale EDN generation using K-mer concept/Biologically-informed multiscale representation (paragraph 1); Experimental setup, results, and discussion/Comparison with state-of-the-art methods (paragraph 12)"},{"label":"Parameters","value":"117 million parameters as specified in the comparison.","status":"source_checked","source_ids":["eden-genomic-classification-2026"],"source_locator":"Experimental setup, results, and discussion/Discussion (paragraph 3); Table Tab3 (paragraph 1)"},{"label":"Known versions / configuration","value":"DNABERT-2 is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["eden-genomic-classification-2026"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"This row is quoted from the DNABERT-2/GUE comparison rather than a new EDEN training run. The EDEN table does not independently establish the exact upstream fine-tuning configuration.","status":"source_checked","source_ids":["eden-genomic-classification-2026"],"source_locator":"Table 5 and corresponding comparison description"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["eden-genomic-classification-2026","evidence-reported-base-dnabert2-readme-md"],"source_locator":"Proposed method: EDEN framework; Proposed method: EDEN framework/EDN numerical representation; Proposed method: EDEN framework/Multiscale EDN generation using K-mer concept; Proposed method: EDEN framework/Multiscale EDN generation using K-mer concept/A unified framework for sequence encoding; Proposed method: EDEN framework/Multiscale EDN generation using K-mer concept/Biologically-informed multiscale representation; Proposed method: EDEN framework/Multiscale EDN generation using K-mer concept/KDE analysis and bandwidth selection; Proposed method: EDEN framework/Multiscale EDN generation using K-mer concept/Fundamental Distinction from Convolutional Operations; Proposed method: EDEN framework/Multiscale EDN generation using K-mer concept/Computational efficiency and representation properties; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://github.com/MAGICS-LAB/DNABERT_2/blob/f25bed9ee20db966dff39e5c1571249d04e36404/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-dnabert2-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"Apache 2.0 (upstream repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-base-dnabert2-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-base-dnabert2-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Provides a previously published transformer reference alongside smaller feature-based classifiers.","source_ids":["eden-genomic-classification-2026"],"source_locator":"Proposed method: EDEN framework/Multiscale EDN generation using K-mer concept/Biologically-informed multiscale representation (paragraph 3); Proposed method: EDEN framework/Multiscale EDN generation using K-mer concept/Fundamental Distinction from Convolutional Operations (paragraph 2)"}],"limitations":[{"text":"Table 5 states that comparator values come from their respective studies. They are quoted evidence, not fresh independent DNABERT-2 runs by the EDEN authors.","source_ids":["eden-genomic-classification-2026"],"source_locator":"Experimental setup, results, and discussion/Comparison with state-of-the-art methods (paragraph 2); Experimental setup, results, and discussion/Comparison with state-of-the-art methods (paragraph 10)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["DNA sequences from genomic classification benchmarks","DNABERT-2","Task-specific genomic class predictions"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["eden-genomic-classification-2026"],"source_locator":"Experimental setup, results, and discussion/Comparison with state-of-the-art methods (paragraph 2); Experimental setup, results, and discussion/Comparison with state-of-the-art methods (paragraph 11)"},"coverage":"limited","gaps":["DNABERT-2 is the comparison-table label; that label does not specify an immutable weight revision.","A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-2df975e60d16d1","kind":"model","name":"GenomeOcean","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["microbes-communities"]},"source_ids":["genomeocean-2025"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"GenomeOcean","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"GenomeOcean is a generative genome model trained on metagenomic contigs spanning diverse environments.","summary_source_ids":["genomeocean-2025"],"summary_source_locator":"Introduction (paragraph 5); Abstract (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"A generative language model uses byte-pair tokenisation and architectural optimisations for DNA sequence modelling. Co-assembled metagenomic contigs provide the pretraining corpus, rather than only reference genomes.","source_ids":["genomeocean-2025"],"source_locator":"Methods/Training/Evaluation Datasets/Evaluation of Dataset Complexity (paragraph 1); Abstract (paragraph 1)"},{"title":"What was evaluated","body":"The linked evaluation record identifies GenomeOcean: Natural vs artificial microbial genome sequence. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["genomeocean-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b3-027"}],"facts":[{"label":"Model type","value":"Study-specific predictive method; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["genomeocean-2025"],"source_locator":"Methods/Training/Evaluation Datasets/Evaluation of Dataset Complexity (paragraph 1); Abstract (paragraph 1)"},{"label":"Architecture / procedure","value":"A generative language model uses byte-pair tokenisation and architectural optimisations for DNA sequence modelling. Co-assembled metagenomic contigs provide the pretraining corpus, rather than only reference genomes.","status":"source_checked","source_ids":["genomeocean-2025"],"source_locator":"Methods/Training/Evaluation Datasets/Evaluation of Dataset Complexity (paragraph 1); Abstract (paragraph 1)"},{"label":"Biological inputs","value":"Genomic DNA sequence tokens","status":"source_checked","source_ids":["genomeocean-2025"],"source_locator":"Results/Pre-training (paragraph 2); Results/GenomeOcean Learns Protein-coding Principles from DNA Alone (paragraph 5)"},{"label":"Outputs","value":"Generated DNA sequences and representations usable for task-specific adaptation","status":"source_checked","source_ids":["genomeocean-2025"],"source_locator":"Results/Model Safety (paragraph 1); Methods/Model Evaluation/Generating and Evaluating Synthetic Metagenomes (paragraph 1)"},{"label":"Parameters","value":"4 billion parameters","status":"source_checked","source_ids":["genomeocean-2025"],"source_locator":"Discussion (paragraph 5); Methods/Model Evaluation/Generating and Evaluating Synthetic Metagenomes (paragraph 2)"},{"label":"Known versions / configuration","value":"GenomeOcean is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["genomeocean-2025"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"More than 600 Gbp of high-quality contigs assembled from 220 TB of metagenomic data.","status":"source_checked","source_ids":["genomeocean-2025"],"source_locator":"Methods/Training/Evaluation Datasets/Assembled Metagenome Datasets (paragraph 1); Results/Pre-training (paragraph 1)"},{"label":"Context limits","value":"Initial pretraining uses 1,024 BPE tokens. The 4B model is then continued at 10,240 tokens, approximately 51 kb; this larger limit is not assigned to the 100M or 500M variants.","status":"source_checked","source_ids":["genomeocean-2025"],"source_locator":"Results/Pre-training (paragraph 1); Table T3 (paragraph 1)"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/jgi-genomeocean/genomeocean/blob/06fa433169539a3c84d7366a663933b888a5386b/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-genomeocean-2025-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"Custom Lawrence Berkeley/Northwestern permissive licence; see the pinned licence text for its conditions. (study repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-genomeocean-2025-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-genomeocean-2025-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"The metagenomic corpus broadens representation beyond well-catalogued reference organisms.","source_ids":["genomeocean-2025"],"source_locator":"Abstract (paragraph 1); Introduction (paragraph 4)"}],"limitations":[{"text":"The paper’s sequence-generation and classification evaluations do not show that every generated sequence has an experimentally verified function.","source_ids":["genomeocean-2025"],"source_locator":"Results/GenomeOcean Learns Protein-coding Principles from DNA Alone (paragraph 6); Results/Inference Efficiency Benchmarking (paragraph 2)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Genomic DNA sequence tokens","GenomeOcean","Generated DNA sequences and representations usable for task-specific adaptation"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["genomeocean-2025"],"source_locator":"Methods/Training/Evaluation Datasets/Evaluation of Dataset Complexity (paragraph 1); Abstract (paragraph 1)"},"coverage":"limited","gaps":["GenomeOcean is the comparison-table label; that label does not specify an immutable weight revision.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-32a19f43a4c254","kind":"model","name":"CAMMiQ","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["microbes-communities"]},"source_ids":["cammiq-2022"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"CAMMiQ","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"CAMMiQ identifies and quantifies microbial genomes from metagenomic reads using combinatorial optimisation.","summary_source_ids":["cammiq-2022"],"summary_source_locator":"Results/Overview of CAMMiQ indexing and querying procedure (paragraph 3); Methods (paragraph 2)","sections":[{"title":"How the evaluated method works","body":"An index contains variable-length substrings unique to one reference genome or shared by exactly two. Optimisation resolves the mixture of reference genomes and their abundance from matching reads.","source_ids":["cammiq-2022"],"source_locator":"Methods (paragraph 3); Methods (paragraph 2)"},{"title":"What was evaluated","body":"The linked evaluation record identifies CAMMiQ: Strain-level abundance quantification. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["cammiq-2022"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b3-019"}],"facts":[{"label":"Model type","value":"Study-specific predictive method; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["cammiq-2022"],"source_locator":"Methods (paragraph 3); Methods (paragraph 2)"},{"label":"Architecture / procedure","value":"An index contains variable-length substrings unique to one reference genome or shared by exactly two. Optimisation resolves the mixture of reference genomes and their abundance from matching reads.","status":"source_checked","source_ids":["cammiq-2022"],"source_locator":"Methods (paragraph 3); Methods (paragraph 2)"},{"label":"Biological inputs","value":"Metagenomic reads and a reference-genome index","status":"source_checked","source_ids":["cammiq-2022"],"source_locator":"Results/Precision and recall in read classification across all species level queries (paragraph 2); Methods (paragraph 2)"},{"label":"Outputs","value":"Detected reference genomes and relative abundance estimates","status":"source_checked","source_ids":["cammiq-2022"],"source_locator":"Methods (paragraph 4); Results/Overview of CAMMiQ indexing and querying procedure (paragraph 2)"},{"label":"Parameters","value":"Not applicable: this is a reference-database algorithm rather than a neural model.","status":"inapplicable","source_ids":["cammiq-2022"],"source_locator":"Results/Precision and recall in read classification across all species level queries (paragraph 3); Introduction (paragraph 3)"},{"label":"Known versions / configuration","value":"CAMMiQ is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["cammiq-2022"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"No neural pretraining: the paper builds reference indexes from specified RefSeq releases, including releases 93 and 205.","status":"source_checked","source_ids":["cammiq-2022"],"source_locator":"Results/Datasets (paragraph 1); Results/Datasets (paragraph 2)"},{"label":"Context limits","value":"Not applicable to a learned context window; the analysed reads/genomes and versioned reference database define the workload.","status":"inapplicable","source_ids":["cammiq-2022"],"source_locator":"Methods/When to use unique substrings—the error free case/Theorem 1 (paragraph 2); Introduction (paragraph 4)"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/algo-cancer/CAMMiQ/blob/6142150d427a74cc21a5ee4d8b37a3b78884f163/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-cammiq-2022-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"MIT (study repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-cammiq-2022-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"Not applicable: this is an indexed combinatorial method, not a pretrained weight release.","status":"inapplicable","source_ids":["cammiq-2022"],"source_locator":"Methods (paragraph 7); Discussion (paragraph 3)"}],"strengths":[{"text":"Uses doubly unique substrings to distinguish closely related genomes that share many sequences.","source_ids":["cammiq-2022"],"source_locator":"Results/Performance of CAMMiQ on real single-cell metatranscriptomic queries (paragraph 1); Abstract (paragraph 1)"}],"limitations":[{"text":"Identification is relative to the supplied genome database; novel organisms and incomplete references require separate interpretation.","source_ids":["cammiq-2022"],"source_locator":"Introduction (paragraph 7); Abstract (paragraph 1)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Metagenomic reads and a reference-genome index","CAMMiQ","Detected reference genomes and relative abundance estimates"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["cammiq-2022"],"source_locator":"Methods (paragraph 3); Methods (paragraph 2)"},"coverage":"limited","gaps":["CAMMiQ is the comparison-table label; that label does not specify an immutable weight revision.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-399b1ce87a3f6d","kind":"model","name":"scGPT","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["cells-tissues"]},"source_ids":["genept-2024"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"scGPT","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"scGPT is a learned-expression comparator in the GenePT study.","summary_source_ids":["genept-2024"],"summary_source_locator":"Results/GenePT embedding removes batch effect while preserving underlying biology (paragraph 1); Abstract (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"The study uses pretrained scGPT gene/cell embeddings for downstream applications, including previously documented gene-level results and extracted cell-level features.","source_ids":["genept-2024"],"source_locator":"Methods/Downstream gene-level and cell-level applications: (paragraph 1); Methods/Data Collection and Transformation: (paragraph 1)"},{"title":"Underlying method and version boundaries","body":"The official scGPT implementation supplies pretrained checkpoints and separate workflows for embedding extraction, cell annotation, integration and perturbation modelling. The checkpoint and adaptation procedure must be identified separately for each result.","source_ids":["evidence-reported-base-scgpt-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies scGPT: Cell-type structure in frozen embeddings. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["genept-2024"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b3-012"}],"facts":[{"label":"Model type","value":"Single-cell foundation model; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-scgpt-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"The study uses pretrained scGPT gene/cell embeddings for downstream applications, including previously documented gene-level results and extracted cell-level features.","status":"source_checked","source_ids":["genept-2024"],"source_locator":"Methods/Downstream gene-level and cell-level applications: (paragraph 1); Methods/Data Collection and Transformation: (paragraph 1)"},{"label":"Biological inputs","value":"Single-cell gene-expression profiles","status":"source_checked","source_ids":["genept-2024"],"source_locator":"Related Work/Deciphering natural language embeddings (paragraph 5); Abstract (paragraph 1)"},{"label":"Outputs","value":"Pretrained gene/cell representations and task-specific predictions","status":"source_checked","source_ids":["genept-2024"],"source_locator":"Results/GenePT embeddings capture underlying gene functionality (paragraph 2); Methods/Downstream gene-level and cell-level applications: (paragraph 1)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["genept-2024","evidence-reported-base-scgpt-readme-md"],"source_locator":"Methods/Data Collection and Transformation:; Methods/Downstream gene-level and cell-level applications:; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"scGPT is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["genept-2024"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"The source describes scGPT pretraining on 33 million CELLxGENE cells; the reported comparator uses frozen embeddings and cosine 10-nearest-neighbour classification.","status":"source_checked","source_ids":["genept-2024"],"source_locator":"Related Work / Foundation models for single-cell transcriptomics; Table4"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["genept-2024","evidence-reported-base-scgpt-readme-md"],"source_locator":"Methods/Data Collection and Transformation:; Methods/Downstream gene-level and cell-level applications:; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://github.com/bowang-lab/scGPT/blob/cebd6fae655b9c585a4807daa3ac31bb764f06b4/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-scgpt-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"MIT (upstream repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-base-scgpt-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-base-scgpt-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Provides an expression-pretraining reference against text-derived GenePT embeddings.","source_ids":["genept-2024"],"source_locator":"Methods/Data Collection and Transformation: (paragraph 2); Introduction (paragraph 4)"}],"limitations":[{"text":"Some gene-level comparator scores are drawn from previous documentation; they are not all freshly trained or independently replicated by this study.","source_ids":["genept-2024"],"source_locator":"Appendix/Appendix A Assessing GenePT sensitivity to variations in gene summary inputs (paragraph 2); Related Work/Deciphering natural language embeddings (paragraph 3)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Single-cell gene-expression profiles","scGPT","Pretrained gene/cell representations and task-specific predictions"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["genept-2024"],"source_locator":"Methods/Downstream gene-level and cell-level applications: (paragraph 1); Methods/Data Collection and Transformation: (paragraph 1)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","scGPT is the comparison-table label; that label does not specify an immutable weight revision.","A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-3af86cb274f658","kind":"model","name":"NT-v2","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["dna-genomes"]},"source_ids":["dna-foundation-models-2025"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"NT-v2","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"This genomic-embedding configuration is compared across classification, gene-expression, variant-effect and domain-recognition tasks.","summary_source_ids":["dna-foundation-models-2025"],"summary_source_locator":"Abstract (paragraph 1); Introduction (paragraph 4)","sections":[{"title":"How the evaluated method works","body":"Frozen sequence representations are pooled and supplied to downstream analyses. The study explicitly compares summary-token, mean and maximum pooling.","source_ids":["dna-foundation-models-2025"],"source_locator":"Results/Sequence classification: pooling methods benchmark (paragraph 2); Methods/Benchmarking methods/Sequence classification benchmark: pooling methods (paragraph 4)"},{"title":"What was evaluated","body":"The linked evaluation record identifies NT-v2: Human 5mC detection. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["dna-foundation-models-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-002"}],"facts":[{"label":"Model type","value":"Study-specific predictive method; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["dna-foundation-models-2025"],"source_locator":"Results/Sequence classification: pooling methods benchmark (paragraph 2); Methods/Benchmarking methods/Sequence classification benchmark: pooling methods (paragraph 4)"},{"label":"Architecture / procedure","value":"Frozen sequence representations are pooled and supplied to downstream analyses. The study explicitly compares summary-token, mean and maximum pooling.","status":"source_checked","source_ids":["dna-foundation-models-2025"],"source_locator":"Results/Sequence classification: pooling methods benchmark (paragraph 2); Methods/Benchmarking methods/Sequence classification benchmark: pooling methods (paragraph 4)"},{"label":"Biological inputs","value":"DNA sequence windows, with model-specific context handling","status":"source_checked","source_ids":["dna-foundation-models-2025"],"source_locator":"Discussion/Limitations (paragraph 1); Methods/Benchmarking datasets/Sequence classification datasets (paragraph 1)"},{"label":"Outputs","value":"Sequence embeddings and downstream genomic predictions","status":"source_checked","source_ids":["dna-foundation-models-2025"],"source_locator":"Introduction (paragraph 1); Discussion (paragraph 1)"},{"label":"Parameters","value":"500 million parameters","status":"source_checked","source_ids":["dna-foundation-models-2025"],"source_locator":"Methods/DNA foundation language models (paragraph 4); Methods/DNA foundation language models (paragraph 3)"},{"label":"Known versions / configuration","value":"NT-v2-500M","status":"source_checked","source_ids":["dna-foundation-models-2025"],"source_locator":"Methods/Model configuration selection (paragraph 1); Table Tab6 (paragraph 1)"},{"label":"Training data / fitting","value":"Masked-language-model pretraining on genomes from 850 species, including the human reference genome; this study evaluates frozen representations.","status":"source_checked","source_ids":["dna-foundation-models-2025"],"source_locator":"Methods / DNA foundation language models; Nucleotide Transformer Version2 paragraph"},{"label":"Context limits","value":"12,000-nucleotide model input limit; the benchmark also specifies task-dependent shorter windows.","status":"source_checked","source_ids":["dna-foundation-models-2025"],"source_locator":"Methods / DNA foundation language models; Nucleotide Transformer Version2 paragraph"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/ChongWuLab/dna_foundation_benchmark/blob/3f4c81ce066f3c47422a83466b085aac1a6be902/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-dna-foundation-models-2025-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","status":"unreported","source_ids":["evidence-reported-dna-foundation-models-2025-readme-md"],"source_locator":"README.md and repository-root licence-file search"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-dna-foundation-models-2025-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Tests pooling choices explicitly and compares task-specific specialists.","source_ids":["dna-foundation-models-2025"],"source_locator":"Methods/Benchmarking methods/Sequence classification benchmark (paragraph 3); Methods/Benchmarking methods/Sequence classification benchmark (paragraph 4)"}],"limitations":[{"text":"The paper finds task-dependent performance; a classifier on frozen features is not a label-free zero-shot prediction.","source_ids":["dna-foundation-models-2025"],"source_locator":"Results/Gene expression prediction (paragraph 5); Results/Sequence classification: pooling methods benchmark (paragraph 1)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["DNA sequence windows, with model-specific context handling","NT-v2","Sequence embeddings and downstream genomic predictions"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["dna-foundation-models-2025"],"source_locator":"Results/Sequence classification: pooling methods benchmark (paragraph 2); Methods/Benchmarking methods/Sequence classification benchmark: pooling methods (paragraph 4)"},"coverage":"limited","gaps":["No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-3bdb3093e8d531","kind":"model","name":"scGPT","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["cells-tissues"]},"source_ids":["single-cell-peft-2024"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"scGPT","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"This single-cell model is evaluated in a study of parameter-efficient adaptation for cell-type identification.","summary_source_ids":["single-cell-peft-2024"],"summary_source_locator":"Introduction (paragraph 5); Results/Comparison of proposed PEFT strategies and other finetuning approaches (paragraph 2)","sections":[{"title":"How the evaluated method works","body":"The study compares ordinary fine-tuning with methods that retain original model parameters while learning additional tensors.","source_ids":["single-cell-peft-2024"],"source_locator":"Introduction (paragraph 4); Proposed PEFT strategies for scLLMs/Finetuning and evaluation settings (paragraph 1)"},{"title":"Underlying method and version boundaries","body":"The official scGPT implementation supplies pretrained checkpoints and separate workflows for embedding extraction, cell annotation, integration and perturbation modelling. The checkpoint and adaptation procedure must be identified separately for each result.","source_ids":["evidence-reported-base-scgpt-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies scGPT: Cell-type identification. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["single-cell-peft-2024"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-025"}],"facts":[{"label":"Model type","value":"Single-cell foundation model; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-scgpt-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"The study compares ordinary fine-tuning with methods that retain original model parameters while learning additional tensors.","status":"source_checked","source_ids":["single-cell-peft-2024"],"source_locator":"Introduction (paragraph 4); Proposed PEFT strategies for scLLMs/Finetuning and evaluation settings (paragraph 1)"},{"label":"Biological inputs","value":"Single-cell gene-expression profiles","status":"source_checked","source_ids":["single-cell-peft-2024"],"source_locator":"An overview of current scLLMs (paragraph 1); Data and code availability (paragraph 1)"},{"label":"Outputs","value":"Cell-type predictions","status":"source_checked","source_ids":["single-cell-peft-2024"],"source_locator":"Proposed PEFT strategies for scLLMs/Finetuning and evaluation settings (paragraph 2); Results/Comparison of native scLLMs on cell type identification (paragraph 4)"},{"label":"Parameters","value":"Table 3 reports 51M trainable parameters for full fine-tuning. Prompt/classifier variants train different subsets; this count is not assigned to every adaptation.","status":"source_checked","source_ids":["single-cell-peft-2024"],"source_locator":"Table3; Trainable Parameters column"},{"label":"Known versions / configuration","value":"scGPT is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["single-cell-peft-2024"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"The evaluation separately compares native scGPT, full fine-tuning, classifier-only fitting and four prompting methods across the named cell-annotation datasets.","status":"source_checked","source_ids":["single-cell-peft-2024"],"source_locator":"Proposed PEFT strategies for scLLMs / Data preparation; Table3"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["single-cell-peft-2024","evidence-reported-base-scgpt-readme-md"],"source_locator":"An overview of current scLLMs/Pretrainer:; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://github.com/bowang-lab/scGPT/blob/cebd6fae655b9c585a4807daa3ac31bb764f06b4/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-scgpt-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"MIT (upstream repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-base-scgpt-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-base-scgpt-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Examines the trade-off between adaptation performance and the number of trainable parameters.","source_ids":["single-cell-peft-2024"],"source_locator":"Results/Comparison of proposed PEFT strategies and other finetuning approaches (paragraph 3); Results/Comparison of proposed PEFT strategies and other finetuning approaches (paragraph 2)"}],"limitations":[{"text":"The base-model label alone does not identify which adaptation tensors were trained; exact PEFT and full-fine-tuning rows must remain separate.","source_ids":["single-cell-peft-2024"],"source_locator":"Results/Comparison of proposed PEFT strategies and other finetuning approaches (paragraph 2); Introduction (paragraph 5)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Single-cell gene-expression profiles","scGPT","Cell-type predictions"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["single-cell-peft-2024"],"source_locator":"Introduction (paragraph 4); Proposed PEFT strategies for scLLMs/Finetuning and evaluation settings (paragraph 1)"},"coverage":"limited","gaps":["scGPT is the comparison-table label; that label does not specify an immutable weight revision.","A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-3c196326586fa7","kind":"model","name":"PMF + ECFP + PF (LightGBM)","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["molecular-interactions"]},"source_ids":["fingerprint-scoring-2022"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"PMF + ECFP + PF (LightGBM)","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"This scoring study fits protein–ligand affinity regressors to potential-of-mean-force descriptors and molecular fingerprints.","summary_source_ids":["fingerprint-scoring-2022"],"summary_source_locator":"Methods (paragraph 1); Results and Discussion/Comparison of Binding Affinities Calculated\nby New Scoring Functions with Experimental Values (paragraph 2)","sections":[{"title":"How the evaluated method works","body":"The LightGBM configuration combines PMF interaction descriptors with extended-connectivity ligand fingerprints and protein fingerprints. The LASSO configuration is a distinct sparse linear comparator using PMF descriptors.","source_ids":["fingerprint-scoring-2022"],"source_locator":"Methods/Computational Details (paragraph 4); Methods/Machine Learning with LASSO and LightGBM (paragraph 1)"},{"title":"What was evaluated","body":"The linked evaluation record identifies PMF + ECFP + PF (LightGBM): Protein–ligand binding energy prediction. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["fingerprint-scoring-2022"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b3-050"}],"facts":[{"label":"Model type","value":"Gradient-boosted-tree pipeline; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["fingerprint-scoring-2022"],"source_locator":"Methods/Computational Details (paragraph 4); Methods/Machine Learning with LASSO and LightGBM (paragraph 1)"},{"label":"Architecture / procedure","value":"The LightGBM configuration combines PMF interaction descriptors with extended-connectivity ligand fingerprints and protein fingerprints. The LASSO configuration is a distinct sparse linear comparator using PMF descriptors.","status":"source_checked","source_ids":["fingerprint-scoring-2022"],"source_locator":"Methods/Computational Details (paragraph 4); Methods/Machine Learning with LASSO and LightGBM (paragraph 1)"},{"label":"Biological inputs","value":"Protein–ligand structural descriptors, with ligand and protein fingerprints in the combined configuration","status":"source_checked","source_ids":["fingerprint-scoring-2022"],"source_locator":"Methods (paragraph 1); Results and Discussion/Comparison of Binding Affinities Calculated\nby New Scoring Functions with Experimental Values (paragraph 2)"},{"label":"Outputs","value":"Predicted protein–ligand binding affinity","status":"source_checked","source_ids":["fingerprint-scoring-2022"],"source_locator":"Introduction (paragraph 4); Results and Discussion/Comparison of Binding Affinities Calculated\nby New Scoring Functions with Experimental Values (paragraph 1)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["fingerprint-scoring-2022"],"source_locator":"Methods; Methods/PMF Score; Methods/Ligand\nFingerprints; Methods/Protein\nFingerprints; Methods/Machine Learning with LASSO and LightGBM; Methods/Computational Details; inspected for aggregate parameter count (component sizes are not added without an exact configuration)"},{"label":"Known versions / configuration","value":"PMF + ECFP + PF (LightGBM) is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["fingerprint-scoring-2022"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"Of 6,271 PDBbind complexes, 4,933 are training, 1,234 are pretest for hyperparameter selection, and 104 are test. LASSO and LightGBM are fitted on interaction fingerprints.","status":"source_checked","source_ids":["fingerprint-scoring-2022"],"source_locator":"Back/Availability notes (paragraph 2); Back/Availability notes (paragraph 1)"},{"label":"Context limits","value":"Protein–ligand interaction fingerprints from supplied complex structures; sequence-token context length is inapplicable.","status":"source_checked","source_ids":["fingerprint-scoring-2022"],"source_locator":"Methods/PMF Score (paragraph 1); Methods/Computational Details (paragraph 2)"},{"label":"Access","value":"The paper links the public RDKit, scikit-learn and LightGBM libraries and provides hyperparameters in its Supporting Information. Additional study data are available from the authors; a complete study-specific checkpoint release is not established.","status":"source_checked","source_ids":["fingerprint-scoring-2022"],"source_locator":"Back / Notes, data and software availability"},{"label":"Code licence","value":"No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","status":"unreported","source_ids":["fingerprint-scoring-2022"],"source_locator":"Back/Availability notes (paragraph 2); Back/Availability notes (paragraph 1)"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["fingerprint-scoring-2022"],"source_locator":"Back/Availability notes (paragraph 2); Back/Availability notes (paragraph 1)"}],"strengths":[{"text":"Offers both a sparse linear control and a nonlinear tree-based model over explicitly defined chemical descriptors.","source_ids":["fingerprint-scoring-2022"],"source_locator":"Methods/Machine Learning with LASSO and LightGBM (paragraph 2); Results and Discussion/Analysis of Descriptors\nAffecting the Scoring\nFunction (paragraph 1)"}],"limitations":[{"text":"The descriptor set and regression algorithm differ between rows; performance cannot be assigned to PMF or fingerprinting in isolation.","source_ids":["fingerprint-scoring-2022"],"source_locator":"Results and Discussion/Analysis of Descriptors\nAffecting the Scoring\nFunction (paragraph 1); Results and Discussion/Systems with Improved and\nUnimproved Scores (paragraph 1)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Protein–ligand structural descriptors, with ligand and protein fingerprints in the combined configuration","PMF + ECFP + PF (LightGBM)","Predicted protein–ligand binding affinity"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["fingerprint-scoring-2022"],"source_locator":"Methods/Computational Details (paragraph 4); Methods/Machine Learning with LASSO and LightGBM (paragraph 1)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","PMF + ECFP + PF (LightGBM) is the comparison-table label; that label does not specify an immutable weight revision.","No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-3e58d0faf88d2e","kind":"model","name":"RiNALMo","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["rna-transcriptomes"]},"source_ids":["mrnabench-2025"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"RiNALMo","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"This RNA model is evaluated as a representation source for mature-mRNA prediction in mRNABench.","summary_source_ids":["mrnabench-2025"],"summary_source_locator":"Introduction/Sequence Compressibility: (paragraph 1); Conclusion (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"The benchmark extracts nucleotide-model embeddings and evaluates task-specific predictors under standardised protocols.","source_ids":["mrnabench-2025"],"source_locator":"Results/Nucleotide foundation models perform poorly at compositional generalization (paragraph 3); Methods/Linear Probing (paragraph 1)"},{"title":"Underlying method and version boundaries","body":"RiNALMo produces nucleotide representations from RNA sequences. The documented large model has 650 million parameters and pretraining on 36 million non-coding RNA sequences; the benchmark row still requires its exact checkpoint identity.","source_ids":["evidence-reported-base-rinalmo-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies RiNALMo: Mean ribosome load from MPRA. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["mrnabench-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-009"}],"facts":[{"label":"Model type","value":"RNA sequence language model; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-rinalmo-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"The benchmark extracts nucleotide-model embeddings and evaluates task-specific predictors under standardised protocols.","status":"source_checked","source_ids":["mrnabench-2025"],"source_locator":"Results/Nucleotide foundation models perform poorly at compositional generalization (paragraph 3); Methods/Linear Probing (paragraph 1)"},{"label":"Biological inputs","value":"Mature mRNA sequences or benchmark-defined segments","status":"source_checked","source_ids":["mrnabench-2025"],"source_locator":"Conclusion (paragraph 1); Appendix Contents/Dataset Processing Protocols/Variant Effect Prediction (paragraph 1)"},{"label":"Outputs","value":"Embeddings and task-specific mRNA-property predictions","status":"source_checked","source_ids":["mrnabench-2025"],"source_locator":"Related Works/Deep learning for mRNA property prediction: (paragraph 1); Appendix Contents/Dataset Processing Protocols/Variant Effect Prediction (paragraph 1)"},{"label":"Parameters","value":"650M parameters (Table 1).","status":"source_checked","source_ids":["mrnabench-2025"],"source_locator":"Table1; RiNALMo row"},{"label":"Known versions / configuration","value":"RiNALMo ncRNA encoder listed as rinalmo in the comparison tables.","status":"source_checked","source_ids":["mrnabench-2025"],"source_locator":"Introduction (paragraph 1); Introduction (paragraph 3)"},{"label":"Training data / fitting","value":"Task-specific adaptation uses the mRNABench partitions; the encoder’s original RNA pretraining is distinct.","status":"source_checked","source_ids":["mrnabench-2025"],"source_locator":"Appendix Contents/Dataset Processing Protocols/Variant Effect Prediction (paragraph 1); Appendix Contents/Dataset Processing Protocols/Mean Ribosome Load (paragraph 1)"},{"label":"Context limits","value":"1,024-token context listed in Table 1; the benchmark chunks longer sequences before probing.","status":"source_checked","source_ids":["mrnabench-2025"],"source_locator":"Table1 RiNALMo row; Methods / Linear Probing"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://github.com/lbcb-sci/RiNALMo/blob/2c2c5c14a5ae609d8c560a5d9ca32e51e0288955/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-rinalmo-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"Apache 2.0 (upstream repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-base-rinalmo-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"CC BY 4.0 for RiNALMo model parameters; code is separately Apache 2.0.","status":"source_checked","source_ids":["evidence-reported-base-rinalmo-readme-md"],"source_locator":"README.md; Model Parameters License"}],"strengths":[{"text":"Tests a broad set of mature-mRNA properties across ten datasets and 59 tasks.","source_ids":["mrnabench-2025"],"source_locator":"Conclusion (paragraph 1); Benchmarking Tasks (paragraph 1)"}],"limitations":[{"text":"A model pretrained mainly on non-coding RNA may have different strengths on mature-mRNA tasks; aggregate results do not identify a universal RNA-model winner.","source_ids":["mrnabench-2025"],"source_locator":"Results/Sequence compressibility predicts model performance across genomic regions (paragraph 4); Conclusion (paragraph 1)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Mature mRNA sequences or benchmark-defined segments","RiNALMo","Embeddings and task-specific mRNA-property predictions"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["mrnabench-2025"],"source_locator":"Results/Nucleotide foundation models perform poorly at compositional generalization (paragraph 3); Methods/Linear Probing (paragraph 1)"},"coverage":"limited","gaps":["An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-3f850c08d76410","kind":"model","name":"DEBFold","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["rna-transcriptomes"]},"source_ids":["debfold-2024"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"DEBFold","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"DEBFold combines several RNA folding tools with a neural network and constrained thermodynamic optimisation.","summary_source_ids":["debfold-2024"],"summary_source_locator":"Methods and Data Sets/DEBFold Workflow (paragraph 1); Results and Discussion/DEBFold Is Robust against Different Thermodynamics-Constrained\nOptimization Algorithms (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"One-hot RNA and predictions from RNAfold, IPknot, MaxExpect, ProbKnot, RNAprob and Fold feed convolutional encoding/decoding and self-attention. The network estimates folding probabilities that become SHAPE-like constraints for free-energy minimisation.","source_ids":["debfold-2024"],"source_locator":"Methods and Data Sets/DEBFold Workflow/Stage 2: Score-Constrained Optimization Folding (paragraph 1); Methods and Data Sets/DEBFold Workflow (paragraph 1)"},{"title":"What was evaluated","body":"The linked evaluation record identifies DEBFold: RNA secondary structure. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["debfold-2024"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-015"}],"facts":[{"label":"Model type","value":"Convolutional neural network; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["debfold-2024"],"source_locator":"Methods and Data Sets/DEBFold Workflow/Stage 2: Score-Constrained Optimization Folding (paragraph 1); Methods and Data Sets/DEBFold Workflow (paragraph 1)"},{"label":"Architecture / procedure","value":"One-hot RNA and predictions from RNAfold, IPknot, MaxExpect, ProbKnot, RNAprob and Fold feed convolutional encoding/decoding and self-attention. The network estimates folding probabilities that become SHAPE-like constraints for free-energy minimisation.","status":"source_checked","source_ids":["debfold-2024"],"source_locator":"Methods and Data Sets/DEBFold Workflow/Stage 2: Score-Constrained Optimization Folding (paragraph 1); Methods and Data Sets/DEBFold Workflow (paragraph 1)"},{"label":"Biological inputs","value":"RNA sequences plus computed outputs of six folding procedures","status":"source_checked","source_ids":["debfold-2024"],"source_locator":"Methods and Data Sets/DEBFold Workflow/Stage 1: Structure Location Folding Probability Estimation (paragraph 5); Results and Discussion/DEBFold Is Robust against Different Thermodynamics-Constrained\nOptimization Algorithms (paragraph 1)"},{"label":"Outputs","value":"RNA secondary structures","status":"source_checked","source_ids":["debfold-2024"],"source_locator":"Introduction (paragraph 1); Methods and Data Sets/DEBFold Workflow (paragraph 1)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["debfold-2024"],"source_locator":"Methods and Data Sets/DEBFold Workflow; Methods and Data Sets/DEBFold Workflow/Stage 1: Structure Location Folding Probability Estimation; Methods and Data Sets/DEBFold Workflow/Stage 2: Score-Constrained Optimization Folding; Methods and Data Sets/DEBFold Workflow/Model Training Hyperparameters; Methods and Data Sets/Family-Wise Processed RNA Structure Ground-Truth Data Set; Methods and Data Sets/Family-Wise Processed RNA Structure Ground-Truth Data Set/Contamination-Free Family-Wise Independent Test Set for Evaluating\nExisting Deep-Learning-Based Structure Prediction Tools; Methods and Data Sets/Family-Wise Processed RNA Structure Ground-Truth Data Set/PDB-Derived Source-Independent Test Set; Methods and Data Sets/Structure Prediction Evaluation Metrics; inspected for aggregate parameter count (component sizes are not added without an exact configuration)"},{"label":"Known versions / configuration","value":"DEBFold is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["debfold-2024"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"RNA-structure datasets with family-wise reserved test sets and a separate PDB-derived test set.","status":"source_checked","source_ids":["debfold-2024"],"source_locator":"Introduction (paragraph 5); Results and Discussion/DEBFold Has Better Generalization Performance than Existing\nDeep-Learning-Based Attempts (paragraph 1)"},{"label":"Context limits","value":"Cross-validation inputs are padded to 512 nucleotides for batching; padding positions are excluded from loss and F1 calculations.","status":"source_checked","source_ids":["debfold-2024"],"source_locator":"Methods and Data Sets/DEBFold Workflow/Stage 1: Structure Location Folding Probability Estimation (paragraph 1); Methods and Data Sets/DEBFold Workflow/Model Training Hyperparameters (paragraph 1)"},{"label":"Access","value":"The authors publish a web interface at https://cobis.bme.ncku.edu.tw/DEBFold/. The source establishes this access route; code and model-weight licence terms remain separate.","status":"source_checked","source_ids":["debfold-2024"],"source_locator":"Abstract and Introduction, DEBFold web interface"},{"label":"Code licence","value":"No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","status":"unreported","source_ids":["debfold-2024"],"source_locator":"Methods and Data Sets/Family-Wise Processed RNA Structure Ground-Truth Data Set/Contamination-Free Family-Wise Independent Test Set for Evaluating\nExisting Deep-Learning-Based Structure Prediction Tools (paragraph 1); Conclusions (paragraph 1)"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["debfold-2024"],"source_locator":"Methods and Data Sets/DEBFold Workflow/Stage 1: Structure Location Folding Probability Estimation (paragraph 3); Methods and Data Sets/DEBFold Workflow/Model Training Hyperparameters (paragraph 1)"}],"strengths":[{"text":"Integrates thermodynamic predictions with a learned constraint model and explicitly checks family-wise generalisation.","source_ids":["debfold-2024"],"source_locator":"Results and Discussion/DEBFold Outperforms Previous Thermodynamics-Based RNA Structure\nPrediction Tools (paragraph 2); Abstract (paragraph 1)"}],"limitations":[{"text":"This is a multistage ensemble pipeline, with dependencies on six upstream tools; its result is not attributable to a single RNA language model.","source_ids":["debfold-2024"],"source_locator":"Introduction (paragraph 5); Results and Discussion/Limitations of DEBFold (paragraph 1)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["RNA sequences plus computed outputs of six folding procedures","DEBFold","RNA secondary structures"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["debfold-2024"],"source_locator":"Methods and Data Sets/DEBFold Workflow/Stage 2: Score-Constrained Optimization Folding (paragraph 1); Methods and Data Sets/DEBFold Workflow (paragraph 1)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","DEBFold is the comparison-table label; that label does not specify an immutable weight revision.","No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-3fd1e9f6c573b2","kind":"model","name":"GTDB-Tk","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["microbes-communities"]},"source_ids":["kmetashot-2025"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"GTDB-Tk","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"GTDB-Tk classifies metagenome-assembled genomes in the kMetaShot comparison.","summary_source_ids":["kmetashot-2025"],"summary_source_locator":"Materials and methods/GTDBtk taxonomy conversion (paragraph 1); Results (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"Marker identification and sequence alignment place genomes in the Genome Taxonomy Database framework.","source_ids":["kmetashot-2025"],"source_locator":"Materials and methods/GTDBtk taxonomy conversion (paragraph 1); Materials and methods (paragraph 1)"},{"title":"Underlying method and version boundaries","body":"GTDB-Tk assigns bacterial and archaeal genomes to the Genome Taxonomy Database and supports MAGs, isolate genomes and single-cell genomes. Historical GTDB-Tk versions and GTDB data releases must be retained together.","source_ids":["evidence-reported-base-gtdbtk-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies GTDB-Tk: Mock-community MAG taxonomy classification. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["kmetashot-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b3-030"}],"facts":[{"label":"Model type","value":"Genome-taxonomy toolkit; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-gtdbtk-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"Marker identification and sequence alignment place genomes in the Genome Taxonomy Database framework.","status":"source_checked","source_ids":["kmetashot-2025"],"source_locator":"Materials and methods/GTDBtk taxonomy conversion (paragraph 1); Materials and methods (paragraph 1)"},{"label":"Biological inputs","value":"Bacterial or archaeal genome assemblies/MAGs","status":"source_checked","source_ids":["kmetashot-2025"],"source_locator":"Results (paragraph 1); Materials and methods (paragraph 1)"},{"label":"Outputs","value":"GTDB taxonomic assignments","status":"source_checked","source_ids":["kmetashot-2025"],"source_locator":"Materials and methods/Taxonomic classification algorithm (paragraph 5); Materials and methods/GTDBtk taxonomy conversion (paragraph 1)"},{"label":"Parameters","value":"Not applicable: this is a reference-database algorithm rather than a neural model.","status":"inapplicable","source_ids":["kmetashot-2025"],"source_locator":"Materials and methods/kMetaShot reference module (paragraph 9); Materials and methods/kMetaShot reference module (paragraph 10)"},{"label":"Known versions / configuration","value":"GTDB-Tk v1.0.2 in the reported CAMI II comparison","status":"unreported","source_ids":["kmetashot-2025"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"Reference taxonomy and genome database rather than a pretrained neural model.","status":"source_checked","source_ids":["kmetashot-2025"],"source_locator":"Materials and methods/GTDBtk taxonomy conversion (paragraph 1); Results/Benchmarking on the in silico Critical Assessment of Metagenome Interpretation II datasets (paragraph 1)"},{"label":"Context limits","value":"Not applicable to a learned context window; the analysed reads/genomes and versioned reference database define the workload.","status":"inapplicable","source_ids":["kmetashot-2025"],"source_locator":"Results (paragraph 1); Materials and methods (paragraph 1)"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://github.com/Ecogenomics/GTDBTk/blob/f17decef1f9d9cf5b4d31fd21f5c9d32d813abdc/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-gtdbtk-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"GNU GPL version 3 (upstream repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-base-gtdbtk-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"Not applicable to neural weights; reference-database access and reuse terms are separate.","status":"inapplicable","source_ids":["kmetashot-2025"],"source_locator":"Materials and methods/kMetaShot reference module (paragraph 10); Materials and methods/Evaluation of computational requirements (paragraph 1)"}],"strengths":[{"text":"Uses a genome- and phylogeny-based reference procedure.","source_ids":["kmetashot-2025"],"source_locator":"Materials and methods/kMetaShot reference module (paragraph 10); Materials and methods/kMetaShot reference module (paragraph 6)"}],"limitations":[{"text":"Database release and genome completeness affect classification; this record is an older evaluated version, not the current software release.","source_ids":["kmetashot-2025"],"source_locator":"Materials and methods/kMetaShot testing on Human Microbiome Project genomes (paragraph 1); Materials and methods/GTDBtk taxonomy conversion (paragraph 1)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Bacterial or archaeal genome assemblies/MAGs","GTDB-Tk","GTDB taxonomic assignments"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["kmetashot-2025"],"source_locator":"Materials and methods/GTDBtk taxonomy conversion (paragraph 1); Materials and methods (paragraph 1)"},"coverage":"limited","gaps":["GTDB-Tk v1.0.2 in the reported CAMI II comparison","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-4058c43eb73b90","kind":"model","name":"Boltz-1","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["molecular-interactions"]},"source_ids":["boltz1-2025"],"links":[],"attributes":{"entity_level":"method","version":"3 recycling rounds; 200 diffusion steps","reported_name":"Boltz-1","historical_missing_metadata":{"checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Boltz-1 predicts three-dimensional biomolecular complexes with an atom-coordinate diffusion model.","summary_source_ids":["boltz1-2025"],"summary_source_locator":"Modeling/Confidence model (paragraph 2); Boltz steering/Method (paragraph 5)","sections":[{"title":"How the evaluated method works","body":"A token-and-pair representation trunk conditions a multiresolution transformer that denoises atom coordinates. The confidence model reuses trunk components, including a 48-layer PairFormer, to assess the generated structure.","source_ids":["boltz1-2025"],"source_locator":"Modeling/Confidence model (paragraph 2); Modeling/Optimizations/Attention bias sharing and caching (paragraph 2)"},{"title":"Underlying method and version boundaries","body":"The official Boltz repository publishes separate Boltz-1 and Boltz-2 models. Boltz-2 adds affinity prediction and other changes; these are not retroactively attributed to Boltz-1 evaluations.","source_ids":["evidence-reported-base-boltz-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies Boltz-1: Protein–ligand pose prediction. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["boltz1-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b3-039"}],"facts":[{"label":"Model type","value":"Biomolecular structure predictor; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-boltz-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"A token-and-pair representation trunk conditions a multiresolution transformer that denoises atom coordinates. The confidence model reuses trunk components, including a 48-layer PairFormer, to assess the generated structure.","status":"source_checked","source_ids":["boltz1-2025"],"source_locator":"Modeling/Confidence model (paragraph 2); Modeling/Optimizations/Attention bias sharing and caching (paragraph 2)"},{"label":"Biological inputs","value":"Protein, RNA or DNA sequences and small-molecule identities, with the model’s sequence and structural preprocessing","status":"source_checked","source_ids":["boltz1-2025"],"source_locator":"Results (paragraph 1); Data pipeline/Validation and test sets curation (paragraph 4)"},{"label":"Outputs","value":"Three-dimensional complexes and confidence estimates","status":"source_checked","source_ids":["boltz1-2025"],"source_locator":"Results/Evaluation criteria (paragraph 2); Results/Results (paragraph 2)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["boltz1-2025","evidence-reported-base-boltz-readme-md"],"source_locator":"Data pipeline/Data source and processing/Structure prediction training pipeline; Modeling/Architectural modifications/MSA module; Modeling/Architectural modifications/Transformer layer; Modeling/Training and inference procedures/Kabsch diffusion interpolation; Modeling/Training and inference procedures/Diffusion loss weighting; Modeling/Confidence model/Trunk architecture and initialization; Modeling/Confidence model/Overall procedure and training; Boltz steering/Method; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"3 recycling rounds; 200 diffusion steps","status":"source_checked","source_ids":["boltz1-2025"],"source_locator":"Table T4 (paragraph 1); Results/Baselines (paragraph 2)"},{"label":"Training data / fitting","value":"PDB structures and an OpenFold distillation dataset of approximately 270,000 structures; the last training stage uses PDB structures only.","status":"source_checked","source_ids":["boltz1-2025"],"source_locator":"Data pipeline/Data source and processing/Structure prediction training pipeline (paragraph 1); Modeling/Training and inference procedures/Kabsch diffusion interpolation (paragraph 3)"},{"label":"Context limits","value":"Training crops use 384 then 512 tokens and 3,456 then 4,608 atoms; these are training crop sizes, not declared inference maxima.","status":"source_checked","source_ids":["boltz1-2025"],"source_locator":"Data pipeline/Unified cropping algorithm (paragraph 1); Modeling/Optimizations/Sequence-local atom representation (paragraph 1)"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://github.com/jwohlwend/boltz/blob/b1ebfc46ecf57f5414e0d1a6f9027bbb122c53bc/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-boltz-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"MIT (upstream repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-base-boltz-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"MIT, as explicitly stated in the paper for the released weights.","status":"source_checked","source_ids":["boltz1-2025"],"source_locator":"Results/Baselines (paragraph 1); Boltz steering/Method (paragraph 6)"}],"strengths":[{"text":"Publishes model weights and training/inference implementation, allowing inspection of the structure-prediction procedure.","source_ids":["boltz1-2025"],"source_locator":"Modeling/Confidence model/Overall procedure and training (paragraph 1); Modeling/Training and inference procedures/Kabsch diffusion interpolation (paragraph 3)"}],"limitations":[{"text":"Atom-coordinate generation can produce stereochemical violations; confidence estimates do not replace checks of physical plausibility.","source_ids":["boltz1-2025"],"source_locator":"Results/Results (paragraph 4); Modeling/Confidence model (paragraph 2)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Protein, RNA or DNA sequences and small-molecule identities, with the model’s sequence and structural preprocessing","Boltz-1","Three-dimensional complexes and confidence estimates"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["boltz1-2025"],"source_locator":"Modeling/Confidence model (paragraph 2); Modeling/Optimizations/Attention bias sharing and caching (paragraph 2)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-40ce004270dee4","kind":"model","name":"ESM2 OFS pseudo-perplexity","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["proteins-complexes"]},"source_ids":["esm2-ofs-fitness-2025"],"links":[],"attributes":{"entity_level":"method","version":"not stated in table","reported_name":"ESM2 OFS pseudo-perplexity","historical_missing_metadata":{"checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"One Fell Swoop approximates ESM-2 pseudo-perplexity from a single unmasked-sequence embedding pass.","summary_source_ids":["esm2-ofs-fitness-2025"],"summary_source_locator":"Section III, One Fell Swoop; Figure 1; Section IV fitness evaluation","sections":[{"title":"How the evaluated method works","body":"Eight multilayer perceptrons map last-layer ESM-2 650M residue embeddings to the masked residue probability profiles that would otherwise require repeated model passes.","source_ids":["esm2-ofs-fitness-2025"],"source_locator":"Section III, One Fell Swoop; Figure 1; Section IV fitness evaluation"},{"title":"Underlying method and version boundaries","body":"ESM-2 is a transformer protein language-model family. The official repository exposes residue embeddings, sequence-level pooling and models at several sizes; the study configuration determines which of these is evaluated.","source_ids":["evidence-reported-base-esm-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies ESM2 OFS pseudo-perplexity: protein variant fitness prediction. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["esm2-ofs-fitness-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-b2-esm2-ofs-fitness-2025"}],"facts":[{"label":"Model type","value":"Protein sequence transformer; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-esm-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"Eight multilayer perceptrons map last-layer ESM-2 650M residue embeddings to the masked residue probability profiles that would otherwise require repeated model passes.","status":"source_checked","source_ids":["esm2-ofs-fitness-2025"],"source_locator":"Section III, One Fell Swoop; Figure 1; Section IV fitness evaluation"},{"label":"Biological inputs","value":"Protein sequence and unmasked ESM-2 residue embeddings","status":"source_checked","source_ids":["esm2-ofs-fitness-2025"],"source_locator":"Section III, One Fell Swoop; Figure 1; Section IV fitness evaluation"},{"label":"Outputs","value":"Approximate residue probability profiles and OFS pseudo-perplexity","status":"source_checked","source_ids":["esm2-ofs-fitness-2025"],"source_locator":"Section III, One Fell Swoop; Figure 1; Section IV fitness evaluation"},{"label":"Parameters","value":"650-million-parameter ESM-2 encoder plus a 7-million-parameter OFS ensemble.","status":"source_checked","source_ids":["esm2-ofs-fitness-2025"],"source_locator":"Section III, One Fell Swoop; Figure 1; Section IV fitness evaluation"},{"label":"Known versions / configuration","value":"ESM2 OFS pseudo-perplexity is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["esm2-ofs-fitness-2025"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"Cluster-representative sequences provide one-at-a-time masked profiles for a cross-entropy distillation objective.","status":"source_checked","source_ids":["esm2-ofs-fitness-2025"],"source_locator":"Section III, One Fell Swoop; Figure 1; Section IV fitness evaluation"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["esm2-ofs-fitness-2025","evidence-reported-base-esm-readme-md"],"source_locator":"Complete primary text and named comparison table; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://github.com/facebookresearch/esm/blob/2b369911bb5b4b0dda914521b9475cad1656b2ac/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-esm-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"MIT (upstream repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-base-esm-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-base-esm-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Reduces repeated encoder passes while retaining an approximation to the sequence-wide masked profile.","source_ids":["esm2-ofs-fitness-2025"],"source_locator":"Section III, One Fell Swoop; Figure 1; Section IV fitness evaluation"}],"limitations":[{"text":"The output is an approximation, not exact pseudo-perplexity; the paper reports task-specific trade-offs against other mutation-scoring heuristics.","source_ids":["esm2-ofs-fitness-2025"],"source_locator":"Section III, One Fell Swoop; Figure 1; Section IV fitness evaluation"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Protein sequence and unmasked ESM-2 residue embeddings","ESM2 OFS pseudo-perplexity","Approximate residue probability profiles and OFS pseudo-perplexity"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["esm2-ofs-fitness-2025"],"source_locator":"Section III, One Fell Swoop; Figure 1; Section IV fitness evaluation"},"coverage":"limited","gaps":["ESM2 OFS pseudo-perplexity is the comparison-table label; that label does not specify an immutable weight revision.","A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-415ee22f46526c","kind":"model","name":"DiffDock","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["molecular-interactions"]},"source_ids":["mpro-pose-affinity-2025"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"DiffDock","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"This configuration predicts ligand poses for the SARS-CoV-2 and MERS-CoV main proteases in the ASAP challenge setting.","summary_source_ids":["mpro-pose-affinity-2025"],"summary_source_locator":"Conclusions (paragraph 1); Introduction (paragraph 2)","sections":[{"title":"How the evaluated method works","body":"The receptor PDB and ligand SMILES enter the official default DiffDock configuration. Multiple poses are generated, internally confidence-ranked and the top-scoring pose is selected.","source_ids":["mpro-pose-affinity-2025"],"source_locator":"Methodologies/Pose Prediction/Deep Learning-Based Modeling with DiffDock (paragraph 1); Methodologies/Pose Prediction/Deep\nLearning-Based Modeling with Boltz-2 (paragraph 1)"},{"title":"Underlying method and version boundaries","body":"DiffDock is a molecular-docking implementation that produces ligand poses and confidence estimates. Its confidence values and predicted coordinates are different outputs from an experimentally calibrated binding-affinity measurement.","source_ids":["evidence-reported-base-diffdock-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies DiffDock: Ligand potency prediction using generated poses. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["mpro-pose-affinity-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-048"}],"facts":[{"label":"Model type","value":"Molecular docking model; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-diffdock-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"The receptor PDB and ligand SMILES enter the official default DiffDock configuration. Multiple poses are generated, internally confidence-ranked and the top-scoring pose is selected.","status":"source_checked","source_ids":["mpro-pose-affinity-2025"],"source_locator":"Methodologies/Pose Prediction/Deep Learning-Based Modeling with DiffDock (paragraph 1); Methodologies/Pose Prediction/Deep\nLearning-Based Modeling with Boltz-2 (paragraph 1)"},{"label":"Biological inputs","value":"Main-protease protein information and ligand structures/SMILES","status":"source_checked","source_ids":["mpro-pose-affinity-2025"],"source_locator":"Methodologies/Pose Prediction/Deep Learning-Based Modeling with DiffDock (paragraph 1); Methodologies/Pose Prediction/Deep\nLearning-Based Modeling with Boltz-2 (paragraph 1)"},{"label":"Outputs","value":"Predicted ligand binding poses","status":"source_checked","source_ids":["mpro-pose-affinity-2025"],"source_locator":"Methodologies/Pose Prediction/Deep Learning-Based Modeling with Gnina (paragraph 2); Methodologies/Potency Prediction/LRIP-SF/Pose Generation (paragraph 2)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["mpro-pose-affinity-2025","evidence-reported-base-diffdock-readme-md"],"source_locator":"Methodologies/Pose Prediction/Data Preparation; Methodologies/Pose Prediction/Molecular Docking with\nGlide; Methodologies/Pose Prediction/Molecular Docking with AutoDock Vina; Methodologies/Pose Prediction/Flexible\nLigand Superposition with FlexS; Methodologies/Pose Prediction/Deep Learning-Based Modeling with AlphaFold3; Methodologies/Pose Prediction/Deep Learning-Based Modeling with DiffDock; Methodologies/Pose Prediction/Deep\nLearning-Based Modeling with Boltz-2; Methodologies/Pose Prediction/Deep Learning-Based Modeling with Gnina; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"DiffDock is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["mpro-pose-affinity-2025"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"The study contains 770 SARS-CoV-2 training complexes and test sets of 98 SARS-CoV-2 and 97 MERS-CoV complexes; these are study partitions, not necessarily upstream-model training corpora.","status":"source_checked","source_ids":["mpro-pose-affinity-2025"],"source_locator":"Methodologies/Pose Prediction/Data Preparation (paragraph 2); Results/Potency Prediction (paragraph 5)"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["mpro-pose-affinity-2025","evidence-reported-base-diffdock-readme-md"],"source_locator":"Methodologies/Pose Prediction/Data Preparation; Methodologies/Pose Prediction/Molecular Docking with\nGlide; Methodologies/Pose Prediction/Molecular Docking with AutoDock Vina; Methodologies/Pose Prediction/Flexible\nLigand Superposition with FlexS; Methodologies/Pose Prediction/Deep Learning-Based Modeling with AlphaFold3; Methodologies/Pose Prediction/Deep Learning-Based Modeling with DiffDock; Methodologies/Pose Prediction/Deep\nLearning-Based Modeling with Boltz-2; Methodologies/Pose Prediction/Deep Learning-Based Modeling with Gnina; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://github.com/gcorso/DiffDock/blob/85c49b60d3e0b0182a59ee43a34a6d7036981284/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-diffdock-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"MIT (upstream repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-base-diffdock-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-base-diffdock-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Evaluates pose generation and downstream affinity estimation separately.","source_ids":["mpro-pose-affinity-2025"],"source_locator":"Conclusions (paragraph 3); Discussion (paragraph 10)"}],"limitations":[{"text":"An affinity result obtained by LRIP-SF on a predicted pose must not be assigned to the pose generator alone.","source_ids":["mpro-pose-affinity-2025"],"source_locator":"Methodologies/Potency Prediction/LRIP-SF/Pose Generation (paragraph 1); Results/Potency Prediction (paragraph 3)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Main-protease protein information and ligand structures/SMILES","DiffDock","Predicted ligand binding poses"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["mpro-pose-affinity-2025"],"source_locator":"Methodologies/Pose Prediction/Deep Learning-Based Modeling with DiffDock (paragraph 1); Methodologies/Pose Prediction/Deep\nLearning-Based Modeling with Boltz-2 (paragraph 1)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","DiffDock is the comparison-table label; that label does not specify an immutable weight revision.","A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-41ae49bb40ed8e","kind":"model","name":"ProteinBERT LLM-encoding model","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["rna-transcriptomes"]},"source_ids":["mrna-protein-diversity-2026"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"ProteinBERT LLM-encoding model","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"This mRNA–protein interaction predictor uses ProteinBERT to encode RNA-binding-protein sequences.","summary_source_ids":["mrna-protein-diversity-2026"],"summary_source_locator":"Results and discussions/Limitations of sequence-based deep-learning models (paragraph 6); Datasets and methods (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"ProteinBERT-derived protein features enter the paper’s attention-based mRNA–protein prediction framework alongside the RNA representation.","source_ids":["mrna-protein-diversity-2026"],"source_locator":"Datasets and methods/The modern attention-based deep network for mRPI identification/Part 2: binding pattern integration (paragraph 1); Datasets and methods/The implemented RBP encoding schemes (paragraph 1)"},{"title":"What was evaluated","body":"The linked evaluation record identifies ProteinBERT LLM-encoding model: mRNA-protein interaction prediction. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["mrna-protein-diversity-2026"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b4-008"}],"facts":[{"label":"Model type","value":"Study-specific predictive method; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["mrna-protein-diversity-2026"],"source_locator":"Datasets and methods/The modern attention-based deep network for mRPI identification/Part 2: binding pattern integration (paragraph 1); Datasets and methods/The implemented RBP encoding schemes (paragraph 1)"},{"label":"Architecture / procedure","value":"ProteinBERT-derived protein features enter the paper’s attention-based mRNA–protein prediction framework alongside the RNA representation.","status":"source_checked","source_ids":["mrna-protein-diversity-2026"],"source_locator":"Datasets and methods/The modern attention-based deep network for mRPI identification/Part 2: binding pattern integration (paragraph 1); Datasets and methods/The implemented RBP encoding schemes (paragraph 1)"},{"label":"Biological inputs","value":"mRNA and RNA-binding-protein sequences","status":"source_checked","source_ids":["mrna-protein-diversity-2026"],"source_locator":"Results and discussions/Limitations of sequence-based deep-learning models (paragraph 6); Datasets and methods/Evaluation metrics (paragraph 1)"},{"label":"Outputs","value":"mRNA–protein interaction predictions","status":"source_checked","source_ids":["mrna-protein-diversity-2026"],"source_locator":"Results and discussions/Limitations of sequence-based deep-learning models (paragraph 6); Results and discussions/Limitations of sequence-based deep-learning models (paragraph 5)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["mrna-protein-diversity-2026","evidence-reported-mrna-protein-diversity-2026-readme-md"],"source_locator":"Datasets and methods; Datasets and methods/Preparation of the mRNA-protein interaction ground-truth dataset; Datasets and methods/The evaluated data partitioning schemes; Datasets and methods/The implemented RBP encoding schemes; Datasets and methods/The implemented RBP encoding schemes/One-hot encoding of RBPs and mRNA fragments; Datasets and methods/The implemented RBP encoding schemes/LLM-encoding of RBPs; Datasets and methods/The implemented RBP encoding schemes/Structure-aware encoding of RBPs; Datasets and methods/The modern attention-based deep network for mRPI identification; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"ProteinBERT LLM-encoding model is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["mrna-protein-diversity-2026"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"CLIP-derived interaction pairs with separate random-pair and RBP-aware partitions.","status":"source_checked","source_ids":["mrna-protein-diversity-2026"],"source_locator":"Datasets and methods (paragraph 1); Datasets and methods/The evaluated data partitioning schemes (paragraph 1)"},{"label":"Context limits","value":"Protein inputs are padded to 2,804 residues and mRNA fragments to 1,024 nucleotides in the evaluated dataset.","status":"source_checked","source_ids":["mrna-protein-diversity-2026"],"source_locator":"Datasets and methods/The implemented RBP encoding schemes (paragraph 1); Datasets and methods/The implemented RBP encoding schemes/One-hot encoding of RBPs and mRNA fragments (paragraph 1)"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/cobisLab/mRPI-issue/blob/0f2d27666c4876f00d4c4e6cb3bec0e1629d214c/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-mrna-protein-diversity-2026-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","status":"unreported","source_ids":["evidence-reported-mrna-protein-diversity-2026-readme-md"],"source_locator":"README.md and repository-root licence-file search"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-mrna-protein-diversity-2026-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"The benchmark explicitly tests whether an interaction model generalises to proteins excluded from training.","source_ids":["mrna-protein-diversity-2026"],"source_locator":"Abstract (paragraph 1); Introduction (paragraph 4)"}],"limitations":[{"text":"Random interaction-level splits permit protein overlap; their performance must not be treated as unseen-RBP generalisation.","source_ids":["mrna-protein-diversity-2026"],"source_locator":"Results and discussions/Limitations of sequence-based deep-learning models (paragraph 2); Results and discussions/RBP-aware cross-validation further confirmed the requirement for the RBP-aware test set and RBP diversity (paragraph 1)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["mRNA and RNA-binding-protein sequences","ProteinBERT LLM-encoding model","mRNA–protein interaction predictions"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["mrna-protein-diversity-2026"],"source_locator":"Datasets and methods/The modern attention-based deep network for mRPI identification/Part 2: binding pattern integration (paragraph 1); Datasets and methods/The implemented RBP encoding schemes (paragraph 1)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","ProteinBERT LLM-encoding model is the comparison-table label; that label does not specify an immutable weight revision.","No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-43cf51abca83d1","kind":"model","name":"RNA-FM","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["rna-transcriptomes"]},"source_ids":["mrnabench-2025"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"RNA-FM","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"This RNA model is evaluated as a representation source for mature-mRNA prediction in mRNABench.","summary_source_ids":["mrnabench-2025"],"summary_source_locator":"Introduction/Sequence Compressibility: (paragraph 1); Conclusion (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"The benchmark extracts nucleotide-model embeddings and evaluates task-specific predictors under standardised protocols.","source_ids":["mrnabench-2025"],"source_locator":"Results/Nucleotide foundation models perform poorly at compositional generalization (paragraph 3); Methods/Linear Probing (paragraph 1)"},{"title":"Underlying method and version boundaries","body":"RNA-FM uses self-supervised pretraining on more than 23 million non-coding RNA sequences to produce general-purpose RNA embeddings. The distinct mRNA-FM model is trained on coding sequences; these names are not interchangeable.","source_ids":["evidence-reported-base-rnafm-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies RNA-FM: Mean ribosome load from MPRA. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["mrnabench-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-010"}],"facts":[{"label":"Model type","value":"RNA sequence language model; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-rnafm-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"The benchmark extracts nucleotide-model embeddings and evaluates task-specific predictors under standardised protocols.","status":"source_checked","source_ids":["mrnabench-2025"],"source_locator":"Results/Nucleotide foundation models perform poorly at compositional generalization (paragraph 3); Methods/Linear Probing (paragraph 1)"},{"label":"Biological inputs","value":"Mature mRNA sequences or benchmark-defined segments","status":"source_checked","source_ids":["mrnabench-2025"],"source_locator":"Conclusion (paragraph 1); Appendix Contents/Dataset Processing Protocols/Variant Effect Prediction (paragraph 1)"},{"label":"Outputs","value":"Embeddings and task-specific mRNA-property predictions","status":"source_checked","source_ids":["mrnabench-2025"],"source_locator":"Related Works/Deep learning for mRNA property prediction: (paragraph 1); Appendix Contents/Dataset Processing Protocols/Variant Effect Prediction (paragraph 1)"},{"label":"Parameters","value":"100M parameters for RNA-FM; mRNA-FM is a distinct 239M model.","status":"source_checked","source_ids":["mrnabench-2025"],"source_locator":"Table1; RNA-FM rows"},{"label":"Known versions / configuration","value":"rna-fm, distinct from the separately evaluated mrna-fm checkpoint.","status":"source_checked","source_ids":["mrnabench-2025"],"source_locator":"Introduction (paragraph 3); Conclusion (paragraph 1)"},{"label":"Training data / fitting","value":"Task-specific adaptation uses the mRNABench partitions; the encoder’s original RNA pretraining is distinct.","status":"source_checked","source_ids":["mrnabench-2025"],"source_locator":"Appendix Contents/Dataset Processing Protocols/Variant Effect Prediction (paragraph 1); Appendix Contents/Dataset Processing Protocols/Mean Ribosome Load (paragraph 1)"},{"label":"Context limits","value":"1,024-token context in Table 1; longer benchmark sequences are chunked for frozen linear probing.","status":"source_checked","source_ids":["mrnabench-2025"],"source_locator":"Table1 RNA-FM row; Methods / Linear Probing"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://github.com/ml4bio/RNA-FM/blob/348951516e0963d22bbb33b3c9fc18c89081d38e/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-rnafm-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"MIT (upstream repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-base-rnafm-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-base-rnafm-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Tests a broad set of mature-mRNA properties across ten datasets and 59 tasks.","source_ids":["mrnabench-2025"],"source_locator":"Conclusion (paragraph 1); Benchmarking Tasks (paragraph 1)"}],"limitations":[{"text":"A model pretrained mainly on non-coding RNA may have different strengths on mature-mRNA tasks; aggregate results do not identify a universal RNA-model winner.","source_ids":["mrnabench-2025"],"source_locator":"Results/Sequence compressibility predicts model performance across genomic regions (paragraph 4); Conclusion (paragraph 1)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Mature mRNA sequences or benchmark-defined segments","RNA-FM","Embeddings and task-specific mRNA-property predictions"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["mrnabench-2025"],"source_locator":"Results/Nucleotide foundation models perform poorly at compositional generalization (paragraph 3); Methods/Linear Probing (paragraph 1)"},"coverage":"limited","gaps":["The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-4438513d9cd42c","kind":"model","name":"ProkBERT-mini","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["microbes-communities"]},"source_ids":["prokbert-2024"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"ProkBERT-mini","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"ProkBERT-mini learns microbial DNA representations with local-context-aware tokenisation.","summary_source_ids":["prokbert-2024"],"summary_source_locator":"2 Materials and methods/2.1 Sequence data/2.1.1 Sequence segmentation and tokenization (paragraph 6); 4 Conclusion (paragraph 5)","sections":[{"title":"How the evaluated method works","body":"A transformer encoder uses local-context-aware 6-mer tokenisation for the ProkBERT-mini variant, followed by self-supervised pretraining and task-specific fine-tuning.","source_ids":["prokbert-2024"],"source_locator":"2 Materials and methods/2.1 Sequence data/2.1.1 Sequence segmentation and tokenization (paragraph 6); 2 Materials and methods (paragraph 1)"},{"title":"What was evaluated","body":"The linked evaluation record identifies ProkBERT-mini: E. coli sigma70 promoter prediction. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["prokbert-2024"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-033"}],"facts":[{"label":"Model type","value":"Transformer representation pipeline; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["prokbert-2024"],"source_locator":"2 Materials and methods/2.1 Sequence data/2.1.1 Sequence segmentation and tokenization (paragraph 6); 2 Materials and methods (paragraph 1)"},{"label":"Architecture / procedure","value":"A transformer encoder uses local-context-aware 6-mer tokenisation for the ProkBERT-mini variant, followed by self-supervised pretraining and task-specific fine-tuning.","status":"source_checked","source_ids":["prokbert-2024"],"source_locator":"2 Materials and methods/2.1 Sequence data/2.1.1 Sequence segmentation and tokenization (paragraph 6); 2 Materials and methods (paragraph 1)"},{"label":"Biological inputs","value":"Microbial DNA sequences","status":"source_checked","source_ids":["prokbert-2024"],"source_locator":"1 Introduction (paragraph 7); 4 Conclusion (paragraph 10)"},{"label":"Outputs","value":"DNA embeddings and task-specific genomic classifications","status":"source_checked","source_ids":["prokbert-2024"],"source_locator":"2 Materials and methods/2.5 Applied metrics (paragraph 1); 3 Results and discussion/3.1 ProkBERT's learned representations capture genomic structure and phylogeny (paragraph 4)"},{"label":"Parameters","value":"Approximately 20 million parameters.","status":"source_checked","source_ids":["prokbert-2024"],"source_locator":"4 Conclusion (paragraph 8); 2 Materials and methods/2.2 Pretraining and learning sequence representations/2.2.2 Training process/2.2.2.2 Training phases and configuration (paragraph 1)"},{"label":"Known versions / configuration","value":"ProkBERT-mini is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["prokbert-2024"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"Unlabelled microbial genome sequences followed by supervised task data, as specified in the paper.","status":"source_checked","source_ids":["prokbert-2024"],"source_locator":"2 Materials and methods/2.4 Application II: phage sequence analysis/2.4.2 Model training for phage sequence analysis (paragraph 1); 2 Materials and methods/2.2 Pretraining and learning sequence representations/2.2.4 Analysis of encoder outputs (paragraph 6)"},{"label":"Context limits","value":"Approximately 1 kb for ProkBERT-mini; the distinct mini-long variant supports approximately 2 kb.","status":"source_checked","source_ids":["prokbert-2024"],"source_locator":"2 Materials and methods/2.1 Sequence data/2.1.1 Sequence segmentation and tokenization (paragraph 6); Table T3 (paragraph 1)"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/nbrg-ppcu/prokbert/blob/8670ae92b816cff158a0b85647a8dea122e251eb/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-prokbert-2024-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"MIT (study repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-prokbert-2024-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-prokbert-2024-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Targets microbial sequence distributions and evaluates promoter-related downstream tasks.","source_ids":["prokbert-2024"],"source_locator":"2 Materials and methods/2.3 Application I: bacterial promoter prediction/2.3.1 Dataset overview/2.3.1.2 Dataset construction for multispecies train, test and validation sets (paragraph 6); 3 Results and discussion/3.4 ProkBERT swiftly and accurately identifies phage sequences, even in challenging settings (paragraph 2)"}],"limitations":[{"text":"The mini checkpoint and tokenizer settings define the evaluated system; results cannot be transferred to every ProkBERT variant.","source_ids":["prokbert-2024"],"source_locator":"2 Materials and methods/2.1 Sequence data/2.1.1 Sequence segmentation and tokenization (paragraph 6); 3 Results and discussion/3.3 ProkBERT performs accurately and robustly in promoter sequence recognition (paragraph 8)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Microbial DNA sequences","ProkBERT-mini","DNA embeddings and task-specific genomic classifications"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["prokbert-2024"],"source_locator":"2 Materials and methods/2.1 Sequence data/2.1.1 Sequence segmentation and tokenization (paragraph 6); 2 Materials and methods (paragraph 1)"},"coverage":"limited","gaps":["ProkBERT-mini is the comparison-table label; that label does not specify an immutable weight revision.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-47521865af7b04","kind":"model","name":"Caduceus-Ph","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["dna-genomes"]},"source_ids":["dna-foundation-models-2025"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"Caduceus-Ph","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"This genomic-embedding configuration is compared across classification, gene-expression, variant-effect and domain-recognition tasks.","summary_source_ids":["dna-foundation-models-2025"],"summary_source_locator":"Abstract (paragraph 1); Introduction (paragraph 4)","sections":[{"title":"How the evaluated method works","body":"Frozen sequence representations are pooled and supplied to downstream analyses. The study explicitly compares summary-token, mean and maximum pooling.","source_ids":["dna-foundation-models-2025"],"source_locator":"Results/Sequence classification: pooling methods benchmark (paragraph 2); Methods/Benchmarking methods/Sequence classification benchmark: pooling methods (paragraph 4)"},{"title":"Underlying method and version boundaries","body":"Caduceus exposes distinct Ph and PS configurations. The documented Ph-131k checkpoint uses 16 layers, width 256 and reverse-complement data augmentation; PS implements reverse-complement equivariance. These training choices are not interchangeable.","source_ids":["evidence-reported-base-caduceus-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies Caduceus-Ph: Human 5mC detection. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["dna-foundation-models-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-001"}],"facts":[{"label":"Model type","value":"DNA state-space model; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-caduceus-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"Frozen sequence representations are pooled and supplied to downstream analyses. The study explicitly compares summary-token, mean and maximum pooling.","status":"source_checked","source_ids":["dna-foundation-models-2025"],"source_locator":"Results/Sequence classification: pooling methods benchmark (paragraph 2); Methods/Benchmarking methods/Sequence classification benchmark: pooling methods (paragraph 4)"},{"label":"Biological inputs","value":"DNA sequence windows, with model-specific context handling","status":"source_checked","source_ids":["dna-foundation-models-2025"],"source_locator":"Discussion/Limitations (paragraph 1); Methods/Benchmarking datasets/Sequence classification datasets (paragraph 1)"},{"label":"Outputs","value":"Sequence embeddings and downstream genomic predictions","status":"source_checked","source_ids":["dna-foundation-models-2025"],"source_locator":"Introduction (paragraph 1); Discussion (paragraph 1)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["dna-foundation-models-2025","evidence-reported-base-caduceus-readme-md"],"source_locator":"Results/Sequence classification: pooling methods benchmark; Results/Pre-training experiment; Methods/DNA foundation language models; Methods/Benchmarking datasets/Sequence classification datasets; Methods/Benchmarking datasets/Gene expression prediction datasets; Methods/Benchmarking datasets/Variant effect quantification datasets; Methods/Benchmarking datasets/TAD region dataset; Methods/Benchmarking methods/Sequence classification benchmark; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"Caduceus-Ph-131K","status":"source_checked","source_ids":["dna-foundation-models-2025"],"source_locator":"Table Tab4 (paragraph 1); Methods/Model configuration selection (paragraph 1)"},{"label":"Training data / fitting","value":"The benchmark extracts frozen Caduceus-Ph-131K embeddings and fits task-specific downstream classifiers or regressors; it does not retrain the DNA encoder. The exact pretraining-data revision is not supplied by this study.","status":"source_checked","source_ids":["dna-foundation-models-2025"],"source_locator":"Methods / Model configuration selection; Benchmarking methods / Gene expression and variant effect prediction"},{"label":"Context limits","value":"131,072 nucleotides; long-sequence tasks use the central window when necessary.","status":"source_checked","source_ids":["dna-foundation-models-2025"],"source_locator":"Methods/Benchmarking methods/Variant effect quantification benchmark (paragraph 3); Methods/Benchmarking datasets/TAD region dataset (paragraph 2)"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://github.com/kuleshov-group/caduceus/blob/0060a6d8079b6a040fc55d505e15972a327b70a6/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-caduceus-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"Apache 2.0 (upstream repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-base-caduceus-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-base-caduceus-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Tests pooling choices explicitly and compares task-specific specialists.","source_ids":["dna-foundation-models-2025"],"source_locator":"Methods/Benchmarking methods/Sequence classification benchmark (paragraph 3); Methods/Benchmarking methods/Sequence classification benchmark (paragraph 4)"}],"limitations":[{"text":"The paper finds task-dependent performance; a classifier on frozen features is not a label-free zero-shot prediction.","source_ids":["dna-foundation-models-2025"],"source_locator":"Results/Gene expression prediction (paragraph 5); Results/Sequence classification: pooling methods benchmark (paragraph 1)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["DNA sequence windows, with model-specific context handling","Caduceus-Ph","Sequence embeddings and downstream genomic predictions"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["dna-foundation-models-2025"],"source_locator":"Results/Sequence classification: pooling methods benchmark (paragraph 2); Methods/Benchmarking methods/Sequence classification benchmark: pooling methods (paragraph 4)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-4921459942b45f","kind":"model","name":"ESM-2 650M embeddings + classifier","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["proteins-complexes"]},"source_ids":["antibody-deamidation-plm-2024"],"links":[],"attributes":{"entity_level":"method","version":"esm2_t33_650m_UR50D","reported_name":"ESM-2 650M embeddings + classifier","historical_missing_metadata":{"checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"This antibody-deamidation pipeline combines ESM-2 sequence embeddings with local sequence information to identify susceptible residues.","summary_source_ids":["antibody-deamidation-plm-2024"],"summary_source_locator":"3. Results/3.4. Performance Evaluation of Models (paragraph 4); Abstract (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"A chimeric supervised predictor integrates pretrained protein-language-model embeddings and a local sequence branch. The score belongs to this complete antibody-specific predictor, not to ESM-2 alone.","source_ids":["antibody-deamidation-plm-2024"],"source_locator":"Abstract (paragraph 1); 4. Discussion and Conclusions (paragraph 2)"},{"title":"Underlying method and version boundaries","body":"ESM-2 is a transformer protein language-model family. The official repository exposes residue embeddings, sequence-level pooling and models at several sizes; the study configuration determines which of these is evaluated.","source_ids":["evidence-reported-base-esm-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies ESM-2 650M embeddings + classifier: antibody deamidation-site prediction. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["antibody-deamidation-plm-2024"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-b2-antibody-deamidation-plm-2024"}],"facts":[{"label":"Model type","value":"Protein sequence transformer; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-esm-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"A chimeric supervised predictor integrates pretrained protein-language-model embeddings and a local sequence branch. The score belongs to this complete antibody-specific predictor, not to ESM-2 alone.","status":"source_checked","source_ids":["antibody-deamidation-plm-2024"],"source_locator":"Abstract (paragraph 1); 4. Discussion and Conclusions (paragraph 2)"},{"label":"Biological inputs","value":"Antibody amino-acid sequences and local residue context","status":"source_checked","source_ids":["antibody-deamidation-plm-2024"],"source_locator":"1. Introduction (paragraph 5); 3. Results/3.2. The Use of ESM-2 Embedding for Deamidation Site Prediction (paragraph 4)"},{"label":"Outputs","value":"Deamidation propensity and extent predictions","status":"source_checked","source_ids":["antibody-deamidation-plm-2024"],"source_locator":"3. Results/3.4. Performance Evaluation of Models (paragraph 4); 3. Results/3.2. The Use of ESM-2 Embedding for Deamidation Site Prediction (paragraph 1)"},{"label":"Parameters","value":"650-million-parameter ESM-2 backbone; the total trained pipeline parameter count is not established here.","status":"unreported","source_ids":["antibody-deamidation-plm-2024"],"source_locator":"3. Results/3.2. The Use of ESM-2 Embedding for Deamidation Site Prediction (paragraph 2); 4. Discussion and Conclusions (paragraph 6)"},{"label":"Known versions / configuration","value":"esm2_t33_650m_UR50D","status":"source_checked","source_ids":["antibody-deamidation-plm-2024"],"source_locator":"3. Results/3.2. The Use of ESM-2 Embedding for Deamidation Site Prediction (paragraph 2); Table antibodies-13-00074-t002 (paragraph 1)"},{"label":"Training data / fitting","value":"An antibody deamidation dataset of 2,285 observations assembled with automated peptide mapping.","status":"source_checked","source_ids":["antibody-deamidation-plm-2024"],"source_locator":"3. Results/3.5. Independent Dataset Predicting Deamidation Hot Spots (paragraph 1); 3. Results/3.5. Independent Dataset Predicting Deamidation Hot Spots (paragraph 2)"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["antibody-deamidation-plm-2024","evidence-reported-base-esm-readme-md"],"source_locator":"2. Materials and Methods/2.1. Chemicals and Reagents; 2. Materials and Methods/2.2. Accelerated Thermal Stress ; 2. Materials and Methods/2.3. Automated Peptide Mapping; 2. Materials and Methods/2.4. LC-MS/MS Analysis ; 3. Results/3.7. Model Implementation for High-Throughput Screening Drug Candidates; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://github.com/facebookresearch/esm/blob/2b369911bb5b4b0dda914521b9475cad1656b2ac/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-esm-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"MIT (upstream repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-base-esm-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-base-esm-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Uses sequence inputs without requiring an experimentally solved antibody structure.","source_ids":["antibody-deamidation-plm-2024"],"source_locator":"3. Results/3.5. Independent Dataset Predicting Deamidation Hot Spots (paragraph 1); 3. Results/3.3. Enhanced Prediction by Combining ESM-2 Embedding and Local Sequence Information (paragraph 1)"}],"limitations":[{"text":"Training labels reflect the study’s antibody modalities and peptide-mapping conditions; transfer to other assay conditions requires evidence.","source_ids":["antibody-deamidation-plm-2024"],"source_locator":"2. Materials and Methods/2.2. Accelerated Thermal Stress  (paragraph 1); Abstract (paragraph 1)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Antibody amino-acid sequences and local residue context","ESM-2 650M embeddings + classifier","Deamidation propensity and extent predictions"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["antibody-deamidation-plm-2024"],"source_locator":"Abstract (paragraph 1); 4. Discussion and Conclusions (paragraph 2)"},"coverage":"limited","gaps":["650-million-parameter ESM-2 backbone; the total trained pipeline parameter count is not established here.","A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-49bc768f46b366","kind":"model","name":"CATHe2 + ProstT5","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["proteins-complexes"]},"source_ids":["cathe2-2025"],"links":[],"attributes":{"entity_level":"method","version":"full ProstT5","reported_name":"CATHe2 + ProstT5","historical_missing_metadata":{"checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"CATHe2 classifies protein domains into CATH superfamilies using ProstT5 representations of sequence and structure.","summary_source_ids":["cathe2-2025"],"summary_source_locator":"Abstract (paragraph 1); Introduction (paragraph 5)","sections":[{"title":"How the evaluated method works","body":"A feed-forward classifier consumes ProstT5 amino-acid embeddings and 3Di structural-alphabet embeddings. This record is the complete CATHe2 classifier with the full ProstT5 feature extractor.","source_ids":["cathe2-2025"],"source_locator":"Abstract (paragraph 1); Results/CATHe2 models performance (paragraph 3)"},{"title":"What was evaluated","body":"The linked evaluation record identifies CATHe2 + ProstT5: CATH superfamily annotation. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["cathe2-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-b2-cathe2-2025"}],"facts":[{"label":"Model type","value":"Study-specific predictive method; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["cathe2-2025"],"source_locator":"Abstract (paragraph 1); Results/CATHe2 models performance (paragraph 3)"},{"label":"Architecture / procedure","value":"A feed-forward classifier consumes ProstT5 amino-acid embeddings and 3Di structural-alphabet embeddings. This record is the complete CATHe2 classifier with the full ProstT5 feature extractor.","status":"source_checked","source_ids":["cathe2-2025"],"source_locator":"Abstract (paragraph 1); Results/CATHe2 models performance (paragraph 3)"},{"label":"Biological inputs","value":"Protein-domain amino-acid sequences and 3Di structural-alphabet sequences","status":"source_checked","source_ids":["cathe2-2025"],"source_locator":"Abstract (paragraph 1); Introduction (paragraph 2)"},{"label":"Outputs","value":"CATH superfamily assignments","status":"source_checked","source_ids":["cathe2-2025"],"source_locator":"Abstract (paragraph 1); Discussion/Critical analysis of the study (paragraph 1)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["cathe2-2025","evidence-reported-cathe2-2025-readme-md"],"source_locator":"Materials and methods/Data and datasets; Materials and methods/Additional experiments; Materials and methods/3Di processing; Materials and methods/Models; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"full ProstT5","status":"source_checked","source_ids":["cathe2-2025"],"source_locator":"Table bpaf080-T3 (paragraph 1); Results/CATHe2 models performance (paragraph 1)"},{"label":"Training data / fitting","value":"CATH-Gene3D-derived domains; the larger training set has 1,039,135 domains, with less than 20% sequence identity between and within train, validation and test partitions.","status":"source_checked","source_ids":["cathe2-2025"],"source_locator":"Materials and methods/Data and datasets (paragraph 2); Materials and methods/Data and datasets (paragraph 1)"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["cathe2-2025","evidence-reported-cathe2-2025-readme-md"],"source_locator":"Materials and methods/Data and datasets; Materials and methods/Additional experiments; Materials and methods/3Di processing; Materials and methods/Models; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/Mouret-Orfeu/CATHe2/blob/cdaf5f4c8d9dfda86f78fd0efc83ab22a30e9ffd/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-cathe2-2025-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"MIT (study repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-cathe2-2025-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-cathe2-2025-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Combines sequence and structural representations and evaluates remote homology under a stringent identity filter.","source_ids":["cathe2-2025"],"source_locator":"Materials and methods/Data and datasets (paragraph 2); Introduction (paragraph 3)"}],"limitations":[{"text":"The evaluated label space covers selected represented CATH superfamilies; it does not establish classification of every rare or unseen superfamily.","source_ids":["cathe2-2025"],"source_locator":"Materials and methods/Data and datasets (paragraph 3); Materials and methods/Data and datasets (paragraph 2)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Protein-domain amino-acid sequences and 3Di structural-alphabet sequences","CATHe2 + ProstT5","CATH superfamily assignments"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["cathe2-2025"],"source_locator":"Abstract (paragraph 1); Results/CATHe2 models performance (paragraph 3)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-4c73500c39e9d0","kind":"model","name":"ESM2 650M","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["proteins-complexes"]},"source_ids":["viral-immune-mimicry-2025"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"ESM2 650M","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"This classifier uses ESM-2 650M protein representations to distinguish viral from human proteins.","summary_source_ids":["viral-immune-mimicry-2025"],"summary_source_locator":"3. Results (paragraph 1); 3. Results/3.4. Latent Structure Embeddings Clustering (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"Pretrained protein embeddings feed an explainable downstream classifier. Misclassified viral proteins are analysed for associations with immune-mimicry-related characteristics.","source_ids":["viral-immune-mimicry-2025"],"source_locator":"3. Results/3.2. Error Analysis Models Insights (paragraph 1); Abstract (paragraph 1)"},{"title":"Underlying method and version boundaries","body":"ESM-2 is a transformer protein language-model family. The official repository exposes residue embeddings, sequence-level pooling and models at several sizes; the study configuration determines which of these is evaluated.","source_ids":["evidence-reported-base-esm-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies ESM2 650M: human-versus-viral protein classification. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["viral-immune-mimicry-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b4-009"}],"facts":[{"label":"Model type","value":"Protein sequence transformer; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-esm-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"Pretrained protein embeddings feed an explainable downstream classifier. Misclassified viral proteins are analysed for associations with immune-mimicry-related characteristics.","status":"source_checked","source_ids":["viral-immune-mimicry-2025"],"source_locator":"3. Results/3.2. Error Analysis Models Insights (paragraph 1); Abstract (paragraph 1)"},{"label":"Biological inputs","value":"Protein sequences represented by pretrained embeddings","status":"source_checked","source_ids":["viral-immune-mimicry-2025"],"source_locator":"2. Materials and Methods/2.2. Pretrained Deep Language Models (ESM, T5) (paragraph 1); 3. Results/3.4. Latent Structure Embeddings Clustering (paragraph 1)"},{"label":"Outputs","value":"Viral-versus-human classification scores","status":"source_checked","source_ids":["viral-immune-mimicry-2025"],"source_locator":"3. Results/3.5. Immunogenicity Analysis (paragraph 1); 3. Results (paragraph 1)"},{"label":"Parameters","value":"650-million-parameter ESM-2 backbone; downstream classifier parameters are additional.","status":"source_checked","source_ids":["viral-immune-mimicry-2025"],"source_locator":"2. Materials and Methods/2.3. Human-Virus Model Training and Implementation (paragraph 2); 2. Materials and Methods/2.2. Pretrained Deep Language Models (ESM, T5) (paragraph 1)"},{"label":"Known versions / configuration","value":"ESM2 650M is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["viral-immune-mimicry-2025"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"UniRef90 duplicate removal and a UniRef50-clustered 80:20 split; sequences above 1,600 residues are excluded.","status":"source_checked","source_ids":["viral-immune-mimicry-2025"],"source_locator":"2. Materials and Methods/2.1. Protein Datasets (paragraph 2); 2. Materials and Methods/2.1. Protein Datasets (paragraph 1)"},{"label":"Context limits","value":"The study excludes proteins longer than 1,600 residues.","status":"source_checked","source_ids":["viral-immune-mimicry-2025"],"source_locator":"2. Materials and Methods/2.1. Protein Datasets (paragraph 2); 3. Results/3.3. Virus Errors Analysis (paragraph 2)"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://github.com/facebookresearch/esm/blob/2b369911bb5b4b0dda914521b9475cad1656b2ac/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-esm-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"MIT (upstream repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-base-esm-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-base-esm-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Cluster-based splitting reduces close-sequence overlap between train and test.","source_ids":["viral-immune-mimicry-2025"],"source_locator":"2. Materials and Methods/2.1. Protein Datasets (paragraph 2); 2. Materials and Methods/2.4. Finding and Analyzing Model Mistakes (paragraph 1)"}],"limitations":[{"text":"Classification errors correlated with immune features do not establish that the model measures actual immune recognition or viral pathogenicity.","source_ids":["viral-immune-mimicry-2025"],"source_locator":"3. Results (paragraph 1); Abstract (paragraph 1)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Protein sequences represented by pretrained embeddings","ESM2 650M","Viral-versus-human classification scores"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["viral-immune-mimicry-2025"],"source_locator":"3. Results/3.2. Error Analysis Models Insights (paragraph 1); Abstract (paragraph 1)"},"coverage":"limited","gaps":["ESM2 650M is the comparison-table label; that label does not specify an immutable weight revision.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-4ce8cae0f2eafc","kind":"model","name":"TCINet + HTRS","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["microbes-communities"]},"source_ids":["metagenomic-pathogens-2025"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"TCINet + HTRS","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"TCINet plus HTRS is a structured metagenomic inference pipeline for taxonomic identification and abundance.","summary_source_ids":["metagenomic-pathogens-2025"],"summary_source_locator":"3 Method/3.3 Taxon-aware Compositional Inference Network (TCINet)/3.3.1 Taxonomy-structured feature embedding (paragraph 6); 5 Conclusions and future work (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"TCINet produces taxonomic embeddings from reads, uses masked activations for sparse abundance estimates and log-normal variance modelling for uncertainty. Hierarchical taxonomic and ecological constraints form part of the inference procedure.","source_ids":["metagenomic-pathogens-2025"],"source_locator":"3 Method/3.3 Taxon-aware Compositional Inference Network (TCINet)/3.3.3 Phylogenetic and ecological regularization (paragraph 1); Abstract (paragraph 1)"},{"title":"What was evaluated","body":"The linked evaluation record identifies TCINet + HTRS: pathogen detection. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["metagenomic-pathogens-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b4-017"}],"facts":[{"label":"Model type","value":"Study-specific predictive method; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["metagenomic-pathogens-2025"],"source_locator":"3 Method/3.3 Taxon-aware Compositional Inference Network (TCINet)/3.3.3 Phylogenetic and ecological regularization (paragraph 1); Abstract (paragraph 1)"},{"label":"Architecture / procedure","value":"TCINet produces taxonomic embeddings from reads, uses masked activations for sparse abundance estimates and log-normal variance modelling for uncertainty. Hierarchical taxonomic and ecological constraints form part of the inference procedure.","status":"source_checked","source_ids":["metagenomic-pathogens-2025"],"source_locator":"3 Method/3.3 Taxon-aware Compositional Inference Network (TCINet)/3.3.3 Phylogenetic and ecological regularization (paragraph 1); Abstract (paragraph 1)"},{"label":"Biological inputs","value":"Metagenomic sequencing reads with taxonomic/ecological reference information","status":"source_checked","source_ids":["metagenomic-pathogens-2025"],"source_locator":"3 Method/3.3 Taxon-aware Compositional Inference Network (TCINet)/3.3.1 Taxonomy-structured feature embedding (paragraph 6); 1 Introduction (paragraph 2)"},{"label":"Outputs","value":"Taxonomic identification and abundance estimates","status":"source_checked","source_ids":["metagenomic-pathogens-2025"],"source_locator":"3 Method/3.3 Taxon-aware Compositional Inference Network (TCINet)/3.3.1 Taxonomy-structured feature embedding (paragraph 6); 3 Method/3.3 Taxon-aware Compositional Inference Network (TCINet) (paragraph 1)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["metagenomic-pathogens-2025"],"source_locator":"3 Method/3.1 Overview; 3 Method/3.2 Preliminaries; 3 Method/3.3 Taxon-aware Compositional Inference Network (TCINet); 3 Method/3.3 Taxon-aware Compositional Inference Network (TCINet)/3.3.1 Taxonomy-structured feature embedding; 3 Method/3.3 Taxon-aware Compositional Inference Network (TCINet)/3.3.2 Sparse and uncertain presence modeling; 3 Method/3.3 Taxon-aware Compositional Inference Network (TCINet)/3.3.3 Phylogenetic and ecological regularization; 3 Method/3.4 Hierarchical Taxonomic Reasoning Strategy (HTRS); 3 Method/3.4 Hierarchical Taxonomic Reasoning Strategy (HTRS)/3.4.1 Tree-Based Signal Aggregation; inspected for aggregate parameter count (component sizes are not added without an exact configuration)"},{"label":"Known versions / configuration","value":"TCINet + HTRS is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["metagenomic-pathogens-2025"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"Section 4.3 names MetaHIT and iHMP metagenomic experiments but does not specify sample accessions, split manifests or exact reference-label construction. The five-fold protocol in Section 4.2 applies to a separately described text/image experiment and cannot be assigned to these genomic rows.","status":"source_checked","source_ids":["metagenomic-pathogens-2025"],"source_locator":"4 Experimental setup/4.1 Dataset (paragraph 1); 4 Experimental setup/4.3 Comparison with SOTA methods (paragraph 5)"},{"label":"Context limits","value":"Section 4.2 describes 150-bp paired-end Illumina reads, adapter/quality filtering, host depletion, and 6-mer frequency vectors.","status":"source_checked","source_ids":["metagenomic-pathogens-2025"],"source_locator":"4 Experimental setup/4.2 Experimental details (paragraph 2); 4 Experimental setup/4.3 Comparison with SOTA methods (paragraph 5)"},{"label":"Access","value":"The source does not identify a public TCINet/HTRS implementation or checkpoint release. Its availability statement refers to the article/supplement and enquiries to the corresponding authors.","status":"unreported","source_ids":["metagenomic-pathogens-2025"],"source_locator":"Data availability statement; full-text search for code, GitHub and checkpoint access"},{"label":"Code licence","value":"No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","status":"unreported","source_ids":["metagenomic-pathogens-2025"],"source_locator":"Generative AI statement (paragraph 1); 5 Conclusions and future work (paragraph 2)"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["metagenomic-pathogens-2025"],"source_locator":"3 Method/3.3 Taxon-aware Compositional Inference Network (TCINet)/3.3.1 Taxonomy-structured feature embedding (paragraph 2); 3 Method/3.4 Hierarchical Taxonomic Reasoning Strategy (HTRS)/3.4.3 Context-aware and scalable inference (paragraph 3)"}],"strengths":[{"text":"Models uncertainty and taxonomic hierarchy explicitly.","source_ids":["metagenomic-pathogens-2025"],"source_locator":"3 Method/3.3 Taxon-aware Compositional Inference Network (TCINet)/3.3.2 Sparse and uncertain presence modeling (paragraph 1); 3 Method/3.4 Hierarchical Taxonomic Reasoning Strategy (HTRS)/3.4.1 Tree-Based Signal Aggregation (paragraph 2)"}],"limitations":[{"text":"The source combines inconsistent text/image and sequencing experiment descriptions. Section 4.3 does report MetaHIT/iHMP results, but the exact samples, splits and taxonomic ground truth remain unresolved; the table is not sufficient evidence of independently validated pathogen detection.","source_ids":["metagenomic-pathogens-2025"],"source_locator":"4 Experimental setup/4.3 Comparison with SOTA methods (paragraph 5); 3 Method/3.3 Taxon-aware Compositional Inference Network (TCINet) (paragraph 1)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Metagenomic sequencing reads with taxonomic/ecological reference information","TCINet + HTRS","Taxonomic identification and abundance estimates"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["metagenomic-pathogens-2025"],"source_locator":"3 Method/3.3 Taxon-aware Compositional Inference Network (TCINet)/3.3.3 Phylogenetic and ecological regularization (paragraph 1); Abstract (paragraph 1)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","TCINet + HTRS is the comparison-table label; that label does not specify an immutable weight revision.","The source does not identify a public TCINet/HTRS implementation or checkpoint release. Its availability statement refers to the article/supplement and enquiries to the corresponding authors.","No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-51ed86132346a0","kind":"model","name":"DiffDock-L","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["molecular-interactions"]},"source_ids":["lipp-2026"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"DiffDock-L","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"This configuration predicts lipid–protein poses in the LiPP benchmark.","summary_source_ids":["lipp-2026"],"summary_source_locator":"Results/Curation of the LiPP Benchmark\nSet for Modeling Lipid–Protein\nInteractions (paragraph 1); Materials\nand Methods/Curation of\nthe LiPP Test Set, Precutoff Set, and Similarity\nAssessments between the Different LiPP Data Sets (paragraph 2)","sections":[{"title":"How the evaluated method works","body":"DiffDock-L performs blind docking; the benchmark retains the highest-ranked pose and applies physical-plausibility checks without additional refinement.","source_ids":["lipp-2026"],"source_locator":"Materials\nand Methods/Molecular Docking and Structure Prediction\nProtocols (paragraph 1); Results/Physical\nPlausibility of Docking and Structure Prediction Generated\nModels of Lipid–Protein Complex (paragraph 2)"},{"title":"Underlying method and version boundaries","body":"DiffDock is a molecular-docking implementation that produces ligand poses and confidence estimates. Its confidence values and predicted coordinates are different outputs from an experimentally calibrated binding-affinity measurement.","source_ids":["evidence-reported-base-diffdock-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies DiffDock-L: Lipid–protein binding pose. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["lipp-2026"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-042"}],"facts":[{"label":"Model type","value":"Molecular docking model; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-diffdock-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"DiffDock-L performs blind docking; the benchmark retains the highest-ranked pose and applies physical-plausibility checks without additional refinement.","status":"source_checked","source_ids":["lipp-2026"],"source_locator":"Materials\nand Methods/Molecular Docking and Structure Prediction\nProtocols (paragraph 1); Results/Physical\nPlausibility of Docking and Structure Prediction Generated\nModels of Lipid–Protein Complex (paragraph 2)"},{"label":"Biological inputs","value":"Experimental protein and lipid coordinates, without a predefined binding pocket","status":"source_checked","source_ids":["lipp-2026"],"source_locator":"Materials\nand Methods/Molecular Docking and Structure Prediction\nProtocols (paragraph 1); Materials\nand Methods/DiffDock-L (paragraph 1)"},{"label":"Outputs","value":"Lipid–protein complex or docking-pose coordinates","status":"source_checked","source_ids":["lipp-2026"],"source_locator":"Materials\nand Methods/Protein Structure Errors (paragraph 1); Materials\nand Methods/Molecular Docking and Structure Prediction\nProtocols (paragraph 1)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["lipp-2026","evidence-reported-base-diffdock-readme-md"],"source_locator":"Materials\nand Methods/Curation of the LiPP Benchmark Set; Materials\nand Methods/Curation of\nthe LiPP Test Set, Precutoff Set, and Similarity\nAssessments between the Different LiPP Data Sets; Materials\nand Methods/Protein Similarity; Materials\nand Methods/Lipid Similarity; Materials\nand Methods/Molecular Docking and Structure Prediction\nProtocols; Materials\nand Methods/AlphaFold 3; Materials\nand Methods/Chai-1; Materials\nand Methods/RoseTTAFold AA; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"DiffDock-L1.1.3;20 diffusion steps, tenposes, expbeta noise schedule; best confidence pose is evaluated.","status":"source_checked","source_ids":["lipp-2026"],"source_locator":"Materials and Methods / DiffDock-L"},{"label":"Training data / fitting","value":"Table 1 identifies PDBbind complexes before 2019 as the training set. The LiPP evaluation uses released inference defaults without additional refinement or retraining.","status":"source_checked","source_ids":["lipp-2026"],"source_locator":"Table 1, DiffDock-L row; Materials and Methods / Molecular Docking and Structure Prediction Protocols"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["lipp-2026","evidence-reported-base-diffdock-readme-md"],"source_locator":"Materials\nand Methods/Curation of the LiPP Benchmark Set; Materials\nand Methods/Curation of\nthe LiPP Test Set, Precutoff Set, and Similarity\nAssessments between the Different LiPP Data Sets; Materials\nand Methods/Protein Similarity; Materials\nand Methods/Lipid Similarity; Materials\nand Methods/Molecular Docking and Structure Prediction\nProtocols; Materials\nand Methods/AlphaFold 3; Materials\nand Methods/Chai-1; Materials\nand Methods/RoseTTAFold AA; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://github.com/gcorso/DiffDock/blob/85c49b60d3e0b0182a59ee43a34a6d7036981284/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-diffdock-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"MIT (upstream repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-base-diffdock-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-base-diffdock-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Tests lipid chemistry and physical plausibility alongside RMSD.","source_ids":["lipp-2026"],"source_locator":"Results/Protein Prediction Errors in Structure Prediction Methods (paragraph 1); Materials\nand Methods/Protein Structure Errors (paragraph 1)"}],"limitations":[{"text":"Input information differs between docking and co-folding methods, and the post-cutoff test subset is much smaller than the full benchmark.","source_ids":["lipp-2026"],"source_locator":"Results/Protein Prediction Errors in Structure Prediction Methods (paragraph 1); Results/Curation of the LiPP Benchmark\nSet for Modeling Lipid–Protein\nInteractions (paragraph 3)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Experimental protein and lipid coordinates, without a predefined binding pocket","DiffDock-L","Lipid–protein complex or docking-pose coordinates"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["lipp-2026"],"source_locator":"Materials\nand Methods/Molecular Docking and Structure Prediction\nProtocols (paragraph 1); Results/Physical\nPlausibility of Docking and Structure Prediction Generated\nModels of Lipid–Protein Complex (paragraph 2)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-52b9c685d99290","kind":"model","name":"RLsite","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["rna-transcriptomes"]},"source_ids":["rlsite-rna-binding-2025"],"links":[],"attributes":{"entity_level":"method","version":"not stated in table","reported_name":"RLsite","historical_missing_metadata":{"checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"RLsite combines pretrained RNA representations with structural graphs to identify small-molecule-binding sites.","summary_source_ids":["rlsite-rna-binding-2025"],"summary_source_locator":"2 Materials and methods/2.4 Graph attention network/2.4.2 Feature aggregation and update (paragraph 2); 1 Introduction (paragraph 7)","sections":[{"title":"How the evaluated method works","body":"RNA-language-model sequence features and graph attention capture nucleotide context and spatial/topological relationships for binding-site prediction.","source_ids":["rlsite-rna-binding-2025"],"source_locator":"4 Conclusions/4.1 Hybrid architecture leveraging complementary modalities (paragraph 1); 2 Materials and methods/2.2 Model architecture overview (paragraph 1)"},{"title":"What was evaluated","body":"The linked evaluation record identifies RLsite: RNA-small-molecule binding-site prediction. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["rlsite-rna-binding-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-b2-rlsite-rna-binding-2025"}],"facts":[{"label":"Model type","value":"Graph-based predictive method; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["rlsite-rna-binding-2025"],"source_locator":"4 Conclusions/4.1 Hybrid architecture leveraging complementary modalities (paragraph 1); 2 Materials and methods/2.2 Model architecture overview (paragraph 1)"},{"label":"Architecture / procedure","value":"RNA-language-model sequence features and graph attention capture nucleotide context and spatial/topological relationships for binding-site prediction.","status":"source_checked","source_ids":["rlsite-rna-binding-2025"],"source_locator":"4 Conclusions/4.1 Hybrid architecture leveraging complementary modalities (paragraph 1); 2 Materials and methods/2.2 Model architecture overview (paragraph 1)"},{"label":"Biological inputs","value":"RNA sequence and graph-based RNA structural information","status":"source_checked","source_ids":["rlsite-rna-binding-2025"],"source_locator":"1 Introduction (paragraph 7); 3 Results and discussions/3.3 Complementarity of graph attention networks and RNA language models (paragraph 2)"},{"label":"Outputs","value":"Small-molecule-binding nucleotide predictions","status":"source_checked","source_ids":["rlsite-rna-binding-2025"],"source_locator":"2 Materials and methods/2.4 Graph attention network/2.4.2 Feature aggregation and update (paragraph 2); 3 Results and discussions/3.4 Consistent high performance of RLsite on the latest dataset T10 (paragraph 3)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["rlsite-rna-binding-2025","evidence-reported-rlsite-rna-binding-2025-readme-md"],"source_locator":"2 Materials and methods/2.1 Benchmark datasets; 2 Materials and methods/2.2 Model architecture overview; 2 Materials and methods/2.3 Pre-trained RNA language model; 2 Materials and methods/2.4 Graph attention network; 2 Materials and methods/2.4 Graph attention network/2.4.1 Graph attention mechanism; 2 Materials and methods/2.4 Graph attention network/2.4.2 Feature aggregation and update; 2 Materials and methods/2.4 Graph attention network/2.4.3 Multi-head attention for improved representation; 2 Materials and methods/2.5 Graph node representations; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"RLsite is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["rlsite-rna-binding-2025"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"A merged RNAglib/RNAsite training collection; RNAglib filtering excludes chains above 440 nt and those with fewer than four binding sites.","status":"source_checked","source_ids":["rlsite-rna-binding-2025"],"source_locator":"2 Materials and methods/2.1 Benchmark datasets (paragraph 1); 3 Results and discussions/3.4 Consistent high performance of RLsite on the latest dataset T10 (paragraph 2)"},{"label":"Context limits","value":"RNA chains longer than 440 nucleotides are excluded from the curated benchmark.","status":"source_checked","source_ids":["rlsite-rna-binding-2025"],"source_locator":"2 Materials and methods/2.1 Benchmark datasets (paragraph 1); 2 Materials and methods/2.1 Benchmark datasets (paragraph 2)"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/SaisaiSun/RLsite/blob/3d4a294aab295cbc1afd56b9ea6ced96e0717304/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-rlsite-rna-binding-2025-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","status":"unreported","source_ids":["evidence-reported-rlsite-rna-binding-2025-readme-md"],"source_locator":"README.md and repository-root licence-file search"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-rlsite-rna-binding-2025-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Combines sequence and structural evidence and evaluates an independent public test set.","source_ids":["rlsite-rna-binding-2025"],"source_locator":"2 Materials and methods/2.1 Benchmark datasets (paragraph 1); 2 Materials and methods/2.4 Graph attention network/2.4.3 Multi-head attention for improved representation (paragraph 1)"}],"limitations":[{"text":"The training filters restrict sequence lengths and binding-site density; sequence-only methods do not receive equivalent inputs.","source_ids":["rlsite-rna-binding-2025"],"source_locator":"3 Results and discussions/3.3 Complementarity of graph attention networks and RNA language models (paragraph 2); 3 Results and discussions/3.2 Secondary structure emerges as the dominant feature for RNA binding site prediction (paragraph 1)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["RNA sequence and graph-based RNA structural information","RLsite","Small-molecule-binding nucleotide predictions"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["rlsite-rna-binding-2025"],"source_locator":"4 Conclusions/4.1 Hybrid architecture leveraging complementary modalities (paragraph 1); 2 Materials and methods/2.2 Model architecture overview (paragraph 1)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","RLsite is the comparison-table label; that label does not specify an immutable weight revision.","No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-52eee4cc67ca26","kind":"model","name":"RNAfold","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["rna-transcriptomes"]},"source_ids":["bpfold-2025"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"RNAfold","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"RNAfold is the thermodynamic RNA-structure baseline in the BPfold comparison.","summary_source_ids":["bpfold-2025"],"summary_source_locator":"Abstract (paragraph 2); Methods/Datasets and evaluation (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"ViennaRNA predicts a minimum-free-energy secondary structure using its thermodynamic energy model; the paper runs default settings.","source_ids":["bpfold-2025"],"source_locator":"Abstract (paragraph 2); Methods/Base pair motif energy as thermodynamic prior (paragraph 1)"},{"title":"Underlying method and version boundaries","body":"The ViennaRNA package computes minimum-free-energy structures, partition functions and associated structure probabilities. RNAfold is a procedure with energy parameters, not a neural language-model checkpoint.","source_ids":["evidence-reported-base-rnafold-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies RNAfold: RNA secondary structure. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["bpfold-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-012"}],"facts":[{"label":"Model type","value":"Thermodynamic RNA folding algorithm; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-rnafold-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"ViennaRNA predicts a minimum-free-energy secondary structure using its thermodynamic energy model; the paper runs default settings.","status":"source_checked","source_ids":["bpfold-2025"],"source_locator":"Abstract (paragraph 2); Methods/Base pair motif energy as thermodynamic prior (paragraph 1)"},{"label":"Biological inputs","value":"RNA nucleotide sequence","status":"source_checked","source_ids":["bpfold-2025"],"source_locator":"Methods/Deep neural network with base pair attention (paragraph 5); Introduction (paragraph 1)"},{"label":"Outputs","value":"RNA secondary structure","status":"source_checked","source_ids":["bpfold-2025"],"source_locator":"Methods/Datasets and evaluation (paragraph 1); Discussion (paragraph 5)"},{"label":"Parameters","value":"Not applicable as a neural parameter count.","status":"inapplicable","source_ids":["bpfold-2025"],"source_locator":"Methods/Training strategy and structure refinement (paragraph 3); Results/Assessing the effectiveness of base pair motif energy (paragraph 3)"},{"label":"Known versions / configuration","value":"ViennaRNA RNAfold version 2.6.4","status":"unreported","source_ids":["bpfold-2025"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"Experimentally informed thermodynamic parameters rather than neural pretraining.","status":"source_checked","source_ids":["bpfold-2025"],"source_locator":"Methods/Training strategy and structure refinement (paragraph 3); Results/Evaluating BPfold on family-wise datasets (paragraph 3)"},{"label":"Context limits","value":"RNA length is constrained by the RNAfold implementation, algorithm and available memory; there is no learned fixed-token context window.","status":"inapplicable","source_ids":["bpfold-2025"],"source_locator":"Supplementary information (paragraph 1); Code availability (paragraph 1)"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://github.com/ViennaRNA/ViennaRNA/blob/1ffec79f5e258896160f7362ced8263450f371dc/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-rnafold-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"ViennaRNA licence/disclaimer; see the pinned full text for scope and conditions. (upstream repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-base-rnafold-license-txt"],"source_locator":"license.txt; complete licence text"},{"label":"Weights licence","value":"Not applicable: RNAfold uses thermodynamic energy parameters rather than pretrained neural weights.","status":"inapplicable","source_ids":["bpfold-2025"],"source_locator":"Methods/Training strategy and structure refinement (paragraph 3); Discussion (paragraph 2)"}],"strengths":[{"text":"Provides a non-neural physical prior against which learned folding models can be assessed.","source_ids":["bpfold-2025"],"source_locator":"Discussion (paragraph 4); Introduction (paragraph 4)"}],"limitations":[{"text":"The conventional nested-pair structure formulation does not model arbitrary pseudoknots or the complete in-vivo folding environment.","source_ids":["bpfold-2025"],"source_locator":"Introduction (paragraph 7); Results/Overview of the BPfold approach (paragraph 1)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["RNA nucleotide sequence","RNAfold","RNA secondary structure"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["bpfold-2025"],"source_locator":"Abstract (paragraph 2); Methods/Base pair motif energy as thermodynamic prior (paragraph 1)"},"coverage":"limited","gaps":["ViennaRNA RNAfold version 2.6.4","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-53d6515bcc1f39","kind":"model","name":"GREmLN","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["cells-tissues"]},"source_ids":["gremln-2026"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"GREmLN","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"GREmLN learns single-cell representations using a gene-interaction graph inside its attention mechanism.","summary_source_ids":["gremln-2026"],"summary_source_locator":"Abstract (paragraph 1); Validation Experiments/Graph Structure Understanding (paragraph 3)","sections":[{"title":"How the evaluated method works","body":"Graph signal processing supplies gene-token structure to attention, replacing an arbitrary sequence ordering with regulatory or interaction relationships.","source_ids":["gremln-2026"],"source_locator":"Model Architecture (paragraph 1); Abstract (paragraph 1)"},{"title":"What was evaluated","body":"The linked evaluation record identifies GREmLN: cell-type annotation. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["gremln-2026"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b4-014"}],"facts":[{"label":"Model type","value":"Graph-based predictive method; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["gremln-2026"],"source_locator":"Model Architecture (paragraph 1); Abstract (paragraph 1)"},{"label":"Architecture / procedure","value":"Graph signal processing supplies gene-token structure to attention, replacing an arbitrary sequence ordering with regulatory or interaction relationships.","status":"source_checked","source_ids":["gremln-2026"],"source_locator":"Model Architecture (paragraph 1); Abstract (paragraph 1)"},{"label":"Biological inputs","value":"Single-cell expression profiles and a molecular-interaction graph","status":"source_checked","source_ids":["gremln-2026"],"source_locator":"Validation Experiments/Transcriptomic Landscape Learning & Cell Type Annotation/Bayesian Graph Integration (paragraph 3); Model Architecture (paragraph 1)"},{"label":"Outputs","value":"Gene/cell embeddings and adapted cell-annotation or reverse-perturbation predictions","status":"source_checked","source_ids":["gremln-2026"],"source_locator":"Abstract (paragraph 1); Validation Experiments/Fine-tuning for Reverse Perturbation Prediction (paragraph 1)"},{"label":"Parameters","value":"The principal model is reported as 10.3M learnable parameters. A separate scaling study uses one-, three- and six-layer models reported as 7.4M, 10M and 24.2M; these are distinct configurations.","status":"source_checked","source_ids":["gremln-2026"],"source_locator":"Validation Experiments/Scaling Behavior (paragraph 1); Appendix/A Appendix/A.6 Computational Costs/Pretraining (paragraph 1)"},{"label":"Known versions / configuration","value":"GREmLN is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["gremln-2026"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"Masked expression-bin prediction on a CELLxGENE-derived corpus with cell-type-specific gene regulatory graphs. Appendix A.2 holds out 15% of cell types/graphs and leaves about 70% of cells for training; Appendix A.5 describes one pretraining epoch.","status":"source_checked","source_ids":["gremln-2026"],"source_locator":"Appendix/A Appendix/A.9 Details of Validation Datasets/Human Immune Cells (paragraph 1); Model Architecture/Training Objectives (paragraph 1)"},{"label":"Context limits","value":"The pinned implementation’s example tokenizer uses max_seq_length=4096. This is an explicit documented configuration, not evidence that every historical evaluation used the same limit.","status":"source_checked","source_ids":["evidence-reported-gremln-2026-readme-md"],"source_locator":"README.md; Custom Tokenization Example"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/czi-ai/GREmLN/blob/e1c5d8edbe2fe96568ff5451f15bc691722bb6a1/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-gremln-2026-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"MIT (study repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-gremln-2026-license-md"],"source_locator":"LICENSE.md; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-gremln-2026-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Makes graph relationships explicit in the representation of unordered gene-expression features.","source_ids":["gremln-2026"],"source_locator":"Validation Experiments/Graph Ablation (paragraph 1); Conclusion & Discussion (paragraph 1)"}],"limitations":[{"text":"The graph is an additional information source; comparisons with expression-only models must account for its provenance and coverage.","source_ids":["gremln-2026"],"source_locator":"Validation Experiments/Graph Ablation (paragraph 1); Appendix/A Appendix/A.1 Gene Regulatory Network Construction (paragraph 1)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Single-cell expression profiles and a molecular-interaction graph","GREmLN","Gene/cell embeddings and adapted cell-annotation or reverse-perturbation predictions"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["gremln-2026"],"source_locator":"Model Architecture (paragraph 1); Abstract (paragraph 1)"},"coverage":"limited","gaps":["GREmLN is the comparison-table label; that label does not specify an immutable weight revision.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-54be8a811c206e","kind":"model","name":"AK-score-ensemble","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["molecular-interactions"]},"source_ids":["akscore-2020"],"links":[],"attributes":{"entity_level":"method","version":"ensemble; learning rate 0.0007","reported_name":"AK-score-ensemble","historical_missing_metadata":{"checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"AK-score predicts protein–ligand binding affinity from a three-dimensional complex.","summary_source_ids":["akscore-2020"],"summary_source_locator":"3. Methods/3.5. Performance Assessment (paragraph 8); 3. Methods/3.2. Convolutional Neural Network (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"The binding pocket and ligand are voxelised on a 30 Å cube with 1 Å spacing. Multichannel 3D convolutional networks learn interaction patterns; the ensemble averages independently trained models, while the single configuration uses one network.","source_ids":["akscore-2020"],"source_locator":"3. Methods/3.2. Convolutional Neural Network (paragraph 1); 3. Methods/3.4. Ensemble Prediction (paragraph 1)"},{"title":"What was evaluated","body":"The linked evaluation record identifies AK-score-ensemble: Protein–ligand binding affinity scoring. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["akscore-2020"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b3-048"}],"facts":[{"label":"Model type","value":"Convolutional neural network; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["akscore-2020"],"source_locator":"3. Methods/3.2. Convolutional Neural Network (paragraph 1); 3. Methods/3.4. Ensemble Prediction (paragraph 1)"},{"label":"Architecture / procedure","value":"The binding pocket and ligand are voxelised on a 30 Å cube with 1 Å spacing. Multichannel 3D convolutional networks learn interaction patterns; the ensemble averages independently trained models, while the single configuration uses one network.","status":"source_checked","source_ids":["akscore-2020"],"source_locator":"3. Methods/3.2. Convolutional Neural Network (paragraph 1); 3. Methods/3.4. Ensemble Prediction (paragraph 1)"},{"label":"Biological inputs","value":"Protein–ligand complex coordinates represented as atomic-density grids","status":"source_checked","source_ids":["akscore-2020"],"source_locator":"3. Methods/3.2. Convolutional Neural Network (paragraph 1); 3. Methods/3.2. Convolutional Neural Network (paragraph 3)"},{"label":"Outputs","value":"Predicted binding affinity","status":"source_checked","source_ids":["akscore-2020"],"source_locator":"3. Methods/3.5. Performance Assessment (paragraph 8); 3. Methods/3.5. Performance Assessment (paragraph 10)"},{"label":"Parameters","value":"Each network has 1,294,925 parameters:1,293,447 trainable and 1,478 non-trainable. The ensemble uses separately trained networks; this is the per-network count.","status":"source_checked","source_ids":["akscore-2020"],"source_locator":"Methods3.3 Network Architecture; total-parameter paragraph"},{"label":"Known versions / configuration","value":"ensemble; learning rate 0.0007","status":"source_checked","source_ids":["akscore-2020"],"source_locator":"Table ijms-21-08424-t002 (paragraph 1); Table ijms-21-08424-t001 (paragraph 1)"},{"label":"Training data / fitting","value":"3,772 PDBbind-2016 refined-set complexes for training; the 285-complex core set is held out for testing.","status":"source_checked","source_ids":["akscore-2020"],"source_locator":"2. Results and Discussion/2.4. Assessment with an Additional Dataset (paragraph 1); 3. Methods/3.5. Performance Assessment (paragraph 14)"},{"label":"Context limits","value":"30 × 30 × 30 spatial grid at 1 Å spacing","status":"source_checked","source_ids":["akscore-2020"],"source_locator":"3. Methods/3.2. Convolutional Neural Network (paragraph 1); 3. Methods/3.2. Convolutional Neural Network (paragraph 4)"},{"label":"Access","value":"A public release of the original 2020 AK-score model was not established from the paper or targeted official-repository search. AK-Score2 is a separate later method and is not substituted.","status":"unreported","source_ids":["akscore-2020"],"source_locator":"Complete 2020 paper, including Methods and Supplementary Materials statement; exact-name repository discovery"},{"label":"Code licence","value":"No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","status":"unreported","source_ids":["akscore-2020"],"source_locator":"4. Conclusions (paragraph 1); 3. Methods/3.5. Performance Assessment (paragraph 14)"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["akscore-2020"],"source_locator":"2. Results and Discussion/2.5. Identifying Hot Spots for Binding Affinity Determination Using Grad-CAM (paragraph 1); 4. Conclusions (paragraph 1)"}],"strengths":[{"text":"The same framework supports affinity regression and relative affinity ranking within the evaluated structural benchmark.","source_ids":["akscore-2020"],"source_locator":"3. Methods/3.5. Performance Assessment (paragraph 11); 3. Methods/3.5. Performance Assessment (paragraph 3)"}],"limitations":[{"text":"Requires a complex structure and inherits the chemical and structural coverage of PDBbind; a favourable core-set result does not establish prospective docking accuracy.","source_ids":["akscore-2020"],"source_locator":"Abstract (paragraph 1); 2. Results and Discussion/2.5. Identifying Hot Spots for Binding Affinity Determination Using Grad-CAM (paragraph 2)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Protein–ligand complex coordinates represented as atomic-density grids","AK-score-ensemble","Predicted binding affinity"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["akscore-2020"],"source_locator":"3. Methods/3.2. Convolutional Neural Network (paragraph 1); 3. Methods/3.4. Ensemble Prediction (paragraph 1)"},"coverage":"limited","gaps":["A public release of the original 2020 AK-score model was not established from the paper or targeted official-repository search. AK-Score2 is a separate later method and is not substituted.","No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-54d974e8e08043","kind":"model","name":"mRNA-LM","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["rna-transcriptomes"]},"source_ids":["mrna-lm-2025"],"links":[],"attributes":{"entity_level":"method","version":"not stated in table","reported_name":"mRNA-LM","historical_missing_metadata":{"checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"mRNA-LM represents complete mRNAs by integrating separate models of the 5′ UTR, coding sequence and 3′ UTR.","summary_source_ids":["mrna-lm-2025"],"summary_source_locator":"Results/Evaluation of UTRBERTs (paragraph 1); Materials and methods/Datasets for pretraining UTRBERTs (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"Two nucleotide-token UTRBERTs and a codon-token CodonBERT component are integrated using contrastive learning inspired by CLIP. Each UTR encoder has 12 layers, 12 heads and 768-dimensional hidden states.","source_ids":["mrna-lm-2025"],"source_locator":"Materials and methods/Datasets for pretraining UTRBERTs/Pretraining our models (paragraph 1); Materials and methods/mRNA-LM: joint language model/Learning the joint representation of segment sequence using CLIP (paragraph 5)"},{"title":"What was evaluated","body":"The linked evaluation record identifies mRNA-LM: mRNA half-life prediction. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["mrna-lm-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-b2-mrna-lm-2025"}],"facts":[{"label":"Model type","value":"Learned representation pipeline; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["mrna-lm-2025"],"source_locator":"Materials and methods/Datasets for pretraining UTRBERTs/Pretraining our models (paragraph 1); Materials and methods/mRNA-LM: joint language model/Learning the joint representation of segment sequence using CLIP (paragraph 5)"},{"label":"Architecture / procedure","value":"Two nucleotide-token UTRBERTs and a codon-token CodonBERT component are integrated using contrastive learning inspired by CLIP. Each UTR encoder has 12 layers, 12 heads and 768-dimensional hidden states.","status":"source_checked","source_ids":["mrna-lm-2025"],"source_locator":"Materials and methods/Datasets for pretraining UTRBERTs/Pretraining our models (paragraph 1); Materials and methods/mRNA-LM: joint language model/Learning the joint representation of segment sequence using CLIP (paragraph 5)"},{"label":"Biological inputs","value":"The 5′ UTR, CDS and 3′ UTR of an mRNA","status":"source_checked","source_ids":["mrna-lm-2025"],"source_locator":"Conclusion (paragraph 14); Results/Evaluation of UTRBERTs (paragraph 1)"},{"label":"Outputs","value":"Integrated mRNA representations and task-specific property predictions","status":"source_checked","source_ids":["mrna-lm-2025"],"source_locator":"Abstract (paragraph 1); Conclusion (paragraph 14)"},{"label":"Parameters","value":"The integrated mRNA-LM model is reported to contain more than 260 million parameters; its UTR modules each use 12 layers, 12 heads and hidden width 768.","status":"source_checked","source_ids":["mrna-lm-2025"],"source_locator":"Materials and methods/Fine-tuning using LoRA (paragraph 1); Materials and methods/Datasets for pretraining UTRBERTs/Pretraining our models (paragraph 1)"},{"label":"Known versions / configuration","value":"mRNA-LM is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["mrna-lm-2025"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"Millions of mRNA sequences from multiple species; UTR encoders use masked-language pretraining.","status":"source_checked","source_ids":["mrna-lm-2025"],"source_locator":"Materials and methods/Datasets for pretraining UTRBERTs/Pretraining our models (paragraph 1); Materials and methods/Datasets for pretraining UTRBERTs (paragraph 1)"},{"label":"Context limits","value":"5UTRBERT maximum length 512 tokens; 3UTRBERT maximum length 1,024 tokens. CDS uses a separate codon model.","status":"source_checked","source_ids":["mrna-lm-2025"],"source_locator":"Materials and methods/Datasets for pretraining UTRBERTs/Pretraining our models (paragraph 1); Conclusion (paragraph 9)"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/Sanofi-Public/mRNA-LM/blob/d7538c9aadbceb59a8832904292b279d0a4c2d12/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-mrna-lm-2025-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"Sanofi academic/non-commercial licence; use outside those permissions requires the separate terms described in the file. (study repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-mrna-lm-2025-license-txt"],"source_locator":"LICENSE.txt; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-mrna-lm-2025-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Models interactions among mRNA regions rather than using only a coding-sequence representation.","source_ids":["mrna-lm-2025"],"source_locator":"Introduction (paragraph 5); Results/Evaluation of UTRBERTs (paragraph 1)"}],"limitations":[{"text":"Component-specific tokenisation and context limits matter; property heads must be evaluated on their own assay and split.","source_ids":["mrna-lm-2025"],"source_locator":"Materials and methods/Datasets for pretraining UTRBERTs/Pretraining our models (paragraph 1); Abstract (paragraph 1)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["The 5′ UTR, CDS and 3′ UTR of an mRNA","mRNA-LM","Integrated mRNA representations and task-specific property predictions"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["mrna-lm-2025"],"source_locator":"Materials and methods/Datasets for pretraining UTRBERTs/Pretraining our models (paragraph 1); Materials and methods/mRNA-LM: joint language model/Learning the joint representation of segment sequence using CLIP (paragraph 5)"},"coverage":"limited","gaps":["mRNA-LM is the comparison-table label; that label does not specify an immutable weight revision.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-57dbab30462150","kind":"model","name":"CLAPE-SMB with ESM-2","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["proteins-complexes"]},"source_ids":["clape-smb-2024"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"CLAPE-SMB with ESM-2","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"CLAPE-SMB identifies small-molecule-binding residues from protein sequences using frozen ESM-2 features and contrastive learning.","summary_source_ids":["clape-smb-2024"],"summary_source_locator":"Conclusion (paragraph 1); Results/Comparison of CLAPE-SMB with DeepProSite and heuristic analyses in three protein case studies (paragraph 3)","sections":[{"title":"How the evaluated method works","body":"The frozen esm2_t33_650M_UR50D encoder emits 1,280-dimensional residue embeddings, which feed the supervised binding-site predictor and contrastive-learning objective.","source_ids":["clape-smb-2024"],"source_locator":"Methods/Sequence embedding (paragraph 2); Methods/Loss function (paragraph 5)"},{"title":"Underlying method and version boundaries","body":"ESM-2 is a transformer protein language-model family. The official repository exposes residue embeddings, sequence-level pooling and models at several sizes; the study configuration determines which of these is evaluated.","source_ids":["evidence-reported-base-esm-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies CLAPE-SMB with ESM-2: protein-small molecule binding-site prediction. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["clape-smb-2024"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b4-011"}],"facts":[{"label":"Model type","value":"Protein sequence transformer; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-esm-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"The frozen esm2_t33_650M_UR50D encoder emits 1,280-dimensional residue embeddings, which feed the supervised binding-site predictor and contrastive-learning objective.","status":"source_checked","source_ids":["clape-smb-2024"],"source_locator":"Methods/Sequence embedding (paragraph 2); Methods/Loss function (paragraph 5)"},{"label":"Biological inputs","value":"Protein sequences using 20 standard amino-acid tokens and X for unknown residues","status":"source_checked","source_ids":["clape-smb-2024"],"source_locator":"Methods/Sequence embedding (paragraph 2); Conclusion (paragraph 1)"},{"label":"Outputs","value":"Per-residue small-molecule-binding predictions","status":"source_checked","source_ids":["clape-smb-2024"],"source_locator":"Results/CLAPE-SMB can rationally distinguish binding and non-binding sites based on residue features (paragraph 1); Results/Necessity of merging multiple binding sites of similar proteins (paragraph 2)"},{"label":"Parameters","value":"650-million-parameter, 33-layer ESM-2 feature extractor; total pipeline size is not reported here.","status":"source_checked","source_ids":["clape-smb-2024"],"source_locator":"Table Tab5 (paragraph 1); Methods/Sequence embedding (paragraph 2)"},{"label":"Known versions / configuration","value":"CLAPE-SMB with ESM-2 is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["clape-smb-2024"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"SJC, UniProtSMB and intrinsically disordered protein datasets described in the paper; encoder layers are not fine-tuned.","status":"source_checked","source_ids":["clape-smb-2024"],"source_locator":"Results/UniProtSMB dataset preparation (paragraph 3); Results/UniProtSMB dataset preparation (paragraph 1)"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["clape-smb-2024","evidence-reported-base-esm-readme-md"],"source_locator":"Methods/Sequence embedding; Methods/Backbone model; Methods/Loss function; Methods/Evaluation metrics; Results/The model architecture of CLAPE-SMB; Results/Influence of model architecture on prediction accuracy; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://github.com/facebookresearch/esm/blob/2b369911bb5b4b0dda914521b9475cad1656b2ac/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-esm-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"MIT (upstream repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-base-esm-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-base-esm-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Predicts binding sites without requiring an experimentally determined protein structure.","source_ids":["clape-smb-2024"],"source_locator":"Results/Comparison of CLAPE-SMB with DeepProSite and heuristic analyses in three protein case studies (paragraph 3); Introduction (paragraph 2)"}],"limitations":[{"text":"The ESM-2 representation and the supervised binding-site head are jointly part of the evaluated pipeline; reported scores are not raw ESM-2 capabilities.","source_ids":["clape-smb-2024"],"source_locator":"Results/Computational efficiency of CLAPE-SMB (paragraph 1); Results/Necessity of merging multiple binding sites of similar proteins (paragraph 2)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Protein sequences using 20 standard amino-acid tokens and X for unknown residues","CLAPE-SMB with ESM-2","Per-residue small-molecule-binding predictions"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["clape-smb-2024"],"source_locator":"Methods/Sequence embedding (paragraph 2); Methods/Loss function (paragraph 5)"},"coverage":"limited","gaps":["CLAPE-SMB with ESM-2 is the comparison-table label; that label does not specify an immutable weight revision.","A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-5b70fccb70bb70","kind":"model","name":"Eco70PromBERT","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["microbes-communities"]},"source_ids":["cyaprombert-2022"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"Eco70PromBERT","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Eco70PromBERT is the E. coli promoter-prediction transformer configuration evaluated in the CyaPromBERT study.","summary_source_ids":["cyaprombert-2022"],"summary_source_locator":"Materials and methods/Datasets (paragraph 2); Materials and methods/Model training (paragraph 2)","sections":[{"title":"How the evaluated method works","body":"The Eco70PromBERT-1 bp configuration uses BERT-base with a single-base tokenizer for E. coli σ70 promoter classification.","source_ids":["cyaprombert-2022"],"source_locator":"Results and discussion/Evaluating model performance compared to existing promoter prediction models using independent datasets from E. coli\n (paragraph 1); Results and discussion/Evaluating model performance compared to existing promoter prediction models using independent datasets from E. coli\n (paragraph 3)"},{"title":"What was evaluated","body":"The linked evaluation record identifies Eco70PromBERT: E. coli sigma70 promoter prediction. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["cyaprombert-2022"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-035"}],"facts":[{"label":"Model type","value":"Study-specific predictive method; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["cyaprombert-2022"],"source_locator":"Results and discussion/Evaluating model performance compared to existing promoter prediction models using independent datasets from E. coli\n (paragraph 1); Results and discussion/Evaluating model performance compared to existing promoter prediction models using independent datasets from E. coli\n (paragraph 3)"},{"label":"Architecture / procedure","value":"The Eco70PromBERT-1 bp configuration uses BERT-base with a single-base tokenizer for E. coli σ70 promoter classification.","status":"source_checked","source_ids":["cyaprombert-2022"],"source_locator":"Results and discussion/Evaluating model performance compared to existing promoter prediction models using independent datasets from E. coli\n (paragraph 1); Results and discussion/Evaluating model performance compared to existing promoter prediction models using independent datasets from E. coli\n (paragraph 3)"},{"label":"Biological inputs","value":"DNA promoter and non-promoter sequence windows","status":"source_checked","source_ids":["cyaprombert-2022"],"source_locator":"Materials and methods/Datasets (paragraph 2); Results and discussion/Evaluating model performance compared to existing promoter prediction models using independent datasets from E. coli\n (paragraph 1)"},{"label":"Outputs","value":"Promoter classification","status":"source_checked","source_ids":["cyaprombert-2022"],"source_locator":"Results and discussion/Evaluating model performance compared to existing promoter prediction models using independent datasets from E. coli\n (paragraph 3); Materials and methods/Datasets (paragraph 2)"},{"label":"Parameters","value":"86.8 million trainable parameters in the BERT-base configuration described by the study.","status":"source_checked","source_ids":["cyaprombert-2022"],"source_locator":"Results and discussion/Interpreting the model’s behavior through Monte Carlo sampling and attention score visualization (paragraph 1); Results and discussion/Evaluating model performance compared to existing promoter prediction models using independent datasets from E. coli\n (paragraph 1)"},{"label":"Known versions / configuration","value":"Eco70PromBERT is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["cyaprombert-2022"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"The paper’s E. coli σ70 promoter comparison; the cyanobacterial data and models are a separate experimental setting.","status":"source_checked","source_ids":["cyaprombert-2022"],"source_locator":"Materials and methods/Datasets (paragraph 2); Results and discussion/Evaluating model performance compared to existing promoter prediction models using independent datasets from E. coli\n (paragraph 1)"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["cyaprombert-2022","evidence-reported-cyaprombert-2022-readme-md"],"source_locator":"Materials and methods/Datasets; Materials and methods/Constructing promoter extracting module from dRNA-seq datasets; Materials and methods/Promoter and non-promoter sequences extraction; Materials and methods/Model training; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/hanepira/TSSnote-CyaPromBert/blob/e86f5449e2e2af3fead1b418ba721f38feb61318/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-cyaprombert-2022-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","status":"unreported","source_ids":["evidence-reported-cyaprombert-2022-readme-md"],"source_locator":"README.md and repository-root licence-file search"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-cyaprombert-2022-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Provides a learned-sequence comparator within an explicit promoter classification test.","source_ids":["cyaprombert-2022"],"source_locator":"Results and discussion/Evaluating model performance compared to existing promoter prediction models using independent datasets from E. coli\n (paragraph 3); Results and discussion/Evaluating model performance compared to existing promoter prediction models using independent datasets from E. coli\n (paragraph 2)"}],"limitations":[{"text":"The E. coli result cannot be substituted for the cross-species cyanobacterial results in the same paper.","source_ids":["cyaprombert-2022"],"source_locator":"Materials and methods/Model training (paragraph 7); Introduction (paragraph 3)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["DNA promoter and non-promoter sequence windows","Eco70PromBERT","Promoter classification"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["cyaprombert-2022"],"source_locator":"Results and discussion/Evaluating model performance compared to existing promoter prediction models using independent datasets from E. coli\n (paragraph 1); Results and discussion/Evaluating model performance compared to existing promoter prediction models using independent datasets from E. coli\n (paragraph 3)"},"coverage":"limited","gaps":["Eco70PromBERT is the comparison-table label; that label does not specify an immutable weight revision.","A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-60455ff7cc0c15","kind":"model","name":"scRegNet (Geneformer backbone)","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["cells-tissues"]},"source_ids":["scregnet-2025"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"scRegNet (Geneformer backbone)","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"scRegNet predicts gene-regulatory links using single-cell foundation-model features and graph learning.","summary_source_ids":["scregnet-2025"],"summary_source_locator":"Conclusions and Discussions (paragraph 1); Results/Performance on benchmark datasets (paragraph 2)","sections":[{"title":"How the evaluated method works","body":"A specified pretrained backbone supplies context-aware gene representations; a graph-based supervised predictor learns regulatory connections from known links. Geneformer and scBERT are separate backbone configurations.","source_ids":["scregnet-2025"],"source_locator":"Method/Graph-based learning with GNNs (paragraph 1); Method/Gene representations from foundation models (paragraph 1)"},{"title":"Underlying method and version boundaries","body":"Geneformer ranks genes by expression scaled against its pretraining corpus, then uses a transformer encoder with a masked-gene objective. V1 and V2 have different corpora, vocabularies, sizes and context limits, so a historical paper name is not replaced with today’s default checkpoint.","source_ids":["evidence-reported-base-geneformer-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies scRegNet (Geneformer backbone): Gene-regulatory link prediction. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["scregnet-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-031"}],"facts":[{"label":"Model type","value":"Single-cell transformer; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-geneformer-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"A specified pretrained backbone supplies context-aware gene representations; a graph-based supervised predictor learns regulatory connections from known links. Geneformer and scBERT are separate backbone configurations.","status":"source_checked","source_ids":["scregnet-2025"],"source_locator":"Method/Graph-based learning with GNNs (paragraph 1); Method/Gene representations from foundation models (paragraph 1)"},{"label":"Biological inputs","value":"Single-cell expression data, pretrained gene embeddings and known regulatory links","status":"source_checked","source_ids":["scregnet-2025"],"source_locator":"Method/Gene representations from foundation models/Geneformer (paragraph 2); Method/Graph-based learning with GNNs (paragraph 1)"},{"label":"Outputs","value":"Predicted gene-regulatory connections","status":"source_checked","source_ids":["scregnet-2025"],"source_locator":"Method/Model training (paragraph 1); Method/Link prediction layer (paragraph 2)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["scregnet-2025","evidence-reported-base-geneformer-readme-md"],"source_locator":"Method; Method/Gene representations from foundation models; Method/Gene representations from foundation models/scBERT; Method/Gene representations from foundation models/scFoundation; Method/Gene representations from foundation models/Geneformer; Method/Gene representations from foundation models/Mean pooling; Method/Graph-based learning with GNNs; Method/Unified gene representations; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"scRegNet (Geneformer backbone) is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["scregnet-2025"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"Backbone pretraining is separate from supervised regulatory-link fitting; the study uses known TF–DNA binding information for the latter.","status":"source_checked","source_ids":["scregnet-2025"],"source_locator":"Experimental Setup/Datasets and data pre-processing (paragraph 2); Experimental Setup/Datasets and data pre-processing (paragraph 4)"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["scregnet-2025","evidence-reported-base-geneformer-readme-md"],"source_locator":"Method; Method/Gene representations from foundation models; Method/Gene representations from foundation models/scBERT; Method/Gene representations from foundation models/scFoundation; Method/Gene representations from foundation models/Geneformer; Method/Gene representations from foundation models/Mean pooling; Method/Graph-based learning with GNNs; Method/Unified gene representations; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://huggingface.co/ctheodoris/Geneformer/blob/1f7fbae4e469a5f4f1af8c111a529cfe1b3829f5/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-geneformer-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","status":"unreported","source_ids":["evidence-reported-base-geneformer-readme-md"],"source_locator":"README.md and repository-root licence-file search"},{"label":"Weights licence","value":"Apache 2.0 is declared in the official Geneformer model-card metadata; the exact historical configuration still needs its checkpoint identity.","status":"source_checked","source_ids":["evidence-reported-base-geneformer-readme-md"],"source_locator":"README.md front matter, license field; model-version list"}],"strengths":[{"text":"Tests whether pretrained gene representations help when experimentally supported regulatory links are limited.","source_ids":["scregnet-2025"],"source_locator":"Method/Graph-based learning with GNNs (paragraph 1); Introduction (paragraph 3)"}],"limitations":[{"text":"Supervised link prediction on known binding data does not establish causality or intervention effects for all inferred edges.","source_ids":["scregnet-2025"],"source_locator":"Abstract (paragraph 1); Method/Graph-based learning with GNNs (paragraph 1)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Single-cell expression data, pretrained gene embeddings and known regulatory links","scRegNet (Geneformer backbone)","Predicted gene-regulatory connections"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["scregnet-2025"],"source_locator":"Method/Graph-based learning with GNNs (paragraph 1); Method/Gene representations from foundation models (paragraph 1)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","scRegNet (Geneformer backbone) is the comparison-table label; that label does not specify an immutable weight revision.","A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-62bc5e5ba13e7d","kind":"model","name":"Kraken 2","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["microbes-communities"]},"source_ids":["lemur-magnet-2024"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"Kraken 2","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Kraken 2 is the reference-database taxonomic classifier used in this metagenomic comparison.","summary_source_ids":["lemur-magnet-2024"],"summary_source_locator":"Discussion (paragraph 4); Methods/Method Comparison (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"Sequence evidence is matched against a taxonomic reference index to assign reads or contigs. Database construction and the chosen confidence/settings remain part of the evaluated configuration.","source_ids":["lemur-magnet-2024"],"source_locator":"Methods/Competitive read alignment with Magnet (paragraph 1); Results/Method overview (paragraph 1)"},{"title":"Underlying method and version boundaries","body":"Kraken 2 classifies sequences against a taxonomic reference index. The software version and the database used to construct that index are distinct reproducibility requirements.","source_ids":["evidence-reported-base-kraken2-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies Kraken 2: Long-read taxonomic profiling. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["lemur-magnet-2024"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b3-032"}],"facts":[{"label":"Model type","value":"Reference-index taxonomic classifier; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-kraken2-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"Sequence evidence is matched against a taxonomic reference index to assign reads or contigs. Database construction and the chosen confidence/settings remain part of the evaluated configuration.","status":"source_checked","source_ids":["lemur-magnet-2024"],"source_locator":"Methods/Competitive read alignment with Magnet (paragraph 1); Results/Method overview (paragraph 1)"},{"label":"Biological inputs","value":"Metagenomic reads or contigs and a taxonomic reference database","status":"source_checked","source_ids":["lemur-magnet-2024"],"source_locator":"Discussion (paragraph 4); Methods/Lemur (paragraph 8)"},{"label":"Outputs","value":"Taxonomic assignments","status":"source_checked","source_ids":["lemur-magnet-2024"],"source_locator":"Methods/Lemur (paragraph 2); Methods/Lemur (paragraph 8)"},{"label":"Parameters","value":"Not applicable as a neural model size.","status":"inapplicable","source_ids":["lemur-magnet-2024"],"source_locator":"Methods/Synthetic and simulated datasets/Simulated data from (Dilthey et al. 2019) (paragraph 1); Methods/Synthetic and simulated datasets/Zymo Fecal Reference with TruMatrix (paragraph 1)"},{"label":"Known versions / configuration","value":"Kraken 2 is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["lemur-magnet-2024"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"Reference-index construction rather than foundation-model pretraining.","status":"source_checked","source_ids":["lemur-magnet-2024"],"source_locator":"Methods/Synthetic and simulated datasets/Zymo Fecal Reference with TruMatrix (paragraph 1); Methods/Synthetic and simulated datasets/Zymo Fecal Reference with TruMatrix (paragraph 2)"},{"label":"Context limits","value":"Not applicable to a pretrained sequence-token window; read/contig lengths and the reference-database or comparison configuration determine the analysed input.","status":"inapplicable","source_ids":["lemur-magnet-2024"],"source_locator":"Methods/Competitive read alignment with Magnet (paragraph 1); Discussion (paragraph 2)"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://github.com/DerrickWood/kraken2/blob/8c190b1b668825935dbf6dee5f969227dc8269bb/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-kraken2-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"MIT (upstream repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-base-kraken2-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"Not applicable: uses a reference index rather than a pretrained neural checkpoint.","status":"inapplicable","source_ids":["lemur-magnet-2024"],"source_locator":"Methods/Competitive read alignment with Magnet (paragraph 1); Results/ZymoBIOMICS Fecal Reference (paragraph 1)"}],"strengths":[{"text":"Provides an established reference-based comparator for learned or alignment-based taxonomic methods.","source_ids":["lemur-magnet-2024"],"source_locator":"Methods/Lemur (paragraph 8); Methods/Competitive read alignment with Magnet (paragraph 1)"}],"limitations":[{"text":"Unrepresented organisms and reference contamination can alter assignments; results are specific to the database and query distribution.","source_ids":["lemur-magnet-2024"],"source_locator":"Discussion (paragraph 4); Results/Prior simulated data (paragraph 2)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Metagenomic reads or contigs and a taxonomic reference database","Kraken 2","Taxonomic assignments"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["lemur-magnet-2024"],"source_locator":"Methods/Competitive read alignment with Magnet (paragraph 1); Results/Method overview (paragraph 1)"},"coverage":"limited","gaps":["Kraken 2 is the comparison-table label; that label does not specify an immutable weight revision.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-65059c3a806306","kind":"model","name":"PlantCAD2","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["dna-genomes"]},"source_ids":["plantcad2-2025"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"PlantCAD2","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"PlantCAD2 is a plant-specific, single-nucleotide genomic language model.","summary_source_ids":["plantcad2-2025"],"summary_source_locator":"Abstract (paragraph 1); Methods/Accessible chromatin region prediction (paragraph 3)","sections":[{"title":"How the evaluated method works","body":"A Caduceus-style bidirectional, reverse-complement-equivariant architecture uses Mamba2 blocks and masked-language pretraining.","source_ids":["plantcad2-2025"],"source_locator":"Methods/Zero-shot evaluation of Evo2 model (paragraph 2); Results/PlantCAD2: a long-context DNA language model for angiosperms (paragraph 1)"},{"title":"What was evaluated","body":"The linked evaluation record identifies PlantCAD2: cross-species conservation prediction. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["plantcad2-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b4-002"}],"facts":[{"label":"Model type","value":"Study-specific predictive method; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["plantcad2-2025"],"source_locator":"Methods/Zero-shot evaluation of Evo2 model (paragraph 2); Results/PlantCAD2: a long-context DNA language model for angiosperms (paragraph 1)"},{"label":"Architecture / procedure","value":"A Caduceus-style bidirectional, reverse-complement-equivariant architecture uses Mamba2 blocks and masked-language pretraining.","status":"source_checked","source_ids":["plantcad2-2025"],"source_locator":"Methods/Zero-shot evaluation of Evo2 model (paragraph 2); Results/PlantCAD2: a long-context DNA language model for angiosperms (paragraph 1)"},{"label":"Biological inputs","value":"Plant genomic DNA at single-nucleotide resolution","status":"source_checked","source_ids":["plantcad2-2025"],"source_locator":"Methods/Repeat annotation and loss re-weighting (paragraph 2); Introduction (paragraph 4)"},{"label":"Outputs","value":"Sequence representations, zero-shot scores and adapted regulatory predictions","status":"source_checked","source_ids":["plantcad2-2025"],"source_locator":"Results/Transcription factor binding sites revealed by high-confidence predictions of PlantCAD2 (paragraph 1); Methods/Evolutionary constraint prediction using the zero-shot strategy (paragraph 3)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["plantcad2-2025","evidence-reported-plantcad2-2025-readme-md"],"source_locator":"Methods/Preparing pre-training genomes; Methods/PlantCAD2 model architecture and pre-training; Methods/Repeat annotation and loss re-weighting; Methods/Evolutionary constraint prediction using the zero-shot strategy; Methods/Core and non-core gene classification using the zero-shot strategy; Methods/Accessible chromatin region prediction; Methods/Gene expression prediction in leaf; Methods/Leaf protein abundance prediction task; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"PlantCAD2 is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["plantcad2-2025"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"65 angiosperm genomes","status":"source_checked","source_ids":["plantcad2-2025"],"source_locator":"Methods/Preparing pre-training genomes (paragraph 1); Methods/PlantCAD2 model architecture and pre-training (paragraph 3)"},{"label":"Context limits","value":"8,192 base pairs","status":"source_checked","source_ids":["plantcad2-2025"],"source_locator":"Methods/PlantCAD2 model architecture and pre-training (paragraph 3); Introduction (paragraph 4)"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/plantcad/plantcad/blob/7240f0238f869b3ac25e4b5ad0996fad96ede9db/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-plantcad2-2025-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"Apache 2.0 (study repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-plantcad2-2025-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-plantcad2-2025-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Combines plant-specific phylogenetic coverage with a longer context and strand-aware architecture.","source_ids":["plantcad2-2025"],"source_locator":"Discussions (paragraph 1); Results/Fine-tuning PlantCAD2 accurately predicts cross-species chromatin accessible regions and cell-type-specific accessible regions (paragraph 2)"}],"limitations":[{"text":"Evidence concerns the paper’s plant benchmarks and adaptations; superiority in selected tasks does not imply superiority across all genomes or biological capabilities.","source_ids":["plantcad2-2025"],"source_locator":"Introduction (paragraph 6); Abstract (paragraph 1)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Plant genomic DNA at single-nucleotide resolution","PlantCAD2","Sequence representations, zero-shot scores and adapted regulatory predictions"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["plantcad2-2025"],"source_locator":"Methods/Zero-shot evaluation of Evo2 model (paragraph 2); Results/PlantCAD2: a long-context DNA language model for angiosperms (paragraph 1)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","PlantCAD2 is the comparison-table label; that label does not specify an immutable weight revision.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity.","The abstract/Methods and Results report conflicting large-model parameter counts (676M versus 694M)."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-673b8f46361000","kind":"model","name":"Kraken2","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["microbes-communities"]},"source_ids":["lazypipe-2020"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"Kraken2","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Kraken 2 is the reference-database taxonomic classifier used in this metagenomic comparison.","summary_source_ids":["lazypipe-2020"],"summary_source_locator":"2. Materials and methods/2.3 Benchmarking performance (paragraph 4); 4. Discussion (paragraph 3)","sections":[{"title":"How the evaluated method works","body":"Sequence evidence is matched against a taxonomic reference index to assign reads or contigs. Database construction and the chosen confidence/settings remain part of the evaluated configuration.","source_ids":["lazypipe-2020"],"source_locator":"2. Materials and methods/2.2 Unix pipeline for assembly, taxonomic profiling and binning of NGS data (paragraph 6); 2. Materials and methods/2.3 Benchmarking performance (paragraph 4)"},{"title":"Underlying method and version boundaries","body":"Kraken 2 classifies sequences against a taxonomic reference index. The software version and the database used to construct that index are distinct reproducibility requirements.","source_ids":["evidence-reported-base-kraken2-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies Kraken2: Simulated metagenome virus-taxon retrieval. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["lazypipe-2020"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b3-022"}],"facts":[{"label":"Model type","value":"Reference-index taxonomic classifier; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-kraken2-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"Sequence evidence is matched against a taxonomic reference index to assign reads or contigs. Database construction and the chosen confidence/settings remain part of the evaluated configuration.","status":"source_checked","source_ids":["lazypipe-2020"],"source_locator":"2. Materials and methods/2.2 Unix pipeline for assembly, taxonomic profiling and binning of NGS data (paragraph 6); 2. Materials and methods/2.3 Benchmarking performance (paragraph 4)"},{"label":"Biological inputs","value":"Metagenomic reads or contigs and a taxonomic reference database","status":"source_checked","source_ids":["lazypipe-2020"],"source_locator":"1. Introduction (paragraph 6); 2. Materials and methods/2.2 Unix pipeline for assembly, taxonomic profiling and binning of NGS data (paragraph 4)"},{"label":"Outputs","value":"Taxonomic assignments","status":"source_checked","source_ids":["lazypipe-2020"],"source_locator":"1. Introduction (paragraph 6); 2. Materials and methods/2.3 Benchmarking performance (paragraph 2)"},{"label":"Parameters","value":"Not applicable as a neural model size.","status":"inapplicable","source_ids":["lazypipe-2020"],"source_locator":"4. Discussion (paragraph 2); Table veaa091-T5 (paragraph 1)"},{"label":"Known versions / configuration","value":"Kraken2 is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["lazypipe-2020"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"Reference-index construction rather than foundation-model pretraining.","status":"source_checked","source_ids":["lazypipe-2020"],"source_locator":"3. Results/3.1 Excellent recall and precision for both simulated and real datasets (paragraph 3); 3. Results/3.1 Excellent recall and precision for both simulated and real datasets (paragraph 4)"},{"label":"Context limits","value":"Not applicable to a pretrained sequence-token window; read/contig lengths and the reference-database or comparison configuration determine the analysed input.","status":"inapplicable","source_ids":["lazypipe-2020"],"source_locator":"1. Introduction (paragraph 6); 2. Materials and methods/2.2 Unix pipeline for assembly, taxonomic profiling and binning of NGS data (paragraph 7)"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://github.com/DerrickWood/kraken2/blob/8c190b1b668825935dbf6dee5f969227dc8269bb/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-kraken2-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"MIT (upstream repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-base-kraken2-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"Not applicable: uses a reference index rather than a pretrained neural checkpoint.","status":"inapplicable","source_ids":["lazypipe-2020"],"source_locator":"4. Discussion (paragraph 2); 2. Materials and methods/2.3 Benchmarking performance (paragraph 1)"}],"strengths":[{"text":"Provides an established reference-based comparator for learned or alignment-based taxonomic methods.","source_ids":["lazypipe-2020"],"source_locator":"2. Materials and methods/2.2 Unix pipeline for assembly, taxonomic profiling and binning of NGS data (paragraph 4); 2. Materials and methods/2.3 Benchmarking performance (paragraph 1)"}],"limitations":[{"text":"Unrepresented organisms and reference contamination can alter assignments; results are specific to the database and query distribution.","source_ids":["lazypipe-2020"],"source_locator":"1. Introduction (paragraph 6); 2. Materials and methods/2.2 Unix pipeline for assembly, taxonomic profiling and binning of NGS data (paragraph 7)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Metagenomic reads or contigs and a taxonomic reference database","Kraken2","Taxonomic assignments"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["lazypipe-2020"],"source_locator":"2. Materials and methods/2.2 Unix pipeline for assembly, taxonomic profiling and binning of NGS data (paragraph 6); 2. Materials and methods/2.3 Benchmarking performance (paragraph 4)"},"coverage":"limited","gaps":["Kraken2 is the comparison-table label; that label does not specify an immutable weight revision.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-67ea6bd77b2ed1","kind":"model","name":"ESM2_AMPS","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["molecular-interactions"]},"source_ids":["esm2-amp-2025"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"ESM2_AMPS","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"ESM2_AMPS predicts binary protein–protein interactions by combining segment-level ESM-2 features.","summary_source_ids":["esm2-amp-2025"],"summary_source_locator":"Methods/Construction of dual models based on feature integration (paragraph 3); Results/Attention mechanism reveal protein–protein interaction mechanism (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"ESM-2 produces representations of sequence segments; a transformer fuses the segment information for interaction prediction. AMPS is the segment-feature variant and is distinct from the special-token-combination variant.","source_ids":["esm2-amp-2025"],"source_locator":"Methods/Interpretability analysis utilizing attention mechanisms (paragraph 1); Methods/Construction of dual models based on feature integration (paragraph 3)"},{"title":"Underlying method and version boundaries","body":"ESM-2 is a transformer protein language-model family. The official repository exposes residue embeddings, sequence-level pooling and models at several sizes; the study configuration determines which of these is evaluated.","source_ids":["evidence-reported-base-esm-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies ESM2_AMPS: protein-protein interaction prediction. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["esm2-amp-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b4-024"}],"facts":[{"label":"Model type","value":"Protein sequence transformer; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-esm-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"ESM-2 produces representations of sequence segments; a transformer fuses the segment information for interaction prediction. AMPS is the segment-feature variant and is distinct from the special-token-combination variant.","status":"source_checked","source_ids":["esm2-amp-2025"],"source_locator":"Methods/Interpretability analysis utilizing attention mechanisms (paragraph 1); Methods/Construction of dual models based on feature integration (paragraph 3)"},{"label":"Biological inputs","value":"Pairs of protein sequences","status":"source_checked","source_ids":["esm2-amp-2025"],"source_locator":"Results/Attention mechanism reveal protein–protein interaction mechanism (paragraph 1); Results/Benchmarking evaluation for ESM2_AMP on PPIs prediction (paragraph 1)"},{"label":"Outputs","value":"Binary protein–protein interaction predictions","status":"source_checked","source_ids":["esm2-amp-2025"],"source_locator":"Methods/Overview of ESM2_AMP framework (paragraph 1); Results/Benchmarking evaluation for ESM2_AMP on PPIs prediction (paragraph 1)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["esm2-amp-2025","evidence-reported-base-esm-readme-md"],"source_locator":"Methods/Protein–protein interaction datasets; Methods/Overview of ESM2_AMP framework; Methods/Feature extraction and fusion method; Methods/Construction of dual models based on feature integration; Methods/Methodology for model training and performance evaluation; Methods/Interpretability analysis utilizing attention mechanisms; Methods/Autoencoder model construction and feature importance calculation methods; Methods/Identification and computational methods of functional amino acid regions; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"ESM2_AMPS is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["esm2-amp-2025"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"Pan and Bernett human PPI datasets with separate evaluation schemes, plus a curated real_test set; similarity filtering is described for independence checks.","status":"source_checked","source_ids":["esm2-amp-2025"],"source_locator":"Methods/Protein–protein interaction datasets (paragraph 1); Methods/Protein–protein interaction datasets (paragraph 2)"},{"label":"Context limits","value":"Proteins between 15 and 4,000 residues are retained in the Pan dataset preparation.","status":"source_checked","source_ids":["esm2-amp-2025"],"source_locator":"Methods/Identification and computational methods of functional amino acid regions (paragraph 1); Methods/Protein–protein interaction datasets (paragraph 1)"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://github.com/facebookresearch/esm/blob/2b369911bb5b4b0dda914521b9475cad1656b2ac/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-esm-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"MIT (upstream repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-base-esm-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-base-esm-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Segment attention can be inspected alongside known functional regions.","source_ids":["esm2-amp-2025"],"source_locator":"Methods/Identification and computational methods of functional amino acid regions (paragraph 4); Results/Correlation between feature attribution value and functional amino acid regions (paragraph 7)"}],"limitations":[{"text":"Attention alignment with annotated domains is an association, not proof of the causal interaction mechanism; random-pair and strict-protein splits are different evaluations.","source_ids":["esm2-amp-2025"],"source_locator":"Methods/Feature extraction and fusion method (paragraph 4); Methods/Construction of dual models based on feature integration (paragraph 2)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Pairs of protein sequences","ESM2_AMPS","Binary protein–protein interaction predictions"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["esm2-amp-2025"],"source_locator":"Methods/Interpretability analysis utilizing attention mechanisms (paragraph 1); Methods/Construction of dual models based on feature integration (paragraph 3)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","ESM2_AMPS is the comparison-table label; that label does not specify an immutable weight revision.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-67eaf766fa9877","kind":"model","name":"ICCTax","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["microbes-communities"]},"source_ids":["icctax-2025"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"ICCTax","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"ICCTax adapts HyenaDNA for hierarchical taxonomic classification across four superkingdoms.","summary_source_ids":["icctax-2025"],"summary_source_locator":"Abstract (paragraph 1); 3 Experiments/3.2 Experiments on ICCTax performance/3.2.2 Performance on the complete dataset (paragraph 2)","sections":[{"title":"How the evaluated method works","body":"HyenaDNA sequence features feed complementary-view hierarchical metric learning, with a hierarchical compactness loss and outputs linked across taxonomic levels.","source_ids":["icctax-2025"],"source_locator":"3 Experiments/3.4 Model analysis of ICCTax/3.4.2 Ablation study (paragraph 2); 2 Methods/2.4 Total loss function (paragraph 1)"},{"title":"What was evaluated","body":"The linked evaluation record identifies ICCTax: Hierarchical metagenomic taxonomy classification. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["icctax-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b3-035"}],"facts":[{"label":"Model type","value":"Study-specific predictive method; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["icctax-2025"],"source_locator":"3 Experiments/3.4 Model analysis of ICCTax/3.4.2 Ablation study (paragraph 2); 2 Methods/2.4 Total loss function (paragraph 1)"},{"label":"Architecture / procedure","value":"HyenaDNA sequence features feed complementary-view hierarchical metric learning, with a hierarchical compactness loss and outputs linked across taxonomic levels.","status":"source_checked","source_ids":["icctax-2025"],"source_locator":"3 Experiments/3.4 Model analysis of ICCTax/3.4.2 Ablation study (paragraph 2); 2 Methods/2.4 Total loss function (paragraph 1)"},{"label":"Biological inputs","value":"1,500-bp metagenomic contigs","status":"source_checked","source_ids":["icctax-2025"],"source_locator":"3 Experiments/3.6 Taxonomic profiling on real metagenomic data (paragraph 1); 2 Methods (paragraph 1)"},{"label":"Outputs","value":"Hierarchical taxonomic assignments, including 155 genera and 43 phyla in the study","status":"source_checked","source_ids":["icctax-2025"],"source_locator":"Abstract (paragraph 1); 1 Introduction (paragraph 6)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["icctax-2025","evidence-reported-icctax-2025-readme-md"],"source_locator":"2 Methods; 2 Methods/2.1 HyenaDNA; 2 Methods/2.2 Complementary-view based hierarchical metric learning; 2 Methods/2.3 Compactness loss; 2 Methods/2.4 Total loss function; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"ICCTax is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["icctax-2025"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"Models use ID, OOD and Complete genomic datasets constructed from 1,500-bp fragments of archaea, bacteria, eukaryotes andviruses. The Complete-trained classifier is then applied to external marine,16S and wastewater datasets.","status":"source_checked","source_ids":["icctax-2025"],"source_locator":"3 Experiments/3.1 Datasets (paragraph 1); Abstract (paragraph 1)"},{"label":"Context limits","value":"1,500-bp trimmed contigs in the evaluated pipeline.","status":"source_checked","source_ids":["icctax-2025"],"source_locator":"3 Experiments/3.6 Taxonomic profiling on real metagenomic data (paragraph 1); 2 Methods (paragraph 1)"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/Ying-Lab/ICCTax/blob/6b7381c7111bde6d40324cd033501204bf3ac3bc/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-icctax-2025-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","status":"unreported","source_ids":["evidence-reported-icctax-2025-readme-md"],"source_locator":"README.md and repository-root licence-file search"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-icctax-2025-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Explicitly evaluates imbalanced and out-of-distribution taxonomic settings.","source_ids":["icctax-2025"],"source_locator":"2 Methods/2.3 Compactness loss (paragraph 2); 2 Methods/2.2 Complementary-view based hierarchical metric learning (paragraph 3)"}],"limitations":[{"text":"The trained taxonomy and sequence sampling define the classifier’s coverage; unobserved labels are not automatically identifiable.","source_ids":["icctax-2025"],"source_locator":"3 Experiments/3.1 Datasets (paragraph 1); 2 Methods/2.3 Compactness loss (paragraph 2)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["1,500-bp metagenomic contigs","ICCTax","Hierarchical taxonomic assignments, including 155 genera and 43 phyla in the study"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["icctax-2025"],"source_locator":"3 Experiments/3.4 Model analysis of ICCTax/3.4.2 Ablation study (paragraph 2); 2 Methods/2.4 Total loss function (paragraph 1)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","ICCTax is the comparison-table label; that label does not specify an immutable weight revision.","No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-688eb780ef7d2e","kind":"model","name":"PC-mer + LR","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["microbes-communities"]},"source_ids":["pc-mer-2024"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"PC-mer + LR","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"PC-mer plus logistic regression is a physicochemical-sequence-feature baseline for bacterial taxonomy.","summary_source_ids":["pc-mer-2024"],"summary_source_locator":"3 Results/3.5 PC-mer in use by the metagenomics ML-based classifier/3.5.2 Execution times (paragraph 1); 1 Introduction (paragraph 6)","sections":[{"title":"How the evaluated method works","body":"PC-mer encodes nucleotide physicochemical properties in a feature vector, followed by a fitted logistic-regression classifier.","source_ids":["pc-mer-2024"],"source_locator":"2 Method/2.2 Learning unit (paragraph 1); 2 Method (paragraph 1)"},{"title":"What was evaluated","body":"The linked evaluation record identifies PC-mer + LR: metagenomic genus classification. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["pc-mer-2024"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b4-019"}],"facts":[{"label":"Model type","value":"Logistic-regression pipeline; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["pc-mer-2024"],"source_locator":"2 Method/2.2 Learning unit (paragraph 1); 2 Method (paragraph 1)"},{"label":"Architecture / procedure","value":"PC-mer encodes nucleotide physicochemical properties in a feature vector, followed by a fitted logistic-regression classifier.","status":"source_checked","source_ids":["pc-mer-2024"],"source_locator":"2 Method/2.2 Learning unit (paragraph 1); 2 Method (paragraph 1)"},{"label":"Biological inputs","value":"DNA/RNA sequence-derived physicochemical feature vectors","status":"source_checked","source_ids":["pc-mer-2024"],"source_locator":"2 Method/2.1 Feature extraction method (paragraph 1); 4. Discussion (paragraph 3)"},{"label":"Outputs","value":"Taxonomic class labels","status":"source_checked","source_ids":["pc-mer-2024"],"source_locator":"3 Results/3.1 Datasets (paragraph 1); 2 Method (paragraph 1)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["pc-mer-2024","evidence-reported-pc-mer-2024-readme-md"],"source_locator":"2 Method; 2 Method/2.1 Feature extraction method; 2 Method/2.2 Learning unit; 3 Results/3.3 PC-mer in use by distance-based methods for comparison and classifying metagenomics sequences; 3 Results/3.5 PC-mer in use by the metagenomics ML-based classifier/3.5.1 Training and testing procedure; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"PC-mer + LR is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["pc-mer-2024"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"A ten-fold study compares feature transformations and eight classifiers on AMP/shotgun metagenomic datasets. The balanced HTL dataset includes 1,000 sequences from 100 genera; a separate unbalanced Qiita-derived dataset supports species-level analysis.","status":"source_checked","source_ids":["pc-mer-2024"],"source_locator":"3 Results/3.1 Datasets (paragraph 1); 3 Results/3.5 PC-mer in use by the metagenomics ML-based classifier/3.5.1 Training and testing procedure (paragraph 1)"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["pc-mer-2024","evidence-reported-pc-mer-2024-readme-md"],"source_locator":"2 Method; 2 Method/2.1 Feature extraction method; 2 Method/2.2 Learning unit; 3 Results/3.3 PC-mer in use by distance-based methods for comparison and classifying metagenomics sequences; 3 Results/3.5 PC-mer in use by the metagenomics ML-based classifier/3.5.1 Training and testing procedure; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/SAkbari93/PC-mer_Metagenomics/blob/5c5f89dcaec5098372ad1fe82d4215186fe417c1/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-pc-mer-2024-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","status":"unreported","source_ids":["evidence-reported-pc-mer-2024-readme-md"],"source_locator":"README.md and repository-root licence-file search"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-pc-mer-2024-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Provides a comparatively simple fitted classifier to test whether complex pretrained models outperform structured sequence statistics.","source_ids":["pc-mer-2024"],"source_locator":"3 Results/3.5 PC-mer in use by the metagenomics ML-based classifier (paragraph 1); 2 Method/2.2 Learning unit (paragraph 1)"}],"limitations":[{"text":"The feature dimension, k setting and taxonomic label set belong to the evaluated dataset; headline results across taxonomic levels are not interchangeable.","source_ids":["pc-mer-2024"],"source_locator":"3 Results/3.5 PC-mer in use by the metagenomics ML-based classifier/3.5.1 Training and testing procedure (paragraph 1); 3 Results/3.4 Feature extraction capability of PC-mer (paragraph 1)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["DNA/RNA sequence-derived physicochemical feature vectors","PC-mer + LR","Taxonomic class labels"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["pc-mer-2024"],"source_locator":"2 Method/2.2 Learning unit (paragraph 1); 2 Method (paragraph 1)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","PC-mer + LR is the comparison-table label; that label does not specify an immutable weight revision.","A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-6ac0730e8481de","kind":"model","name":"Lemur","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["microbes-communities"]},"source_ids":["lemur-magnet-2024"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"Lemur","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Lemur profiles taxonomy from long-read shotgun metagenomic data using marker genes.","summary_source_ids":["lemur-magnet-2024"],"summary_source_locator":"Abstract (paragraph 1); Introduction (paragraph 3)","sections":[{"title":"How the evaluated method works","body":"A marker-gene matching procedure and expectation–maximisation reduce false-positive taxonomic calls while retaining supported taxa. Magnet is a separate whole-genome verification tool.","source_ids":["lemur-magnet-2024"],"source_locator":"Abstract (paragraph 1); Methods/Competitive read alignment with Magnet (paragraph 1)"},{"title":"What was evaluated","body":"The linked evaluation record identifies Lemur: Long-read taxonomic profiling. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["lemur-magnet-2024"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b3-031"}],"facts":[{"label":"Model type","value":"Study-specific predictive method; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["lemur-magnet-2024"],"source_locator":"Abstract (paragraph 1); Methods/Competitive read alignment with Magnet (paragraph 1)"},{"label":"Architecture / procedure","value":"A marker-gene matching procedure and expectation–maximisation reduce false-positive taxonomic calls while retaining supported taxa. Magnet is a separate whole-genome verification tool.","status":"source_checked","source_ids":["lemur-magnet-2024"],"source_locator":"Abstract (paragraph 1); Methods/Competitive read alignment with Magnet (paragraph 1)"},{"label":"Biological inputs","value":"Long-read shotgun metagenomic reads and a marker-gene database","status":"source_checked","source_ids":["lemur-magnet-2024"],"source_locator":"Abstract (paragraph 1); Introduction (paragraph 3)"},{"label":"Outputs","value":"Taxonomic profiles","status":"source_checked","source_ids":["lemur-magnet-2024"],"source_locator":"Methods/Synthetic and simulated datasets/Zymo Fecal Reference with TruMatrix (paragraph 1); Results/Chicken gut metagenome (paragraph 2)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["lemur-magnet-2024","evidence-reported-lemur-magnet-2024-readme-md"],"source_locator":"Results/Method overview; Methods/Lemur; Methods/Competitive read alignment with Magnet; Methods/Method Comparison; Methods/Synthetic and simulated datasets/Simulated data from (Dilthey et al. 2019); Methods/Synthetic and simulated datasets/Zymo EVEN & Zymo LOG; Methods/Synthetic and simulated datasets/Simulated metagenome; Methods/Synthetic and simulated datasets/Zymo Fecal Reference with TruMatrix; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"Lemur is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["lemur-magnet-2024"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"The paper’s approximately 4-GB marker database contains information from more than 300,000 RefSeq genomes; this is a reference index rather than neural pretraining.","status":"source_checked","source_ids":["lemur-magnet-2024"],"source_locator":"Methods/Synthetic and simulated datasets/Simulated data from (Dilthey et al. 2019) (paragraph 1); Methods/Competitive read alignment with Magnet (paragraph 1)"},{"label":"Context limits","value":"Not applicable to a pretrained sequence-token window; read/contig lengths and the reference-database or comparison configuration determine the analysed input.","status":"inapplicable","source_ids":["lemur-magnet-2024"],"source_locator":"Methods/Competitive read alignment with Magnet (paragraph 1); Discussion (paragraph 2)"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/treangenlab/lemur/blob/eda2cb57727b72fc5b1fb28be1fe45a4826100f9/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-lemur-magnet-2024-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"MIT (study repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-lemur-magnet-2024-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-lemur-magnet-2024-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Designed for long-read data with a compact marker reference.","source_ids":["lemur-magnet-2024"],"source_locator":"Discussion (paragraph 4); Results/Computational performance (paragraph 3)"}],"limitations":[{"text":"Lemur profiles and Magnet presence/absence validation are separate procedures; one tool’s results should not be assigned to the other.","source_ids":["lemur-magnet-2024"],"source_locator":"Introduction (paragraph 3); Results/Method overview (paragraph 2)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Long-read shotgun metagenomic reads and a marker-gene database","Lemur","Taxonomic profiles"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["lemur-magnet-2024"],"source_locator":"Abstract (paragraph 1); Methods/Competitive read alignment with Magnet (paragraph 1)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","Lemur is the comparison-table label; that label does not specify an immutable weight revision.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-6c0bc8d297cc7a","kind":"model","name":"DiffDock-NMDN","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["molecular-interactions"]},"source_ids":["nmdn-2025"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"DiffDock-NMDN","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"DiffDock-NMDN combines diffusion-based pose generation with a learned pose-selection and affinity-scoring pipeline.","summary_source_ids":["nmdn-2025"],"summary_source_locator":"DiffDock-NMDN\nBlind Docking and Virtual Screening\nProtocol (paragraph 1); Abstract (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"DiffDock samples candidate poses. A normalised mixture-density network scores protein-residue/ligand-atom distances and selects a pose; an additional interaction module estimates affinity. The protein encoder uses ESM-2 650M.","source_ids":["nmdn-2025"],"source_locator":"Methods/Ligand Conformation Stability and Solvation\nEnergetics (paragraph 1); Methods/Normalized Mixture Density Network Module/Training Loss for the NMDN Modules (paragraph 2)"},{"title":"Underlying method and version boundaries","body":"DiffDock is a molecular-docking implementation that produces ligand poses and confidence estimates. Its confidence values and predicted coordinates are different outputs from an experimentally calibrated binding-affinity measurement.","source_ids":["evidence-reported-base-diffdock-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies DiffDock-NMDN: Protein–ligand virtual screening. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["nmdn-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-043"}],"facts":[{"label":"Model type","value":"Molecular docking model; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-diffdock-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"DiffDock samples candidate poses. A normalised mixture-density network scores protein-residue/ligand-atom distances and selects a pose; an additional interaction module estimates affinity. The protein encoder uses ESM-2 650M.","status":"source_checked","source_ids":["nmdn-2025"],"source_locator":"Methods/Ligand Conformation Stability and Solvation\nEnergetics (paragraph 1); Methods/Normalized Mixture Density Network Module/Training Loss for the NMDN Modules (paragraph 2)"},{"label":"Biological inputs","value":"Protein structure/sequence and ligand information","status":"source_checked","source_ids":["nmdn-2025"],"source_locator":"Methods/Interaction\nModule/Calculation of Protein–Ligand Pair\nContributions (paragraph 8); Methods/Normalized Mixture Density Network Module/Inference (paragraph 1)"},{"label":"Outputs","value":"Selected binding poses and estimated binding affinities","status":"source_checked","source_ids":["nmdn-2025"],"source_locator":"Results and Discussion/DiffDock-NMDN Blind Docking Protocol (paragraph 4); Data Sets/Evaluation Metrics/CASF-2016 (paragraph 2)"},{"label":"Parameters","value":"650-million-parameter ESM-2 protein encoder; this is not the total pipeline size.","status":"source_checked","source_ids":["nmdn-2025"],"source_locator":"Methods/Encoders\nfor Protein, Ligand, and Metal (paragraph 1); Methods/Normalized Mixture Density Network Module/Inference (paragraph 1)"},{"label":"Known versions / configuration","value":"DiffDock-NMDN is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["nmdn-2025"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"NMDN uses PDBbind 2020 with 12,554 training and 1,083 evaluation complexes from the first RTMScore split. An additional affinity-module fine-tuning stage draws weak binders from EquiVS and Papyrus: 60,000 pairs are sampled from 250,267 eligible pairs. The pretrained DiffDock pose generator is a separate component.","status":"source_checked","source_ids":["nmdn-2025"],"source_locator":"Data Sets / Training Data Set Preparation / Binder Data Set and Weak-Binder Data Set"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["nmdn-2025","evidence-reported-base-diffdock-readme-md"],"source_locator":"Methods/Model Overview; Methods/Encoders\nfor Protein, Ligand, and Metal; Methods/Normalized Mixture Density Network Module; Methods/Normalized Mixture Density Network Module/Protein–ligand NMDN Module Architecture; Methods/Normalized Mixture Density Network Module/Metal–Ligand NMDN\nModule; Methods/Normalized Mixture Density Network Module/Training Loss for the NMDN Modules; Methods/Normalized Mixture Density Network Module/Inference; Methods/Interaction\nModule; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://github.com/gcorso/DiffDock/blob/85c49b60d3e0b0182a59ee43a34a6d7036981284/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-diffdock-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"MIT (upstream repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-base-diffdock-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-base-diffdock-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Separates pose sampling, pose selection and affinity estimation explicitly.","source_ids":["nmdn-2025"],"source_locator":"Methods/Ligand Conformation Stability and Solvation\nEnergetics (paragraph 1); Methods/Interaction\nModule/Calculation of Protein–Ligand Pair\nContributions (paragraph 16)"}],"limitations":[{"text":"A DiffDock confidence score is not itself binding affinity; the complete NMDN pipeline must be identified when reporting screening results.","source_ids":["nmdn-2025"],"source_locator":"Methods/Normalized Mixture Density Network Module/Inference (paragraph 6); Methods/Interaction\nModule (paragraph 1)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Protein structure/sequence and ligand information","DiffDock-NMDN","Selected binding poses and estimated binding affinities"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["nmdn-2025"],"source_locator":"Methods/Ligand Conformation Stability and Solvation\nEnergetics (paragraph 1); Methods/Normalized Mixture Density Network Module/Training Loss for the NMDN Modules (paragraph 2)"},"coverage":"limited","gaps":["DiffDock-NMDN is the comparison-table label; that label does not specify an immutable weight revision.","A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-6d9dbac97852d8","kind":"model","name":"DETIRE","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["microbes-communities"]},"source_ids":["detire-viral-metagenomes-2023"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"DETIRE","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"DETIRE classifies short metagenomic DNA fragments as viral or non-viral.","summary_source_ids":["detire-viral-metagenomes-2023"],"summary_source_locator":"3. Results/3.4. Performance on the testing dataset (paragraph 1); 3. Results/3.2. A real human gut metagenome dataset (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"TextGCN learns embeddings of 3-mers from a heterogeneous sequence/token graph. CNN and bidirectional-LSTM branches extract spatial and sequential features, which are weighted together for classification.","source_ids":["detire-viral-metagenomes-2023"],"source_locator":"2. Materials and methods/2.2. Composition of DETIRE (paragraph 3); 2. Materials and methods/2.2. Composition of DETIRE (paragraph 2)"},{"title":"What was evaluated","body":"The linked evaluation record identifies DETIRE: viral sequence detection. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["detire-viral-metagenomes-2023"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b4-018"}],"facts":[{"label":"Model type","value":"Graph-based predictive method; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["detire-viral-metagenomes-2023"],"source_locator":"2. Materials and methods/2.2. Composition of DETIRE (paragraph 3); 2. Materials and methods/2.2. Composition of DETIRE (paragraph 2)"},{"label":"Architecture / procedure","value":"TextGCN learns embeddings of 3-mers from a heterogeneous sequence/token graph. CNN and bidirectional-LSTM branches extract spatial and sequential features, which are weighted together for classification.","status":"source_checked","source_ids":["detire-viral-metagenomes-2023"],"source_locator":"2. Materials and methods/2.2. Composition of DETIRE (paragraph 3); 2. Materials and methods/2.2. Composition of DETIRE (paragraph 2)"},{"label":"Biological inputs","value":"Short DNA fragments, with 500-bp fragments used for training","status":"source_checked","source_ids":["detire-viral-metagenomes-2023"],"source_locator":"3. Results/3.2. A real human gut metagenome dataset (paragraph 1); 4. Discussions (paragraph 2)"},{"label":"Outputs","value":"Viral-sequence classification scores","status":"source_checked","source_ids":["detire-viral-metagenomes-2023"],"source_locator":"3. Results/3.4. Performance on the testing dataset (paragraph 1); 3. Results/3.5. Performance on the CAMI Marine metagenome (paragraph 2)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["detire-viral-metagenomes-2023","evidence-reported-detire-viral-metagenomes-2023-readme-md"],"source_locator":"2. Materials and methods/2.1. Virus and host RefSeq genome datasets for training and testing; 2. Materials and methods/2.2. Composition of DETIRE; 2. Materials and methods/2.3. Evaluation criteria; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"DETIRE is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["detire-viral-metagenomes-2023"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"220,000 sampled 500-bp Virus/Host RefSeq fragments for the classifier; the graph-embedding corpus uses viral RefSeq available through 11 October 2022 and prokaryotic host sequences.","status":"source_checked","source_ids":["detire-viral-metagenomes-2023"],"source_locator":"2. Materials and methods/2.1. Virus and host RefSeq genome datasets for training and testing (paragraph 1); Abstract (paragraph 1)"},{"label":"Context limits","value":"500 bp during the reported classifier training; the paper separately examines short fragments below 1,000 bp.","status":"source_checked","source_ids":["detire-viral-metagenomes-2023"],"source_locator":"4. Discussions (paragraph 2); 1. Introduction (paragraph 2)"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/crazyinter/DETIRE/blob/6b48c5bcb1303abe593173633d1f13da1d8d5869/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-detire-viral-metagenomes-2023-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","status":"unreported","source_ids":["evidence-reported-detire-viral-metagenomes-2023-readme-md"],"source_locator":"README.md and repository-root licence-file search"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-detire-viral-metagenomes-2023-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Combines local and sequential features and evaluates the difficult short-fragment regime.","source_ids":["detire-viral-metagenomes-2023"],"source_locator":"1. Introduction (paragraph 5); 4. Discussions (paragraph 1)"}],"limitations":[{"text":"Reference-derived fragments may differ from highly novel environmental viruses; this is a viral detector, not a complete taxonomy or abundance estimator.","source_ids":["detire-viral-metagenomes-2023"],"source_locator":"Abstract (paragraph 1); 4. Discussions (paragraph 4)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Short DNA fragments, with 500-bp fragments used for training","DETIRE","Viral-sequence classification scores"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["detire-viral-metagenomes-2023"],"source_locator":"2. Materials and methods/2.2. Composition of DETIRE (paragraph 3); 2. Materials and methods/2.2. Composition of DETIRE (paragraph 2)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","DETIRE is the comparison-table label; that label does not specify an immutable weight revision.","No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-70c732770a200f","kind":"model","name":"Chai-1","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["molecular-interactions"]},"source_ids":["ibex-2025"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"Chai-1","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Chai-1 is a general structure-prediction comparator on the Ibex immunoglobulin benchmark.","summary_source_ids":["ibex-2025"],"summary_source_locator":"Discussion (paragraph 6); Results/Benchmarking on the ImmuneBuilder test set (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"The study uses one seed and one diffusion trajectory per prediction, then aligns frameworks and measures region-specific backbone RMSD.","source_ids":["ibex-2025"],"source_locator":"Results/Benchmarking on the ImmuneBuilder test set (paragraph 1); Results/Benchmarking on a private dataset (paragraph 3)"},{"title":"Underlying method and version boundaries","body":"Chai-1 predicts biomolecular structures through the released inference implementation. Its input options, MSA/template use and sampling budget are part of each evaluated configuration.","source_ids":["evidence-reported-base-chai-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies Chai-1: Antibody loop structure prediction. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["ibex-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b3-041"}],"facts":[{"label":"Model type","value":"Biomolecular structure predictor; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-chai-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"The study uses one seed and one diffusion trajectory per prediction, then aligns frameworks and measures region-specific backbone RMSD.","status":"source_checked","source_ids":["ibex-2025"],"source_locator":"Results/Benchmarking on the ImmuneBuilder test set (paragraph 1); Results/Benchmarking on a private dataset (paragraph 3)"},{"label":"Biological inputs","value":"Antibody, nanobody or T-cell-receptor sequences","status":"source_checked","source_ids":["ibex-2025"],"source_locator":"Introduction (paragraph 5); Results/Benchmarking on the ImmuneBuilder test set (paragraph 2)"},{"label":"Outputs","value":"Predicted structures","status":"source_checked","source_ids":["ibex-2025"],"source_locator":"Methods/Training (paragraph 1); The Ibex model (paragraph 2)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["ibex-2025","evidence-reported-base-chai-readme-md"],"source_locator":"Methods/Model; Methods/Data; Methods/Training; Methods/Private dataset generation and characterization; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"One seed and one diffusion trajectory in the ImmuneBuilder comparison","status":"unreported","source_ids":["ibex-2025"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"The study evaluates the released structure predictor on the ImmuneBuilder test set using one seed and one diffusion trajectory. It does not report task-specific retraining of Chai-1; a complete upstream training inventory is not restated.","status":"source_checked","source_ids":["ibex-2025"],"source_locator":"Results / Benchmarking on the ImmuneBuilder test set"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["ibex-2025","evidence-reported-base-chai-readme-md"],"source_locator":"Methods/Model; Methods/Data; Methods/Training; Methods/Private dataset generation and characterization; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://github.com/chaidiscovery/chai-lab/blob/66c38d1fe5c6756a89ff8596b1dea87d305ec06f/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-chai-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"Apache 2.0 (upstream repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-base-chai-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"Apache 2.0 for both Chai-1 code and model weights, explicitly stated in the official README; paper-specific derived artifacts are separate.","status":"source_checked","source_ids":["evidence-reported-base-chai-readme-md"],"source_locator":"README.md; Licence section"}],"strengths":[{"text":"Provides a general biomolecular predictor alongside specialised immunoglobulin methods.","source_ids":["ibex-2025"],"source_locator":"Discussion (paragraph 6); Discussion (paragraph 3)"}],"limitations":[{"text":"The paper reports different behaviour across loop types and modalities; the single-sample configuration is not a maximum-sampling capability estimate.","source_ids":["ibex-2025"],"source_locator":"Methods/Training (paragraph 2); Results/Benchmarking on the ImmuneBuilder test set (paragraph 2)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Antibody, nanobody or T-cell-receptor sequences","Chai-1","Predicted structures"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["ibex-2025"],"source_locator":"Results/Benchmarking on the ImmuneBuilder test set (paragraph 1); Results/Benchmarking on a private dataset (paragraph 3)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","One seed and one diffusion trajectory in the ImmuneBuilder comparison","A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-70f57ebb163a5c","kind":"model","name":"Kraken2","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["microbes-communities"]},"source_ids":["cammiq-2022"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"Kraken2","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Kraken 2 is the reference-database taxonomic classifier used in this metagenomic comparison.","summary_source_ids":["cammiq-2022"],"summary_source_locator":"Introduction (paragraph 3); Introduction (paragraph 4)","sections":[{"title":"How the evaluated method works","body":"Sequence evidence is matched against a taxonomic reference index to assign reads or contigs. Database construction and the chosen confidence/settings remain part of the evaluated configuration.","source_ids":["cammiq-2022"],"source_locator":"Methods (paragraph 7); Methods (paragraph 1)"},{"title":"Underlying method and version boundaries","body":"Kraken 2 classifies sequences against a taxonomic reference index. The software version and the database used to construct that index are distinct reproducibility requirements.","source_ids":["evidence-reported-base-kraken2-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies Kraken2: Strain-level abundance quantification. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["cammiq-2022"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b3-020"}],"facts":[{"label":"Model type","value":"Reference-index taxonomic classifier; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-kraken2-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"Sequence evidence is matched against a taxonomic reference index to assign reads or contigs. Database construction and the chosen confidence/settings remain part of the evaluated configuration.","status":"source_checked","source_ids":["cammiq-2022"],"source_locator":"Methods (paragraph 7); Methods (paragraph 1)"},{"label":"Biological inputs","value":"Metagenomic reads or contigs and a taxonomic reference database","status":"source_checked","source_ids":["cammiq-2022"],"source_locator":"Introduction (paragraph 4); Results/Precision and recall in read classification across all species level queries (paragraph 3)"},{"label":"Outputs","value":"Taxonomic assignments","status":"source_checked","source_ids":["cammiq-2022"],"source_locator":"Results/Performance of CAMMiQ on real single-cell metatranscriptomic queries (paragraph 5); Introduction (paragraph 4)"},{"label":"Parameters","value":"Not applicable as a neural model size.","status":"inapplicable","source_ids":["cammiq-2022"],"source_locator":"Methods/CAMMiQ Index/Sparsifying unique substrings (paragraph 1); Methods/Query processing stage 2: ILP formulation (paragraph 4)"},{"label":"Known versions / configuration","value":"Kraken2 is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["cammiq-2022"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"Reference-index construction rather than foundation-model pretraining.","status":"source_checked","source_ids":["cammiq-2022"],"source_locator":"Results/Datasets (paragraph 1); Results/Datasets (paragraph 2)"},{"label":"Context limits","value":"Not applicable to a pretrained sequence-token window; read/contig lengths and the reference-database or comparison configuration determine the analysed input.","status":"inapplicable","source_ids":["cammiq-2022"],"source_locator":"Methods (paragraph 1); Introduction (paragraph 7)"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://github.com/DerrickWood/kraken2/blob/8c190b1b668825935dbf6dee5f969227dc8269bb/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-kraken2-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"MIT (upstream repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-base-kraken2-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"Not applicable: uses a reference index rather than a pretrained neural checkpoint.","status":"inapplicable","source_ids":["cammiq-2022"],"source_locator":"Results/Precision and recall in read classification across all species level queries (paragraph 3); Results/Evaluation of computational resources on species level queries (paragraph 1)"}],"strengths":[{"text":"Provides an established reference-based comparator for learned or alignment-based taxonomic methods.","source_ids":["cammiq-2022"],"source_locator":"Introduction (paragraph 2); Introduction (paragraph 4)"}],"limitations":[{"text":"Unrepresented organisms and reference contamination can alter assignments; results are specific to the database and query distribution.","source_ids":["cammiq-2022"],"source_locator":"Introduction (paragraph 4); Introduction (paragraph 3)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Metagenomic reads or contigs and a taxonomic reference database","Kraken2","Taxonomic assignments"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["cammiq-2022"],"source_locator":"Methods (paragraph 7); Methods (paragraph 1)"},"coverage":"limited","gaps":["Kraken2 is the comparison-table label; that label does not specify an immutable weight revision.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-7234658bc9c828","kind":"model","name":"RNAret","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["rna-transcriptomes"]},"source_ids":["rnaret-2026"],"links":[],"attributes":{"entity_level":"method","version":"5-mer","reported_name":"RNAret","historical_missing_metadata":{"checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"RNAret learns RNA representations with a retention network instead of quadratic self-attention.","summary_source_ids":["rnaret-2026"],"summary_source_locator":"Methods/Bidirectional encoder representation from Retentive Network (paragraph 1); Methods/Bidirectional encoder representation from Retentive Network (paragraph 7)","sections":[{"title":"How the evaluated method works","body":"Bidirectional retention blocks use multiscale retention, feed-forward layers, residual connections and layer normalisation; masked-language pretraining learns sequence features.","source_ids":["rnaret-2026"],"source_locator":"Methods/Bidirectional encoder representation from Retentive Network (paragraph 1); Methods/Bidirectional encoder representation from Retentive Network (paragraph 2)"},{"title":"What was evaluated","body":"The linked evaluation record identifies RNAret: miRNA-mRNA interaction prediction. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["rnaret-2026"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-b2-rnaret-2026"}],"facts":[{"label":"Model type","value":"Study-specific predictive method; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["rnaret-2026"],"source_locator":"Methods/Bidirectional encoder representation from Retentive Network (paragraph 1); Methods/Bidirectional encoder representation from Retentive Network (paragraph 2)"},{"label":"Architecture / procedure","value":"Bidirectional retention blocks use multiscale retention, feed-forward layers, residual connections and layer normalisation; masked-language pretraining learns sequence features.","status":"source_checked","source_ids":["rnaret-2026"],"source_locator":"Methods/Bidirectional encoder representation from Retentive Network (paragraph 1); Methods/Bidirectional encoder representation from Retentive Network (paragraph 2)"},{"label":"Biological inputs","value":"RNA nucleotide sequences","status":"source_checked","source_ids":["rnaret-2026"],"source_locator":"Introduction (paragraph 2); Methods/RNA tokenization (paragraph 1)"},{"label":"Outputs","value":"RNA representations and downstream interaction, structure or classification predictions","status":"source_checked","source_ids":["rnaret-2026"],"source_locator":"Methods/Fine-tuning strategies for downstream tasks (paragraph 1); Methods/RNA-RNA interaction prediction (paragraph 1)"},{"label":"Parameters","value":"Approximately 12 million trainable parameters.","status":"source_checked","source_ids":["rnaret-2026"],"source_locator":"Methods/Fine-tuning strategies for downstream tasks (paragraph 2); Methods/RNAret model architecture (paragraph 1)"},{"label":"Known versions / configuration","value":"5-mer","status":"source_checked","source_ids":["rnaret-2026"],"source_locator":"Table Tab3 (paragraph 1); Table Tab2 (paragraph 1)"},{"label":"Training data / fitting","value":"29.8 million RNA sequences","status":"source_checked","source_ids":["rnaret-2026"],"source_locator":"Methods/Pretraining strategies (paragraph 1); Methods/Pretraining strategies (paragraph 3)"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["rnaret-2026","evidence-reported-rnaret-2026-readme-md"],"source_locator":"Results/Self-supervised pretraining and feature extraction for RNA sequences; Methods/Bidirectional encoder representation from Retentive Network; Methods/RNAret model architecture; Methods/RNA tokenization; Methods/Pretraining strategies; Methods/Fine-tuning strategies for downstream tasks; Methods/RNA-RNA interaction prediction; Methods/RNA secondary structure prediction; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/DrBlackZJU/RNAret/blob/40ddab25fc038ba2b96bc9b9f88216abe38b2f64/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-rnaret-2026-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"CC0 1.0 (study repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-rnaret-2026-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-rnaret-2026-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"The retention mechanism is designed to support longer sequences with linear scaling in the discussed processing regime.","source_ids":["rnaret-2026"],"source_locator":"Methods/Bidirectional encoder representation from Retentive Network (paragraph 7); Methods/Bidirectional encoder representation from Retentive Network (paragraph 3)"}],"limitations":[{"text":"Algorithmic scaling does not establish a universal maximum useful sequence length or equal accuracy at all lengths.","source_ids":["rnaret-2026"],"source_locator":"Methods/Pretraining strategies (paragraph 3); Results/mRNA/lncRNA classification (paragraph 2)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["RNA nucleotide sequences","RNAret","RNA representations and downstream interaction, structure or classification predictions"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["rnaret-2026"],"source_locator":"Methods/Bidirectional encoder representation from Retentive Network (paragraph 1); Methods/Bidirectional encoder representation from Retentive Network (paragraph 2)"},"coverage":"limited","gaps":["A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-73ae07fb5be204","kind":"model","name":"GSMFormer-PPI + ProstT5","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["proteins-complexes"]},"source_ids":["gsmformer-ppi-2026"],"links":[],"attributes":{"entity_level":"method","version":"not stated in table","reported_name":"GSMFormer-PPI + ProstT5","historical_missing_metadata":{"checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"GSMFormer-PPI combines molecular surfaces, structural graphs and ProstT5 sequence features for interaction prediction.","summary_source_ids":["gsmformer-ppi-2026"],"summary_source_locator":"Results/Ablation experiments for input modalities (paragraph 1); Conclusions (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"MaSIF extracts physicochemical surface descriptors, graph convolutions process protein structure, and a transformer with linear projectors fuses these with residue-level embeddings.","source_ids":["gsmformer-ppi-2026"],"source_locator":"Methods/Model architecture/Integration of protein structural and sequence features (paragraph 2); Methods/Model architecture/Linear projectors (paragraph 1)"},{"title":"What was evaluated","body":"The linked evaluation record identifies GSMFormer-PPI + ProstT5: protein-protein interaction prediction. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["gsmformer-ppi-2026"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-b2-gsmformer-ppi-2026"}],"facts":[{"label":"Model type","value":"Graph-aware transformer pipeline; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["gsmformer-ppi-2026"],"source_locator":"Methods/Model architecture/Integration of protein structural and sequence features (paragraph 2); Methods/Model architecture/Linear projectors (paragraph 1)"},{"label":"Architecture / procedure","value":"MaSIF extracts physicochemical surface descriptors, graph convolutions process protein structure, and a transformer with linear projectors fuses these with residue-level embeddings.","status":"source_checked","source_ids":["gsmformer-ppi-2026"],"source_locator":"Methods/Model architecture/Integration of protein structural and sequence features (paragraph 2); Methods/Model architecture/Linear projectors (paragraph 1)"},{"label":"Biological inputs","value":"Protein surface and 3D structure features plus ProstT5 sequence embeddings","status":"source_checked","source_ids":["gsmformer-ppi-2026"],"source_locator":"Methods/Graph representation of 3D protein structure (paragraph 2); Discussion (paragraph 7)"},{"label":"Outputs","value":"Protein–protein interaction predictions","status":"source_checked","source_ids":["gsmformer-ppi-2026"],"source_locator":"Methods/Graph representation of 3D protein structure/Graph Convolutional Neural Network (GCN) (paragraph 1); Methods/Model architecture/Integration of protein structural and sequence features (paragraph 2)"},{"label":"Parameters","value":"Table 3 reports 2.8M parameters for GSMFormer-PPI; the pretrained ProstT5 feature extractor is a separate component.","status":"source_checked","source_ids":["gsmformer-ppi-2026"],"source_locator":"Table3; GSMFormer-PPI row, Parameters column"},{"label":"Known versions / configuration","value":"GSMFormer-PPI + ProstT5 is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["gsmformer-ppi-2026"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"PINDER ML-ready splits with interface- and sequence-similarity controls, including the paper’s stated 30% sequence-identity threshold.","status":"source_checked","source_ids":["gsmformer-ppi-2026"],"source_locator":"Methods/Dataset/Splits (paragraph 1); Methods/Dataset/Splits (paragraph 2)"},{"label":"Context limits","value":"Curated dimer chains contain 35–300 residues; embeddings are padded to the longest sequence in each batch.","status":"source_checked","source_ids":["gsmformer-ppi-2026"],"source_locator":"Methods/Protein sequence feature extraction (paragraph 3); Methods/Graph representation of 3D protein structure/Graph Convolutional Neural Network (GCN) (paragraph 2)"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/ChervovNikita/gsmformer-ppi/blob/db9886e8b295f35b544a5703659e2a9115ce9e22/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-gsmformer-ppi-2026-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","status":"unreported","source_ids":["evidence-reported-gsmformer-ppi-2026-readme-md"],"source_locator":"README.md and repository-root licence-file search"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-gsmformer-ppi-2026-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Ablation studies examine the contribution of surface features and multimodal fusion.","source_ids":["gsmformer-ppi-2026"],"source_locator":"Abstract (paragraph 1); Conclusions (paragraph 1)"}],"limitations":[{"text":"Cross-dataset comparisons of separately reported scores are not a matched head-to-head benchmark; structural inputs confer information beyond sequence-only models.","source_ids":["gsmformer-ppi-2026"],"source_locator":"Abstract (paragraph 1); Results/Performance of GSMFormer-PPI on Baranwal’s dataset (paragraph 1)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Protein surface and 3D structure features plus ProstT5 sequence embeddings","GSMFormer-PPI + ProstT5","Protein–protein interaction predictions"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["gsmformer-ppi-2026"],"source_locator":"Methods/Model architecture/Integration of protein structural and sequence features (paragraph 2); Methods/Model architecture/Linear projectors (paragraph 1)"},"coverage":"limited","gaps":["GSMFormer-PPI + ProstT5 is the comparison-table label; that label does not specify an immutable weight revision.","No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-75e4e5e5965320","kind":"model","name":"DiffDock holo","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["molecular-interactions"]},"source_ids":["ensemble-idp-docking-2025"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"DiffDock holo","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"This holo-ensemble docking configuration evaluates small-molecule binding to α-synuclein conformations.","summary_source_ids":["ensemble-idp-docking-2025"],"summary_source_locator":"Discussion (paragraph 7); Introduction (paragraph 6)","sections":[{"title":"How the evaluated method works","body":"DiffDock docks once to each protein conformation with default parameters. The trained confidence-model output supplies the pose score, which is min–max normalised within the method/ensemble setting.","source_ids":["ensemble-idp-docking-2025"],"source_locator":"Methods/DiffDock Ensemble docking. (paragraph 1); Results/Ensemble docking protocols for intrinsically disordered proteins. (paragraph 4)"},{"title":"Underlying method and version boundaries","body":"DiffDock is a molecular-docking implementation that produces ligand poses and confidence estimates. Its confidence values and predicted coordinates are different outputs from an experimentally calibrated binding-affinity measurement.","source_ids":["evidence-reported-base-diffdock-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies DiffDock holo: Intrinsically disordered protein ensemble docking. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["ensemble-idp-docking-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b3-047"}],"facts":[{"label":"Model type","value":"Molecular docking model; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-diffdock-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"DiffDock docks once to each protein conformation with default parameters. The trained confidence-model output supplies the pose score, which is min–max normalised within the method/ensemble setting.","status":"source_checked","source_ids":["ensemble-idp-docking-2025"],"source_locator":"Methods/DiffDock Ensemble docking. (paragraph 1); Results/Ensemble docking protocols for intrinsically disordered proteins. (paragraph 4)"},{"label":"Biological inputs","value":"Ligands and holo protein conformations from the study’s molecular-dynamics ensemble","status":"source_checked","source_ids":["ensemble-idp-docking-2025"],"source_locator":"Introduction (paragraph 4); Methods/Docked-pose RMSD calculations. (paragraph 2)"},{"label":"Outputs","value":"Docked poses and normalised relative docking scores","status":"source_checked","source_ids":["ensemble-idp-docking-2025"],"source_locator":"Results (paragraph 3); Results/Ensemble docking accurately predicts the relative affinities of small molecules to α-synuclein. (paragraph 1)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["ensemble-idp-docking-2025","evidence-reported-base-diffdock-readme-md"],"source_locator":"Methods/MD Simulations.; Methods/t-SNE clustering.; Methods/AutoDock Vina ensemble docking.; Methods/DiffDock Ensemble docking.; Methods/Analysis of IDP ligand binding modes.; Methods/Docked-pose RMSD calculations.; Methods/Comparing docking scores.; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"DiffDock holo is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["ensemble-idp-docking-2025"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"The pretrained docking model is evaluated on holo conformations from the study’s ligand-bound molecular-dynamics ensemble; no affinity-label fine-tuning is described.","status":"source_checked","source_ids":["ensemble-idp-docking-2025"],"source_locator":"Results/Quantifying the similarity of ligand binding poses obtained from MD simulations and ensemble docking. (paragraph 4); Methods/t-SNE clustering. (paragraph 1)"},{"label":"Context limits","value":"The evaluated receptor ensemble is the 20-residue α-synuclein C-terminal fragment 121–140; this study-specific input is not a general DiffDock context maximum.","status":"source_checked","source_ids":["ensemble-idp-docking-2025"],"source_locator":"Results/Ensemble docking protocols for intrinsically disordered proteins. (paragraph 1); Results/Ensemble docking accurately predicts the relative affinities of small molecules to α-synuclein. (paragraph 1)"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://github.com/gcorso/DiffDock/blob/85c49b60d3e0b0182a59ee43a34a6d7036981284/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-diffdock-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"MIT (upstream repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-base-diffdock-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-base-diffdock-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Represents an intrinsically disordered target as an ensemble rather than a single rigid structure.","source_ids":["ensemble-idp-docking-2025"],"source_locator":"Abstract (paragraph 1); Introduction (paragraph 1)"}],"limitations":[{"text":"The normalised scores have method-specific meanings and are not directly calibrated binding free energies.","source_ids":["ensemble-idp-docking-2025"],"source_locator":"Methods/Comparing docking scores. (paragraph 1); Results/Ensemble docking accurately reproduces IDP ligand binding modes observed in experimentally validated long timescale MD simulations. (paragraph 2)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Ligands and holo protein conformations from the study’s molecular-dynamics ensemble","DiffDock holo","Docked poses and normalised relative docking scores"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["ensemble-idp-docking-2025"],"source_locator":"Methods/DiffDock Ensemble docking. (paragraph 1); Results/Ensemble docking protocols for intrinsically disordered proteins. (paragraph 4)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","DiffDock holo is the comparison-table label; that label does not specify an immutable weight revision.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-77ad27d4098177","kind":"model","name":"scGPT","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["cells-tissues"]},"source_ids":["scelmo-2025"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"scGPT","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"This single-cell foundation-model configuration is a comparator in the scELMo study.","summary_source_ids":["scelmo-2025"],"summary_source_locator":"Results/scELMo for clustering and batch effect correction. (paragraph 1); Methods/Data pre-processing. (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"Pretrained cell/gene representations are evaluated on the paper’s clustering, integration, annotation or perturbation tasks, with the associated evaluation retaining the task-specific procedure.","source_ids":["scelmo-2025"],"source_locator":"Methods/scELMo under the fine-tuning framework. (paragraph 1); Methods/Data pre-processing./Metrics. (paragraph 12)"},{"title":"Underlying method and version boundaries","body":"The official scGPT implementation supplies pretrained checkpoints and separate workflows for embedding extraction, cell annotation, integration and perturbation modelling. The checkpoint and adaptation procedure must be identified separately for each result.","source_ids":["evidence-reported-base-scgpt-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies scGPT: Cell-type annotation. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["scelmo-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-029"}],"facts":[{"label":"Model type","value":"Single-cell foundation model; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-scgpt-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"Pretrained cell/gene representations are evaluated on the paper’s clustering, integration, annotation or perturbation tasks, with the associated evaluation retaining the task-specific procedure.","status":"source_checked","source_ids":["scelmo-2025"],"source_locator":"Methods/scELMo under the fine-tuning framework. (paragraph 1); Methods/Data pre-processing./Metrics. (paragraph 12)"},{"label":"Biological inputs","value":"Single-cell gene-expression measurements","status":"source_checked","source_ids":["scelmo-2025"],"source_locator":"Methods/Data pre-processing./Metrics. (paragraph 15); Methods/Data pre-processing./Metrics. (paragraph 12)"},{"label":"Outputs","value":"Cell embeddings or adapted task predictions","status":"source_checked","source_ids":["scelmo-2025"],"source_locator":"Methods/Problem definition. (paragraph 4); Results/scELMo for clustering and batch effect correction. (paragraph 1)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["scelmo-2025","evidence-reported-base-scgpt-readme-md"],"source_locator":"Methods/Problem definition.; Methods/Method explanation.; Methods/scELMo under the zero-shot learning framework.; Methods/scELMo under the fine-tuning framework.; Methods/Data pre-processing.; Methods/Data pre-processing./Metrics.; Methods/Explanations of baseline models.; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"scGPT is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["scelmo-2025"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"The comparison includes frozen and fine-tuned embedding settings followed by k-nearest-neighbour classification. Table 1 notes that some results were taken from GenePT; the exact origin and adaptation must therefore remain attached to each row.","status":"source_checked","source_ids":["scelmo-2025"],"source_locator":"Methods / Explanations of baseline models; Table 1 caption"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["scelmo-2025","evidence-reported-base-scgpt-readme-md"],"source_locator":"Methods/Problem definition.; Methods/Method explanation.; Methods/scELMo under the zero-shot learning framework.; Methods/scELMo under the fine-tuning framework.; Methods/Data pre-processing.; Methods/Data pre-processing./Metrics.; Methods/Explanations of baseline models.; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://github.com/bowang-lab/scGPT/blob/cebd6fae655b9c585a4807daa3ac31bb764f06b4/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-scgpt-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"MIT (upstream repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-base-scgpt-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-base-scgpt-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Provides a learned-expression reference for comparison with text-derived biological embeddings.","source_ids":["scelmo-2025"],"source_locator":"Methods/scELMo under the fine-tuning framework. (paragraph 4); Discussion (paragraph 2)"}],"limitations":[{"text":"Different downstream tasks and adaptation settings cannot be combined into a single model-quality claim.","source_ids":["scelmo-2025"],"source_locator":"Introduction (paragraph 3); Results/Overview of scELMo. (paragraph 2)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Single-cell gene-expression measurements","scGPT","Cell embeddings or adapted task predictions"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["scelmo-2025"],"source_locator":"Methods/scELMo under the fine-tuning framework. (paragraph 1); Methods/Data pre-processing./Metrics. (paragraph 12)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","scGPT is the comparison-table label; that label does not specify an immutable weight revision.","A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-790768ed581685","kind":"model","name":"kMetaShot","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["microbes-communities"]},"source_ids":["kmetashot-2025"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"kMetaShot","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"kMetaShot assigns taxonomy to metagenome-assembled genomes using k-mer/minimizer counts.","summary_source_ids":["kmetashot-2025"],"summary_source_locator":"Materials and methods (paragraph 1); Materials and methods/kMetaShot reference module (paragraph 7)","sections":[{"title":"How the evaluated method works","body":"The classifier aggregates k-mer/minimizer evidence for assembled microbial genomes, providing a comparatively concise alternative to marker-identification and alignment workflows.","source_ids":["kmetashot-2025"],"source_locator":"Abstract (paragraph 1); Materials and methods (paragraph 1)"},{"title":"What was evaluated","body":"The linked evaluation record identifies kMetaShot: Mock-community MAG taxonomy classification. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["kmetashot-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b3-029"}],"facts":[{"label":"Model type","value":"Study-specific predictive method; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["kmetashot-2025"],"source_locator":"Abstract (paragraph 1); Materials and methods (paragraph 1)"},{"label":"Architecture / procedure","value":"The classifier aggregates k-mer/minimizer evidence for assembled microbial genomes, providing a comparatively concise alternative to marker-identification and alignment workflows.","status":"source_checked","source_ids":["kmetashot-2025"],"source_locator":"Abstract (paragraph 1); Materials and methods (paragraph 1)"},{"label":"Biological inputs","value":"Metagenome-assembled genomes","status":"source_checked","source_ids":["kmetashot-2025"],"source_locator":"Abstract (paragraph 1); Materials and methods/kMetaShot benchmarking on a real dataset (paragraph 1)"},{"label":"Outputs","value":"Taxonomic assignments","status":"source_checked","source_ids":["kmetashot-2025"],"source_locator":"Materials and methods/Taxonomic classification algorithm (paragraph 5); Materials and methods/Taxonomic classification algorithm (paragraph 3)"},{"label":"Parameters","value":"Not applicable: reference-index matching and taxonomic assignment do not use a neural parameter total.","status":"inapplicable","source_ids":["kmetashot-2025"],"source_locator":"Materials and methods/kMetaShot testing on Human Microbiome Project genomes (paragraph 1); Materials and methods/kMetaShot reference module (paragraph 10)"},{"label":"Known versions / configuration","value":"kMetaShot is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["kmetashot-2025"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"A minimizer reference is constructed from the supplied genomes; the documented representation uses 61-mers and 31-base minimizers. Reference taxonomy and database release remain separate provenance.","status":"source_checked","source_ids":["kmetashot-2025"],"source_locator":"Materials and methods (paragraph 1); Materials and methods/kMetaShot reference module (paragraph 7)"},{"label":"Context limits","value":"Not applicable to a pretrained sequence-token window; read/contig lengths and the reference-database or comparison configuration determine the analysed input.","status":"inapplicable","source_ids":["kmetashot-2025"],"source_locator":"Materials and methods/kMetaShot reference module (paragraph 10); Results (paragraph 1)"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/gdefazio/kMetaShot/blob/95dac648aba94d119d20929478f6c3955206f9d4/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-kmetashot-2025-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"GNU GPL version 3 (study repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-kmetashot-2025-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"Not applicable to pretrained neural weights; reference-data reuse terms are separate.","status":"inapplicable","source_ids":["kmetashot-2025"],"source_locator":"Materials and methods/kMetaShot reference module (paragraph 10); Results (paragraph 1)"}],"strengths":[{"text":"Evaluates both simulated and real mock-community genomes.","source_ids":["kmetashot-2025"],"source_locator":"Materials and methods (paragraph 2); Materials and methods/kMetaShot benchmarking on a real dataset (paragraph 1)"}],"limitations":[{"text":"Genome assembly quality and reference-taxonomy coverage are part of the classification problem; performance on MAGs is not interchangeable with raw-read classification.","source_ids":["kmetashot-2025"],"source_locator":"Materials and methods/kMetaShot testing on Human Microbiome Project genomes (paragraph 1); Introduction (paragraph 3)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Metagenome-assembled genomes","kMetaShot","Taxonomic assignments"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["kmetashot-2025"],"source_locator":"Abstract (paragraph 1); Materials and methods (paragraph 1)"},"coverage":"limited","gaps":["kMetaShot is the comparison-table label; that label does not specify an immutable weight revision.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-7ad28cd57f5f5b","kind":"model","name":"ERNIE-RNA + CoBRA","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["rna-transcriptomes"]},"source_ids":["cobra-rna-binding-2026"],"links":[],"attributes":{"entity_level":"method","version":"not stated in table","reported_name":"ERNIE-RNA + CoBRA","historical_missing_metadata":{"checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"CoBRA uses ERNIE-RNA residue embeddings to predict small-molecule-binding nucleotides.","summary_source_ids":["cobra-rna-binding-2026"],"summary_source_locator":"Material and methods/Dataset preparation (paragraph 1); Abstract (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"A pretrained RNA encoder provides contextual nucleotide vectors to a multilayer-perceptron binary classifier. Explicit RNA coordinates or distances are not supplied to the classifier.","source_ids":["cobra-rna-binding-2026"],"source_locator":"Introduction (paragraph 4); Material and methods/Laplacian-based curvature for RNA 3D structure analysis (paragraph 3)"},{"title":"What was evaluated","body":"The linked evaluation record identifies ERNIE-RNA + CoBRA: RNA compound-binding site prediction. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["cobra-rna-binding-2026"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-b2-cobra-rna-binding-2026"}],"facts":[{"label":"Model type","value":"Learned representation pipeline; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["cobra-rna-binding-2026"],"source_locator":"Introduction (paragraph 4); Material and methods/Laplacian-based curvature for RNA 3D structure analysis (paragraph 3)"},{"label":"Architecture / procedure","value":"A pretrained RNA encoder provides contextual nucleotide vectors to a multilayer-perceptron binary classifier. Explicit RNA coordinates or distances are not supplied to the classifier.","status":"source_checked","source_ids":["cobra-rna-binding-2026"],"source_locator":"Introduction (paragraph 4); Material and methods/Laplacian-based curvature for RNA 3D structure analysis (paragraph 3)"},{"label":"Biological inputs","value":"RNA nucleotide sequences","status":"source_checked","source_ids":["cobra-rna-binding-2026"],"source_locator":"Material and methods/Laplacian-based curvature for RNA 3D structure analysis (paragraph 4); Material and methods/Laplacian-based curvature for RNA 3D structure analysis (paragraph 2)"},{"label":"Outputs","value":"Per-nucleotide ligand-binding predictions","status":"source_checked","source_ids":["cobra-rna-binding-2026"],"source_locator":"Material and methods/Dataset preparation (paragraph 1); Material and methods/Laplacian-based curvature for RNA 3D structure analysis (paragraph 4)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["cobra-rna-binding-2026","evidence-reported-cobra-rna-binding-2026-readme-md"],"source_locator":"Material and methods/Dataset preparation; Material and methods/Model architecture; Material and methods/RNA language models; Material and methods/Loss functions; Material and methods/Evaluation metrics; Material and methods/Laplacian-based curvature for RNA 3D structure analysis; Material and methods/Evaluating computational cost with comparison to other machine learning; Results/Comparison with other existing methods on benchmark sets; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"ERNIE-RNA + CoBRA is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["cobra-rna-binding-2026"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"TR60 and HARIBOSS RNA–ligand datasets; the study tests four independent benchmark sets and additional structure-based splits.","status":"source_checked","source_ids":["cobra-rna-binding-2026"],"source_locator":"Material and methods/Dataset preparation (paragraph 3); Material and methods/Dataset preparation (paragraph 2)"},{"label":"Context limits","value":"Inputs are standardised to 161 RNA nucleotides; shorter sequences are padded and chains exceeding that limit are removed.","status":"source_checked","source_ids":["cobra-rna-binding-2026"],"source_locator":"Material and methods/Model architecture (paragraph 1); Material and methods/Dataset preparation (paragraph 2)"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/kucm-lsbi/CoBRA/blob/415fd05cabf990f28a46cc2ba651531a28f7d249/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-cobra-rna-binding-2026-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"GNU GPL version 3 as explicitly declared by the repository (study repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-cobra-rna-binding-2026-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-cobra-rna-binding-2026-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Provides a sequence-based option when an RNA structure is unavailable.","source_ids":["cobra-rna-binding-2026"],"source_locator":"Results/Comparison with other existing methods on benchmark sets (paragraph 4); Discussion (paragraph 3)"}],"limitations":[{"text":"The result measures an ERNIE-RNA-plus-classifier pipeline; structural information still influences how training labels and evaluation partitions are constructed.","source_ids":["cobra-rna-binding-2026"],"source_locator":"Material and methods/RNA language models (paragraph 2); Results/Overall performance comparison and model selection (paragraph 5)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["RNA nucleotide sequences","ERNIE-RNA + CoBRA","Per-nucleotide ligand-binding predictions"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["cobra-rna-binding-2026"],"source_locator":"Introduction (paragraph 4); Material and methods/Laplacian-based curvature for RNA 3D structure analysis (paragraph 3)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","ERNIE-RNA + CoBRA is the comparison-table label; that label does not specify an immutable weight revision.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-7b052acf17b5ba","kind":"model","name":"NCD-gzip","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["microbes-communities"]},"source_ids":["ncd-metagenomics-2026"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"NCD-gzip","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"NCD-gzip classifies genomic sequences using compression-derived distances and nearest neighbours.","summary_source_ids":["ncd-metagenomics-2026"],"summary_source_locator":"Introduction (paragraph 3); Methods/Metagenomic training data subsampling/Genome fragmentation (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"gzip compression estimates normalised compression distance between sequences; a k-nearest-neighbour rule assigns labels from the reference examples.","source_ids":["ncd-metagenomics-2026"],"source_locator":"Introduction (paragraph 3); Methods/Normalized compression distance (paragraph 3)"},{"title":"What was evaluated","body":"The linked evaluation records identify NCD-gzip: CAMI II superkingdom read classification; NCD-gzip: CAMI II phylum read classification. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["ncd-metagenomics-2026"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b3-023, evaluation-lit-b3-024"}],"facts":[{"label":"Model type","value":"Nearest-neighbour procedure; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["ncd-metagenomics-2026"],"source_locator":"Introduction (paragraph 3); Methods/Normalized compression distance (paragraph 3)"},{"label":"Architecture / procedure","value":"gzip compression estimates normalised compression distance between sequences; a k-nearest-neighbour rule assigns labels from the reference examples.","status":"source_checked","source_ids":["ncd-metagenomics-2026"],"source_locator":"Introduction (paragraph 3); Methods/Normalized compression distance (paragraph 3)"},{"label":"Biological inputs","value":"Nucleotide sequences and a labelled reference set","status":"source_checked","source_ids":["ncd-metagenomics-2026"],"source_locator":"Related Works/HyenaDNA (paragraph 1); Introduction (paragraph 3)"},{"label":"Outputs","value":"Gene labels or taxonomic classes, depending on the experiment","status":"source_checked","source_ids":["ncd-metagenomics-2026"],"source_locator":"Results/Prokaryotic gene classification/Test (paragraph 2); Results/Prokaryotic gene classification/Gene/taxa out (paragraph 1)"},{"label":"Parameters","value":"Not applicable to the compression/nearest-neighbour method.","status":"inapplicable","source_ids":["ncd-metagenomics-2026"],"source_locator":"Introduction (paragraph 3); Methods/Normalized compression distance (paragraph 2)"},{"label":"Known versions / configuration","value":"NCD-gzip is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["ncd-metagenomics-2026"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"No learned foundation-model weights; the labelled comparison set acts as the reference.","status":"source_checked","source_ids":["ncd-metagenomics-2026"],"source_locator":"Methods/Datasets/Metagenomic reads (paragraph 1); Results/CAMI dataset (paragraph 3)"},{"label":"Context limits","value":"Not applicable to a pretrained sequence-token window; read/contig lengths and the reference-database or comparison configuration determine the analysed input.","status":"inapplicable","source_ids":["ncd-metagenomics-2026"],"source_locator":"Methods/Metagenomic training data subsampling/Genome fragmentation (paragraph 1); Methods/Metagenomic training data subsampling (paragraph 1)"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/ghproducts/genomics-ncd-gzip/blob/d5bb37be194a2716bc65543c65a26a65eebc1849/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-ncd-metagenomics-2026-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"CC0 1.0 (study repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-ncd-metagenomics-2026-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"Not applicable: no pretrained weight file is required.","status":"inapplicable","source_ids":["ncd-metagenomics-2026"],"source_locator":"Results/Timing results (paragraph 2); Methods/Normalized compression distance (paragraph 3)"}],"strengths":[{"text":"Provides a non-neural reference based on shared compressibility rather than a pretrained embedding.","source_ids":["ncd-metagenomics-2026"],"source_locator":"Methods/Metagenomic training data subsampling/Genome fragmentation (paragraph 2); Results/CAMI dataset (paragraph 3)"}],"limitations":[{"text":"The paper reports O(MN) comparison scaling with reference and query counts, limiting large-database use.","source_ids":["ncd-metagenomics-2026"],"source_locator":"Methods/Normalized compression distance/NCD timing (paragraph 1); Introduction (paragraph 3)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Nucleotide sequences and a labelled reference set","NCD-gzip","Gene labels or taxonomic classes, depending on the experiment"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["ncd-metagenomics-2026"],"source_locator":"Introduction (paragraph 3); Methods/Normalized compression distance (paragraph 3)"},"coverage":"limited","gaps":["NCD-gzip is the comparison-table label; that label does not specify an immutable weight revision.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-7c595040de69bc","kind":"model","name":"GenePT-w","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["cells-tissues"]},"source_ids":["genept-2024"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"GenePT-w","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"GenePT-w represents cells by expression-weighted averages of gene embeddings derived from gene-description text.","summary_source_ids":["genept-2024"],"summary_source_locator":"Introduction (paragraph 4); Methods/Data Collection and Transformation: (paragraph 2)","sections":[{"title":"How the evaluated method works","body":"Gene descriptions from NCBI are embedded using GPT-3.5. The weighted variant averages gene vectors according to each cell’s expression values; it differs from the sentence-embedding variant.","source_ids":["genept-2024"],"source_locator":"Methods/Data Collection and Transformation: (paragraph 4); Related Work/Using language models for cell biology (paragraph 1)"},{"title":"What was evaluated","body":"The linked evaluation record identifies GenePT-w: Cell-type structure in frozen embeddings. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["genept-2024"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b3-011"}],"facts":[{"label":"Model type","value":"Study-specific predictive method; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["genept-2024"],"source_locator":"Methods/Data Collection and Transformation: (paragraph 4); Related Work/Using language models for cell biology (paragraph 1)"},{"label":"Architecture / procedure","value":"Gene descriptions from NCBI are embedded using GPT-3.5. The weighted variant averages gene vectors according to each cell’s expression values; it differs from the sentence-embedding variant.","status":"source_checked","source_ids":["genept-2024"],"source_locator":"Methods/Data Collection and Transformation: (paragraph 4); Related Work/Using language models for cell biology (paragraph 1)"},{"label":"Biological inputs","value":"Gene-description embeddings and single-cell expression values","status":"source_checked","source_ids":["genept-2024"],"source_locator":"Related Work/Deciphering natural language embeddings (paragraph 4); Related Work/Deciphering natural language embeddings (paragraph 5)"},{"label":"Outputs","value":"Cell embeddings for downstream cell-analysis tasks","status":"source_checked","source_ids":["genept-2024"],"source_locator":"Related Work/Deciphering natural language embeddings (paragraph 4); Methods/Downstream gene-level and cell-level applications: (paragraph 1)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["genept-2024","evidence-reported-genept-2024-readme-md"],"source_locator":"Methods/Data Collection and Transformation:; Methods/Downstream gene-level and cell-level applications:; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"GenePT-w is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["genept-2024"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"The biological representation uses NCBI gene descriptions; the method does not train a new foundation model on a large cell-expression corpus.","status":"source_checked","source_ids":["genept-2024"],"source_locator":"Abstract (paragraph 1); Introduction (paragraph 4)"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["genept-2024","evidence-reported-genept-2024-readme-md"],"source_locator":"Methods/Data Collection and Transformation:; Methods/Downstream gene-level and cell-level applications:; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/yiqunchen/GenePT/blob/3602699e7425a7be577771f8f07e218db6c79b9f/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-genept-2024-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","status":"unreported","source_ids":["evidence-reported-genept-2024-readme-md"],"source_locator":"README.md and repository-root licence-file search"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-genept-2024-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Provides a relatively simple literature-derived representation baseline for single-cell tasks.","source_ids":["genept-2024"],"source_locator":"Results/GenePT learns representations that reflect cell biology (paragraph 2); Appendix/Appendix C Cell type annotation results (paragraph 1)"}],"limitations":[{"text":"Depends on coverage and content of gene annotations and a proprietary text-embedding model; it is not a direct measurement of molecular mechanism.","source_ids":["genept-2024"],"source_locator":"Appendix/Appendix B Additional results for the gene level functionality and property predictions/B.3 Addressing potential information leakage/B.5 Experiments with tissue-dependent gene emebddings (paragraph 4); Methods/Data Collection and Transformation: (paragraph 1)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Gene-description embeddings and single-cell expression values","GenePT-w","Cell embeddings for downstream cell-analysis tasks"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["genept-2024"],"source_locator":"Methods/Data Collection and Transformation: (paragraph 4); Related Work/Using language models for cell biology (paragraph 1)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","GenePT-w is the comparison-table label; that label does not specify an immutable weight revision.","A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-7cf2f9951e1dba","kind":"model","name":"scaLR","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["cells-tissues"]},"source_ids":["scalr-2025"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"scaLR","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"scaLR is a low-resource single-cell annotation pipeline with feature selection and neural classification.","summary_source_ids":["scalr-2025"],"summary_source_locator":"Abstract (paragraph 1); Results/Performance evaluation of scaLR using all genes as features (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"Models are trained on feature subsets to estimate feature importance, then the final model uses the top-K selected features. Batched samples and CPU-parallel training reduce memory requirements.","source_ids":["scalr-2025"],"source_locator":"Abstract (paragraph 1); Methods/The scaLR platform/Feature extraction (paragraph 1)"},{"title":"What was evaluated","body":"The linked evaluation record identifies scaLR: PBMC cell-type classification. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["scalr-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b3-015"}],"facts":[{"label":"Model type","value":"Study-specific predictive method; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["scalr-2025"],"source_locator":"Abstract (paragraph 1); Methods/The scaLR platform/Feature extraction (paragraph 1)"},{"label":"Architecture / procedure","value":"Models are trained on feature subsets to estimate feature importance, then the final model uses the top-K selected features. Batched samples and CPU-parallel training reduce memory requirements.","status":"source_checked","source_ids":["scalr-2025"],"source_locator":"Abstract (paragraph 1); Methods/The scaLR platform/Feature extraction (paragraph 1)"},{"label":"Biological inputs","value":"Single-cell gene-expression data","status":"source_checked","source_ids":["scalr-2025"],"source_locator":"Introduction (paragraph 1); Introduction (paragraph 2)"},{"label":"Outputs","value":"Selected features, cell-type predictions and downstream analyses","status":"source_checked","source_ids":["scalr-2025"],"source_locator":"Results/Comparison of top-K differential gene expression features between full and test set samples (paragraph 1); Abstract (paragraph 1)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["scalr-2025","evidence-reported-scalr-2025-readme-md"],"source_locator":"Methods/Single-cell RNA-Seq data selection and download; Methods/The scaLR platform; Methods/The scaLR platform/Data processing; Methods/The scaLR platform/Feature extraction; Methods/The scaLR platform/Training; Methods/The scaLR platform/Evaluation and downstream analysis; Methods/scaLR performance comparison with different pipelines; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"scaLR is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["scalr-2025"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"Feature-subset linear models rank genes using mean absolute class weights; a final DNN is fitted on the selected top-K genes, with checkpoint selection using validation data.","status":"source_checked","source_ids":["scalr-2025"],"source_locator":"Methods/The scaLR platform/Feature extraction (paragraph 1); Methods/The scaLR platform/Training (paragraph 1)"},{"label":"Context limits","value":"A selected gene-feature matrix rather than a sequence-token context window; the chosen top-K is part of the fitted configuration.","status":"source_checked","source_ids":["scalr-2025"],"source_locator":"Results/Comparison of top-K differential gene expression features between full and test set samples (paragraph 1); Methods/The scaLR platform/Feature extraction (paragraph 1)"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/infocusp/scaLR/blob/b5f72ce8f9bd25cb90f4b6b3112a03288c788f2e/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-scalr-2025-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"GNU GPL version 3 (study repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-scalr-2025-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-scalr-2025-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Makes feature selection and memory-aware processing part of the workflow.","source_ids":["scalr-2025"],"source_locator":"Methods/The scaLR platform (paragraph 1); Abstract (paragraph 1)"}],"limitations":[{"text":"The selected features and final training data are part of the fitted system; resource claims are tied to the study’s hardware and workloads.","source_ids":["scalr-2025"],"source_locator":"Abstract (paragraph 1); Methods/scaLR performance comparison with different pipelines (paragraph 4)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Single-cell gene-expression data","scaLR","Selected features, cell-type predictions and downstream analyses"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["scalr-2025"],"source_locator":"Abstract (paragraph 1); Methods/The scaLR platform/Feature extraction (paragraph 1)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","scaLR is the comparison-table label; that label does not specify an immutable weight revision.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-7dd5992188a868","kind":"model","name":"PST","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["proteins-complexes"]},"source_ids":["pst-2025"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"PST","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"PST adds structural adapters to a pretrained protein sequence transformer.","summary_source_ids":["pst-2025"],"summary_source_locator":"1 Introduction (paragraph 4); 5 Discussion (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"Specialised structural adapters modify self-attention to combine sequence and structure embeddings, while retaining the pretrained language-model backbone.","source_ids":["pst-2025"],"source_locator":"1 Introduction (paragraph 4); 3 Methods/3.2 Protein structure transformer/3.2.2 Protein structure transformer construction (paragraph 1)"},{"title":"What was evaluated","body":"The linked evaluation record identifies PST: Zero-shot variant effect prediction. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["pst-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-023"}],"facts":[{"label":"Model type","value":"Study-specific predictive method; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["pst-2025"],"source_locator":"1 Introduction (paragraph 4); 3 Methods/3.2 Protein structure transformer/3.2.2 Protein structure transformer construction (paragraph 1)"},{"label":"Architecture / procedure","value":"Specialised structural adapters modify self-attention to combine sequence and structure embeddings, while retaining the pretrained language-model backbone.","status":"source_checked","source_ids":["pst-2025"],"source_locator":"1 Introduction (paragraph 4); 3 Methods/3.2 Protein structure transformer/3.2.2 Protein structure transformer construction (paragraph 1)"},{"label":"Biological inputs","value":"Protein sequence and structural information","status":"source_checked","source_ids":["pst-2025"],"source_locator":"1 Introduction (paragraph 7); 3 Methods (paragraph 1)"},{"label":"Outputs","value":"Joint sequence/structure representations for downstream tasks","status":"source_checked","source_ids":["pst-2025"],"source_locator":"4 Experiments/4.4 Ablation studies/4.4.3 Pretraining only structure extractors almost matches full-model pretraining performance (paragraph 2); 1 Introduction (paragraph 5)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["pst-2025","evidence-reported-pst-2025-readme-md"],"source_locator":"3 Methods; 3 Methods/3.1 Evolutionary scale modeling; 3 Methods/3.1 Evolutionary scale modeling/3.1.1 ESM-2 model architecture; 3 Methods/3.2 Protein structure transformer; 3 Methods/3.2 Protein structure transformer/3.2.1 Protein graph representation; 3 Methods/3.2 Protein structure transformer/3.2.2 Protein structure transformer construction; 3 Methods/3.2 Protein structure transformer/3.2.3 Pretraining the PST; 4 Experiments/4.1 Experimental setup/4.1.4 Pretraining; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"PST is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["pst-2025"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"542,000 protein structures with a masked-language objective for the added structural integration.","status":"source_checked","source_ids":["pst-2025"],"source_locator":"3 Methods/3.2 Protein structure transformer/3.2.3 Pretraining the PST (paragraph 1); 3 Methods/3.1 Evolutionary scale modeling (paragraph 2)"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["pst-2025","evidence-reported-pst-2025-readme-md"],"source_locator":"3 Methods; 3 Methods/3.1 Evolutionary scale modeling; 3 Methods/3.1 Evolutionary scale modeling/3.1.1 ESM-2 model architecture; 3 Methods/3.2 Protein structure transformer; 3 Methods/3.2 Protein structure transformer/3.2.1 Protein graph representation; 3 Methods/3.2 Protein structure transformer/3.2.2 Protein structure transformer construction; 3 Methods/3.2 Protein structure transformer/3.2.3 Pretraining the PST; 4 Experiments/4.1 Experimental setup/4.1.4 Pretraining; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/BorgwardtLab/PST/blob/57d9dcd8200900504a19de459450e137867262d7/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-pst-2025-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"BSD 3-Clause (study repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-pst-2025-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-pst-2025-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Tests a lightweight way to incorporate structure without training an entirely new large protein model.","source_ids":["pst-2025"],"source_locator":"2 Related work (paragraph 1); 4 Experiments/4.4 Ablation studies/4.4.1 Simpler structural inputs yield better performance (paragraph 1)"}],"limitations":[{"text":"The extra structural inputs and adaptation data must be preserved when comparing with the sequence-only ESM-2 baseline.","source_ids":["pst-2025"],"source_locator":"3 Methods (paragraph 1); 5 Discussion (paragraph 3)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Protein sequence and structural information","PST","Joint sequence/structure representations for downstream tasks"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["pst-2025"],"source_locator":"1 Introduction (paragraph 4); 3 Methods/3.2 Protein structure transformer/3.2.2 Protein structure transformer construction (paragraph 1)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","PST is the comparison-table label; that label does not specify an immutable weight revision.","A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-7f1165b35f10e2","kind":"model","name":"DNABERT-2","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["microbes-communities"]},"source_ids":["genomeocean-2025"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"DNABERT-2","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"DNABERT-2 is fine-tuned to distinguish natural from GenomeOcean-generated DNA sequences.","summary_source_ids":["genomeocean-2025"],"summary_source_locator":"Results/Model Safety (paragraph 1); Results/GenomeOcean Learns Protein-coding Principles from DNA Alone (paragraph 4)","sections":[{"title":"How the evaluated method works","body":"The pretrained DNA transformer is adapted with a binary classifier using the study’s natural/artificial sequence labels.","source_ids":["genomeocean-2025"],"source_locator":"Results/Model Safety (paragraph 1); Results/Model Safety (paragraph 2)"},{"title":"Underlying method and version boundaries","body":"DNABERT-2 replaces overlapping k-mer tokens with byte-pair encoding and uses ALiBi positional biases. The official 117M model produces 768-dimensional token representations; downstream classifiers and pooling choices are separate configuration details.","source_ids":["evidence-reported-base-dnabert2-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies DNABERT-2: Natural vs artificial microbial genome sequence. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["genomeocean-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b3-028"}],"facts":[{"label":"Model type","value":"DNA sequence transformer; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-dnabert2-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"The pretrained DNA transformer is adapted with a binary classifier using the study’s natural/artificial sequence labels.","status":"source_checked","source_ids":["genomeocean-2025"],"source_locator":"Results/Model Safety (paragraph 1); Results/Model Safety (paragraph 2)"},{"label":"Biological inputs","value":"2-kbp natural or generated DNA sequences","status":"source_checked","source_ids":["genomeocean-2025"],"source_locator":"Results/GenomeOcean Learns Protein-coding Principles from DNA Alone (paragraph 4); Methods/Training/Evaluation Datasets/Generated Sequence Discrimination (paragraph 1)"},{"label":"Outputs","value":"Natural-versus-artificial sequence predictions","status":"source_checked","source_ids":["genomeocean-2025"],"source_locator":"Methods/Training/Evaluation Datasets/Generated Sequence Discrimination (paragraph 1); Results/Model Safety (paragraph 1)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["genomeocean-2025","evidence-reported-base-dnabert2-readme-md"],"source_locator":"Results/Pre-training; Methods/Model and Training/Tokenization; Methods/Model and Training/Architecture and Pre-training; Methods/Model and Training/Fine-tune GenomeOcean as Biosynthetic Gene Clusters Foundation Model (bgcFM); Methods/Training/Evaluation Datasets/Assembled Metagenome Datasets; Methods/Training/Evaluation Datasets/Evaluation of Dataset Complexity; Methods/Training/Evaluation Datasets/Biosynthetic Gene Cluster (BGC) Collection; Methods/Training/Evaluation Datasets/ZymoBiomics Microbial Community Standard Datasets; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"DNABERT-2 is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["genomeocean-2025"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"Balanced 18,000/2,000/20,000 train/validation/test sequences.","status":"source_checked","source_ids":["genomeocean-2025"],"source_locator":"Methods/Training/Evaluation Datasets/Generated Sequence Discrimination (paragraph 1); Data & Code/Metagenome Raw Reads Datasets Used in Assembly (paragraph 1)"},{"label":"Context limits","value":"2,000-bp benchmark windows.","status":"source_checked","source_ids":["genomeocean-2025"],"source_locator":"Abstract (paragraph 1); Table T4 (paragraph 1)"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://github.com/MAGICS-LAB/DNABERT_2/blob/f25bed9ee20db966dff39e5c1571249d04e36404/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-dnabert2-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"Apache 2.0 (upstream repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-base-dnabert2-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-base-dnabert2-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Tests whether generated DNA can be distinguished by a separate pretrained genome model.","source_ids":["genomeocean-2025"],"source_locator":"Results/Model Safety (paragraph 1); Methods/Training/Evaluation Datasets/Generated Sequence Discrimination (paragraph 1)"}],"limitations":[{"text":"A detector for this generator and data distribution is not a universal provenance or sequence-safety detector.","source_ids":["genomeocean-2025"],"source_locator":"Results/Model Safety (paragraph 2); Discussion (paragraph 2)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["2-kbp natural or generated DNA sequences","DNABERT-2","Natural-versus-artificial sequence predictions"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["genomeocean-2025"],"source_locator":"Results/Model Safety (paragraph 1); Results/Model Safety (paragraph 2)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","DNABERT-2 is the comparison-table label; that label does not specify an immutable weight revision.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-7f5b8234967c54","kind":"model","name":"2OMe-LM","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["rna-transcriptomes"]},"source_ids":["2ome-lm-2025"],"links":[],"attributes":{"entity_level":"method","version":"not stated in table","reported_name":"2OMe-LM","historical_missing_metadata":{"checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"2OMe-LM predicts whether the central nucleotide of an RNA window carries a 2′-O-methylation modification.","summary_source_ids":["2ome-lm-2025"],"summary_source_locator":"3 Results/3.4 Motif analysis (paragraph 2); 2 Materials and methods/2.2 Model architecture/2.2.2 Word2vec embedding (paragraph 2)","sections":[{"title":"How the evaluated method works","body":"SpliceBERT embeddings pass through a dimensionality-reduction layer. A second branch encodes k-mers with word2vec and a bidirectional LSTM. Fused features enter an attention block and three fully connected layers for site classification.","source_ids":["2ome-lm-2025"],"source_locator":"2 Materials and methods/2.2 Model architecture (paragraph 1); 2 Materials and methods/2.2 Model architecture/2.2.3 Feature fusion and prediction (paragraph 3)"},{"title":"What was evaluated","body":"The linked evaluation record identifies 2OMe-LM: human RNA 2-prime-O-methylation site prediction. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["2ome-lm-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-b2-2ome-lm-2025"}],"facts":[{"label":"Model type","value":"Recurrent neural network; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["2ome-lm-2025"],"source_locator":"2 Materials and methods/2.2 Model architecture (paragraph 1); 2 Materials and methods/2.2 Model architecture/2.2.3 Feature fusion and prediction (paragraph 3)"},{"label":"Architecture / procedure","value":"SpliceBERT embeddings pass through a dimensionality-reduction layer. A second branch encodes k-mers with word2vec and a bidirectional LSTM. Fused features enter an attention block and three fully connected layers for site classification.","status":"source_checked","source_ids":["2ome-lm-2025"],"source_locator":"2 Materials and methods/2.2 Model architecture (paragraph 1); 2 Materials and methods/2.2 Model architecture/2.2.3 Feature fusion and prediction (paragraph 3)"},{"label":"Biological inputs","value":"RNA sequence windows centred on candidate modification sites","status":"source_checked","source_ids":["2ome-lm-2025"],"source_locator":"3 Results/3.4 Motif analysis (paragraph 2); 4 Conclusion (paragraph 1)"},{"label":"Outputs","value":"Probability that the central site is 2′-O-methylated","status":"source_checked","source_ids":["2ome-lm-2025"],"source_locator":"2 Materials and methods/2.2 Model architecture/2.2.3 Feature fusion and prediction (paragraph 3); 3 Results/3.5 Case study (paragraph 1)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["2ome-lm-2025","evidence-reported-2ome-lm-2025-readme-md"],"source_locator":"2 Materials and methods/2.1 Datasets; 2 Materials and methods/2.2 Model architecture; 2 Materials and methods/2.2 Model architecture/2.2.1 Pretrained RNA language model; 2 Materials and methods/2.2 Model architecture/2.2.2 Word2vec embedding; 2 Materials and methods/2.2 Model architecture/2.2.3 Feature fusion and prediction; 2 Materials and methods/2.3 Deep learning baseline models; 2 Materials and methods/2.4 Evaluation metrics; 2 Materials and methods/2.5 Implementation details; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"2OMe-LM is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["2ome-lm-2025"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"8,037 positive and 8,037 negative examples, with an 80:20 training/test partition; SpliceBERT provides separately pretrained sequence features.","status":"source_checked","source_ids":["2ome-lm-2025"],"source_locator":"2 Materials and methods/2.1 Datasets (paragraph 2); 2 Materials and methods/2.1 Datasets (paragraph 1)"},{"label":"Context limits","value":"SpliceBERT supports up to 1,024 nucleotides; that backbone limit is not an independently verified limit of the complete predictor.","status":"source_checked","source_ids":["2ome-lm-2025"],"source_locator":"3 Results/3.3 Effectiveness of pre-trained RNA language model (paragraph 1); Table btaf417-T2 (paragraph 1)"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/CSUBioGroup/2OMe-LM/blob/2e22439723777b5bacdce72cdcd7cfbde9e88cd1/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-2ome-lm-2025-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","status":"unreported","source_ids":["evidence-reported-2ome-lm-2025-readme-md"],"source_locator":"README.md and repository-root licence-file search"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-2ome-lm-2025-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Combines pretrained RNA representations with a separately learned k-mer branch and provides attention-based motif analysis.","source_ids":["2ome-lm-2025"],"source_locator":"Abstract (paragraph 1); 3 Results/3.4 Motif analysis (paragraph 1)"}],"limitations":[{"text":"The reported human-RNA classification task and sampling scheme do not establish performance on every RNA species or modification type.","source_ids":["2ome-lm-2025"],"source_locator":"3 Results/3.6 Cross-nucleotide evaluation (paragraph 1); 3 Results/3.4 Motif analysis (paragraph 2)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["RNA sequence windows centred on candidate modification sites","2OMe-LM","Probability that the central site is 2′-O-methylated"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["2ome-lm-2025"],"source_locator":"2 Materials and methods/2.2 Model architecture (paragraph 1); 2 Materials and methods/2.2 Model architecture/2.2.3 Feature fusion and prediction (paragraph 3)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","2OMe-LM is the comparison-table label; that label does not specify an immutable weight revision.","No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-7f6ffd9e2a08be","kind":"model","name":"DiffDock","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["molecular-interactions"]},"source_ids":["boltz-stereochemistry-2025"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"DiffDock","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"This protein–ligand pose predictor is evaluated for stereochemical validity as well as pose accuracy.","summary_source_ids":["boltz-stereochemistry-2025"],"summary_source_locator":"Results/Benchmark of Protein–Ligand\nComplex Structure Prediction (paragraph 6); Methods/Benchmark Dataset (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"The study compares generated ligand poses and protein–ligand complexes under a shared geometry assessment. The unrestrained baseline is distinct from the proposed restraint-guided Boltz inference.","source_ids":["boltz-stereochemistry-2025"],"source_locator":"Results/Case Studies for Bond and Angle Geometries of Ligands (paragraph 3); Results/Structure Prediction Using\nRestraint-Guided Inference (paragraph 1)"},{"title":"Underlying method and version boundaries","body":"DiffDock is a molecular-docking implementation that produces ligand poses and confidence estimates. Its confidence values and predicted coordinates are different outputs from an experimentally calibrated binding-affinity measurement.","source_ids":["evidence-reported-base-diffdock-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies DiffDock: Protein–ligand pose prediction. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["boltz-stereochemistry-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-046"}],"facts":[{"label":"Model type","value":"Molecular docking model; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-diffdock-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"The study compares generated ligand poses and protein–ligand complexes under a shared geometry assessment. The unrestrained baseline is distinct from the proposed restraint-guided Boltz inference.","status":"source_checked","source_ids":["boltz-stereochemistry-2025"],"source_locator":"Results/Case Studies for Bond and Angle Geometries of Ligands (paragraph 3); Results/Structure Prediction Using\nRestraint-Guided Inference (paragraph 1)"},{"label":"Biological inputs","value":"Protein and ligand information under the method-specific docking or co-folding protocol","status":"source_checked","source_ids":["boltz-stereochemistry-2025"],"source_locator":"Results/Benchmark of Protein–Ligand\nComplex Structure Prediction (paragraph 1); Results/Case Studies for Bond and Angle Geometries of Ligands (paragraph 1)"},{"label":"Outputs","value":"Predicted protein–ligand binding poses","status":"source_checked","source_ids":["boltz-stereochemistry-2025"],"source_locator":"Results/Case Studies for Bond and Angle Geometries of Ligands (paragraph 2); Methods/Implementation of the Restraints (paragraph 1)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["boltz-stereochemistry-2025","evidence-reported-base-diffdock-readme-md"],"source_locator":"Methods/Benchmark Dataset; Methods/Evaluation Metrics; Methods/Improving Geometry through Restraints; Methods/Implementation of the Restraints; Methods/Multiple Sequence Alignment and Structure Prediction; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"DiffDock is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["boltz-stereochemistry-2025"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"The evaluation separates structures before and after training-date cutoffs and adds chemical/protein similarity filters.","status":"source_checked","source_ids":["boltz-stereochemistry-2025"],"source_locator":"Methods/Benchmark Dataset (paragraph 3); Methods/Benchmark Dataset (paragraph 4)"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["boltz-stereochemistry-2025","evidence-reported-base-diffdock-readme-md"],"source_locator":"Methods/Benchmark Dataset; Methods/Evaluation Metrics; Methods/Improving Geometry through Restraints; Methods/Implementation of the Restraints; Methods/Multiple Sequence Alignment and Structure Prediction; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://github.com/gcorso/DiffDock/blob/85c49b60d3e0b0182a59ee43a34a6d7036981284/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-diffdock-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"MIT (upstream repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-base-diffdock-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-base-diffdock-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Reports chirality, bond geometry and steric clashes alongside RMSD.","source_ids":["boltz-stereochemistry-2025"],"source_locator":"Results/Benchmark of Protein–Ligand\nComplex Structure Prediction (paragraph 6); Results/Structure Prediction Using\nRestraint-Guided Inference (paragraph 4)"}],"limitations":[{"text":"Pose accuracy deteriorates on filtered post-cutoff data, and low RMSD alone does not imply physically valid ligand geometry.","source_ids":["boltz-stereochemistry-2025"],"source_locator":"Results/Benchmark of Protein–Ligand\nComplex Structure Prediction (paragraph 6); Methods/Evaluation Metrics (paragraph 3)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Protein and ligand information under the method-specific docking or co-folding protocol","DiffDock","Predicted protein–ligand binding poses"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["boltz-stereochemistry-2025"],"source_locator":"Results/Case Studies for Bond and Angle Geometries of Ligands (paragraph 3); Results/Structure Prediction Using\nRestraint-Guided Inference (paragraph 1)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","DiffDock is the comparison-table label; that label does not specify an immutable weight revision.","A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-808b23c65fbc89","kind":"model","name":"Cell-DINO ViT-L","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["cells-tissues"]},"source_ids":["cell-dino-2025"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"Cell-DINO ViT-L","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Cell-DINO ViT-L learns representations of fluorescent cell images, including protein-localisation information.","summary_source_ids":["cell-dino-2025"],"summary_source_locator":"Results/Cell-DINO outperforms standard supervised ViT (paragraph 3); Results/Cell-DINO outperforms alternative self-supervised strategies (paragraph 3)","sections":[{"title":"How the evaluated method works","body":"DINOv2 self-supervised learning trains a vision transformer on cell images. Frozen embeddings support downstream classifiers; the protein-localisation experiment includes a separately trained two-layer classifier.","source_ids":["cell-dino-2025"],"source_locator":"Results/Cell-DINO is competitive against highly tuned models (paragraph 1); Results/Cell-DINO outperforms standard supervised ViT (paragraph 3)"},{"title":"What was evaluated","body":"The linked evaluation record identifies Cell-DINO ViT-L: protein localization classification. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["cell-dino-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b4-015"}],"facts":[{"label":"Model type","value":"Vision transformer; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["cell-dino-2025"],"source_locator":"Results/Cell-DINO is competitive against highly tuned models (paragraph 1); Results/Cell-DINO outperforms standard supervised ViT (paragraph 3)"},{"label":"Architecture / procedure","value":"DINOv2 self-supervised learning trains a vision transformer on cell images. Frozen embeddings support downstream classifiers; the protein-localisation experiment includes a separately trained two-layer classifier.","status":"source_checked","source_ids":["cell-dino-2025"],"source_locator":"Results/Cell-DINO is competitive against highly tuned models (paragraph 1); Results/Cell-DINO outperforms standard supervised ViT (paragraph 3)"},{"label":"Biological inputs","value":"Fluorescence-microscopy cell images","status":"source_checked","source_ids":["cell-dino-2025"],"source_locator":"Methods/Cell-DINO algorithm (paragraph 1); Methods/Datasets (paragraph 5)"},{"label":"Outputs","value":"Cell-image embeddings and downstream protein-localisation predictions","status":"source_checked","source_ids":["cell-dino-2025"],"source_locator":"Results/Cell-DINO outperforms standard supervised ViT (paragraph 3); Results/Cell-DINO outperforms standard supervised ViT (paragraph 1)"},{"label":"Parameters","value":"The HPA-SC downstream two-layer classifier has 1.2 million supervised parameters; this is not the total of the frozen ViT-L encoder plus classifier.","status":"source_checked","source_ids":["cell-dino-2025"],"source_locator":"Results/Cell-DINO reduces the dependency on manual annotations (paragraph 1); Results/Cell-DINO is competitive against highly tuned models (paragraph 1)"},{"label":"Known versions / configuration","value":"Cell-DINO ViT-L is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["cell-dino-2025"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"Separate ViT-L pretraining datasets: HPA-FoV about 200,000 images, HPA-SC about 500,000 images, and combined Cell Painting about five million images. The protein-localisation classifier is trained separately.","status":"source_checked","source_ids":["cell-dino-2025"],"source_locator":"Methods/Computational resources used for training Cell-DINO models (paragraph 1); Methods/Datasets (paragraph 4)"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["cell-dino-2025","evidence-reported-cell-dino-2025-docs-readme-cell-dino-md"],"source_locator":"Methods/Datasets; Methods/Cell-DINO algorithm; Methods/Vision Transformer (ViT) architecture; Methods/Supervised ViTs; Methods/Training and evaluation protocol on HPA datasets; Methods/Kaggle evaluation protocol; Methods/Cell Painting evaluation protocol; Methods/Baselines; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); docs/README_CELL_DINO.md at pinned repository revision"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/facebookresearch/dinov2/blob/7764ea0f912e53c92e82eb78a2a1631e92725fc8/docs/README_CELL_DINO.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-cell-dino-2025-docs-readme-cell-dino-md"],"source_locator":"docs/README_CELL_DINO.md; installation, model download and usage instructions"},{"label":"Code licence","value":"Conflicting official statements: the Cell-DINO README says CC BY-NC, but its linked code licence is headed Creative Commons Attribution 4.0 International. Code reuse terms require author clarification; the DINOv2 root licence is not substituted.","status":"unreported","source_ids":["evidence-reported-cell-dino-2025-docs-readme-cell-dino-md","evidence-reported-cell-dino-2025-license-cell-dino-code"],"source_locator":"docs/README_CELL_DINO.md / License; linked LICENSE_CELL_DINO_CODE heading and grant"},{"label":"Weights licence","value":"FAIR Noncommercial Research License, version 1 (18 August 2025), for the Cell-DINO model materials.","status":"source_checked","source_ids":["evidence-reported-cell-dino-2025-license-cell-dino-models"],"source_locator":"LICENSE_CELL_DINO_MODELS; title, definitions and Section 1"}],"strengths":[{"text":"Self-supervised representations can support protein-localisation classification when annotations are limited.","source_ids":["cell-dino-2025"],"source_locator":"Abstract (paragraph 1); Methods/Supervised ViTs (paragraph 1)"}],"limitations":[{"text":"This record is included for molecular protein-localisation evaluation; performance on these microscopy datasets is not evidence for general clinical imaging.","source_ids":["cell-dino-2025"],"source_locator":"Results/Cell-DINO outperforms other pre-trained models (paragraph 3); Results/Cell-DINO outperforms alternative self-supervised strategies (paragraph 3)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Fluorescence-microscopy cell images","Cell-DINO ViT-L","Cell-image embeddings and downstream protein-localisation predictions"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["cell-dino-2025"],"source_locator":"Results/Cell-DINO is competitive against highly tuned models (paragraph 1); Results/Cell-DINO outperforms standard supervised ViT (paragraph 3)"},"coverage":"limited","gaps":["Cell-DINO ViT-L is the comparison-table label; that label does not specify an immutable weight revision.","A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","Conflicting official statements: the Cell-DINO README says CC BY-NC, but its linked code licence is headed Creative Commons Attribution 4.0 International. Code reuse terms require author clarification; the DINOv2 root licence is not substituted.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-8100b3de6c7811","kind":"model","name":"PMF (LASSO)","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["molecular-interactions"]},"source_ids":["fingerprint-scoring-2022"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"PMF (LASSO)","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"The PMF-only LASSO model is a sparse linear binding-affinity baseline in the fingerprint-scoring study.","summary_source_ids":["fingerprint-scoring-2022"],"summary_source_locator":"Conclusions (paragraph 3); Methods/Machine Learning with LASSO and LightGBM (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"LASSO regression fits binding affinities from potential-of-mean-force descriptors alone, without the ECFP and protein-fingerprint features used by the combined LightGBM configuration.","source_ids":["fingerprint-scoring-2022"],"source_locator":"Methods/Computational Details (paragraph 2); Methods/Computational Details (paragraph 4)"},{"title":"What was evaluated","body":"The linked evaluation record identifies PMF (LASSO): Protein–ligand binding energy prediction. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["fingerprint-scoring-2022"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b3-051"}],"facts":[{"label":"Model type","value":"Gradient-boosted-tree pipeline; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["fingerprint-scoring-2022"],"source_locator":"Methods/Computational Details (paragraph 2); Methods/Computational Details (paragraph 4)"},{"label":"Architecture / procedure","value":"LASSO regression fits binding affinities from potential-of-mean-force descriptors alone, without the ECFP and protein-fingerprint features used by the combined LightGBM configuration.","status":"source_checked","source_ids":["fingerprint-scoring-2022"],"source_locator":"Methods/Computational Details (paragraph 2); Methods/Computational Details (paragraph 4)"},{"label":"Biological inputs","value":"PMF protein–ligand interaction descriptors","status":"source_checked","source_ids":["fingerprint-scoring-2022"],"source_locator":"Methods (paragraph 1); Results and Discussion/Comparison of Binding Affinities Calculated\nby New Scoring Functions with Experimental Values (paragraph 2)"},{"label":"Outputs","value":"Predicted protein–ligand binding affinity","status":"source_checked","source_ids":["fingerprint-scoring-2022"],"source_locator":"Introduction (paragraph 4); Results and Discussion/Comparison of Binding Affinities Calculated\nby New Scoring Functions with Experimental Values (paragraph 1)"},{"label":"Parameters","value":"A sparse set of fitted linear coefficients on potential-of-mean-force descriptors; a pretrained neural parameter count is inapplicable.","status":"inapplicable","source_ids":["fingerprint-scoring-2022"],"source_locator":"Methods/Machine Learning with LASSO and LightGBM (paragraph 2); Results and Discussion/Analysis of Descriptors\nAffecting the Scoring\nFunction (paragraph 1)"},{"label":"Known versions / configuration","value":"PMF (LASSO) is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["fingerprint-scoring-2022"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"Of 6,271 PDBbind complexes, 4,933 are training, 1,234 are pretest for hyperparameter selection, and 104 are test. LASSO and LightGBM are fitted on interaction fingerprints.","status":"source_checked","source_ids":["fingerprint-scoring-2022"],"source_locator":"Back/Availability notes (paragraph 2); Back/Availability notes (paragraph 1)"},{"label":"Context limits","value":"Protein–ligand interaction fingerprints from supplied complex structures; sequence-token context length is inapplicable.","status":"source_checked","source_ids":["fingerprint-scoring-2022"],"source_locator":"Methods/PMF Score (paragraph 1); Methods/Computational Details (paragraph 2)"},{"label":"Access","value":"The paper links the public RDKit, scikit-learn and LightGBM libraries and provides hyperparameters in its Supporting Information. Additional study data are available from the authors; a complete study-specific checkpoint release is not established.","status":"source_checked","source_ids":["fingerprint-scoring-2022"],"source_locator":"Back / Notes, data and software availability"},{"label":"Code licence","value":"No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","status":"unreported","source_ids":["fingerprint-scoring-2022"],"source_locator":"Back/Availability notes (paragraph 2); Back/Availability notes (paragraph 1)"},{"label":"Weights licence","value":"Not applicable to pretrained weights; coefficients are fitted from the study training complexes.","status":"inapplicable","source_ids":["fingerprint-scoring-2022"],"source_locator":"Results and Discussion/Comparison of Binding Affinities Calculated\nby New Scoring Functions with Experimental Values (paragraph 1); Methods/Computational Details (paragraph 5)"}],"strengths":[{"text":"Offers both a sparse linear control and a nonlinear tree-based model over explicitly defined chemical descriptors.","source_ids":["fingerprint-scoring-2022"],"source_locator":"Methods/Machine Learning with LASSO and LightGBM (paragraph 2); Results and Discussion/Analysis of Descriptors\nAffecting the Scoring\nFunction (paragraph 1)"}],"limitations":[{"text":"The descriptor set and regression algorithm differ between rows; performance cannot be assigned to PMF or fingerprinting in isolation.","source_ids":["fingerprint-scoring-2022"],"source_locator":"Results and Discussion/Analysis of Descriptors\nAffecting the Scoring\nFunction (paragraph 1); Results and Discussion/Systems with Improved and\nUnimproved Scores (paragraph 1)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["PMF protein–ligand interaction descriptors","PMF (LASSO)","Predicted protein–ligand binding affinity"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["fingerprint-scoring-2022"],"source_locator":"Methods/Computational Details (paragraph 2); Methods/Computational Details (paragraph 4)"},"coverage":"limited","gaps":["PMF (LASSO) is the comparison-table label; that label does not specify an immutable weight revision.","No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-81b0394d5ac3e8","kind":"model","name":"TransBind","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["molecular-interactions"]},"source_ids":["transbind-2026"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"TransBind","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"TransBind predicts transcription-factor binding by combining genomic DNA with protein-aware TF embeddings.","summary_source_ids":["transbind-2026"],"summary_source_locator":"Ablation study/Multimodal fusion strategies (paragraph 1); Materials and methods/Deep learning model for TF–DNA binding site classification/Module 1: DNA sequence encoder (paragraph 5)","sections":[{"title":"How the evaluated method works","body":"Cross-attention allows a TF embedding containing sequence and structural information to attend to DNA regions. The protein language model is pretrained on DNA-binding proteins.","source_ids":["transbind-2026"],"source_locator":"Materials and methods/Deep learning model for TF–DNA binding site classification/Module 1: DNA sequence encoder (paragraph 5); Materials and methods/Deep learning model for TF–DNA binding site classification/Module 3: Bimodal feature aggregation for TF–DNA binding prediction (paragraph 4)"},{"title":"What was evaluated","body":"The linked evaluation record identifies TransBind: transcription-factor DNA binding-site prediction. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["transbind-2026"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b4-023"}],"facts":[{"label":"Model type","value":"Study-specific predictive method; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["transbind-2026"],"source_locator":"Materials and methods/Deep learning model for TF–DNA binding site classification/Module 1: DNA sequence encoder (paragraph 5); Materials and methods/Deep learning model for TF–DNA binding site classification/Module 3: Bimodal feature aggregation for TF–DNA binding prediction (paragraph 4)"},{"label":"Architecture / procedure","value":"Cross-attention allows a TF embedding containing sequence and structural information to attend to DNA regions. The protein language model is pretrained on DNA-binding proteins.","status":"source_checked","source_ids":["transbind-2026"],"source_locator":"Materials and methods/Deep learning model for TF–DNA binding site classification/Module 1: DNA sequence encoder (paragraph 5); Materials and methods/Deep learning model for TF–DNA binding site classification/Module 3: Bimodal feature aggregation for TF–DNA binding prediction (paragraph 4)"},{"label":"Biological inputs","value":"DNA windows and transcription-factor protein representations","status":"source_checked","source_ids":["transbind-2026"],"source_locator":"Ablation study/Effect of input DNA sequence length on model performance (paragraph 3); Materials and methods/Deep learning model for label-zero-shot TF–DNA binding site prediction (paragraph 3)"},{"label":"Outputs","value":"TF-binding predictions for genomic regions","status":"source_checked","source_ids":["transbind-2026"],"source_locator":"Introduction (paragraph 8); Materials and methods/DNA data (paragraph 4)"},{"label":"Parameters","value":"4.6 million parameters for the selected unidirectional cross-attention configuration; this does not count the upstream pretrained feature extractor.","status":"source_checked","source_ids":["transbind-2026"],"source_locator":"Ablation study/Multimodal fusion strategies (paragraph 1); Ablation study/Multimodal fusion strategies (paragraph 2)"},{"label":"Known versions / configuration","value":"TransBind is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["transbind-2026"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"690 ChIP-seq experiments covering 161 TFs and 91 human cell types; the reported final DNA-bin collection contains 1,903,668 unique unambiguous bins before reverse-complement augmentation.","status":"source_checked","source_ids":["transbind-2026"],"source_locator":"Materials and methods/DNA data (paragraph 3); Materials and methods/DNA data (paragraph 1)"},{"label":"Context limits","value":"1,000-bp inputs: a 200-bp genomic bin extended by 400 bases on each side. The assembly is GRCh37/hg19; chromosomes 8–9 are test and chromosome 7 is validation.","status":"source_checked","source_ids":["transbind-2026"],"source_locator":"Materials and methods/DNA data (paragraph 2); Table tbl1 (paragraph 1)"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/jianlin-cheng/TransBind/blob/7537f264c5ad94958bcad05bb57edd8028c323ff/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-transbind-2026-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"GNU GPL version 3 (study repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-transbind-2026-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-transbind-2026-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Makes TF identity and representation explicit instead of using DNA alone.","source_ids":["transbind-2026"],"source_locator":"Results/Convolutional kernels capture general DNA patterns (paragraph 4); Ablation study/Effect of input DNA sequence length on model performance (paragraph 3)"}],"limitations":[{"text":"ChIP-seq labels, cell context and unseen-TF partitioning define the biological claim; protein-aware prediction does not itself confirm binding in a new experiment.","source_ids":["transbind-2026"],"source_locator":"Introduction (paragraph 8); Abstract (paragraph 1)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["DNA windows and transcription-factor protein representations","TransBind","TF-binding predictions for genomic regions"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["transbind-2026"],"source_locator":"Materials and methods/Deep learning model for TF–DNA binding site classification/Module 1: DNA sequence encoder (paragraph 5); Materials and methods/Deep learning model for TF–DNA binding site classification/Module 3: Bimodal feature aggregation for TF–DNA binding prediction (paragraph 4)"},"coverage":"limited","gaps":["TransBind is the comparison-table label; that label does not specify an immutable weight revision.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-86393c76dd8fa9","kind":"model","name":"MegSite + ESM3","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["proteins-complexes"]},"source_ids":["megsite-2025"],"links":[],"attributes":{"entity_level":"method","version":"not stated in table","reported_name":"MegSite + ESM3","historical_missing_metadata":{"checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"MegSite predicts protein residues that bind DNA or RNA using multimodal ESM3 information.","summary_source_ids":["megsite-2025"],"summary_source_locator":"Bidirectional cross-attention fusion/The architecture of MegSite (paragraph 1); Bidirectional cross-attention fusion/Implementation details (paragraph 2)","sections":[{"title":"How the evaluated method works","body":"Sequence, structural and functional representations from ESM3 inform a task-specific nucleic-acid-binding-site predictor.","source_ids":["megsite-2025"],"source_locator":"Results/Performance of different network architectures (paragraph 7); Materials and methods/Multimodal protein language model (paragraph 2)"},{"title":"What was evaluated","body":"The linked evaluation record identifies MegSite + ESM3: DNA-binding residue prediction. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["megsite-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-b2-megsite-2025"}],"facts":[{"label":"Model type","value":"Study-specific predictive method; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["megsite-2025"],"source_locator":"Results/Performance of different network architectures (paragraph 7); Materials and methods/Multimodal protein language model (paragraph 2)"},{"label":"Architecture / procedure","value":"Sequence, structural and functional representations from ESM3 inform a task-specific nucleic-acid-binding-site predictor.","status":"source_checked","source_ids":["megsite-2025"],"source_locator":"Results/Performance of different network architectures (paragraph 7); Materials and methods/Multimodal protein language model (paragraph 2)"},{"label":"Biological inputs","value":"Protein sequence, structure and function representations","status":"source_checked","source_ids":["megsite-2025"],"source_locator":"Results/Comparison with alternative protein language model embeddings (paragraph 3); Protein graph representation/Node feature (paragraph 1)"},{"label":"Outputs","value":"Per-residue DNA- or RNA-binding predictions","status":"source_checked","source_ids":["megsite-2025"],"source_locator":"Results/Comparison with state-of-the-art methods (paragraph 3); Protein graph representation (paragraph 1)"},{"label":"Parameters","value":"The feature extractor is ESM3-Small: 1.4 billion parameters and 48 transformer blocks. The equivariant graph predictor is an additional fitted component.","status":"source_checked","source_ids":["megsite-2025"],"source_locator":"Materials and methods/Multimodal protein language model (paragraph 2); Protein graph representation/Node feature (paragraph 1)"},{"label":"Known versions / configuration","value":"MegSite + ESM3 is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["megsite-2025"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"DNA-573_Train and RNA-495_Train with independent DNA/RNA test sets; the paper filters training/test sequence similarity at 30%.","status":"source_checked","source_ids":["megsite-2025"],"source_locator":"Materials and methods/Benchmark datasets (paragraph 1); Results/Comparison with state-of-the-art methods (paragraph 1)"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["megsite-2025","evidence-reported-megsite-2025-readme-md"],"source_locator":"Materials and methods/Benchmark datasets; Materials and methods/Multimodal protein language model; Bidirectional cross-attention fusion/Implementation details; Bidirectional cross-attention fusion/The architecture of MegSite; Results/Performance of different network architectures; Results/Comparison with state-of-the-art methods; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/pengsl-lab/MegSite/blob/4d1f5441f15bd20eb3e7c9e5be2d4ca5a857f46b/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-megsite-2025-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","status":"unreported","source_ids":["evidence-reported-megsite-2025-readme-md"],"source_locator":"README.md and repository-root licence-file search"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-megsite-2025-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Includes independent tests and analyses of proteins with low structural similarity.","source_ids":["megsite-2025"],"source_locator":"Results/Performance of different network architectures (paragraph 6); Abstract (paragraph 1)"}],"limitations":[{"text":"These are downstream binding-site predictions with multimodal inputs, not raw ESM3 outputs or evidence that predicted binding is experimentally realised.","source_ids":["megsite-2025"],"source_locator":"Bidirectional cross-attention fusion/The architecture of MegSite (paragraph 1); Results/Comparison with alternative protein language model embeddings (paragraph 3)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Protein sequence, structure and function representations","MegSite + ESM3","Per-residue DNA- or RNA-binding predictions"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["megsite-2025"],"source_locator":"Results/Performance of different network architectures (paragraph 7); Materials and methods/Multimodal protein language model (paragraph 2)"},"coverage":"limited","gaps":["MegSite + ESM3 is the comparison-table label; that label does not specify an immutable weight revision.","A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-884582fb0c70dc","kind":"model","name":"Boltz-1","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["molecular-interactions"]},"source_ids":["antibody-flexibility-2025"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"Boltz-1","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"This structure-prediction configuration is evaluated for antibody–antigen complex and interface quality.","summary_source_ids":["antibody-flexibility-2025"],"summary_source_locator":"Results/Fold quality affects predictive performance/Folding quality declines across models without MSAs or templates. (paragraph 3); Results/Fold quality affects predictive performance/AF3 and Chai-1 (no-MSA/template) effectively preserve prediction quality. (paragraph 3)","sections":[{"title":"How the evaluated method works","body":"The generative co-folding predictor builds antibody–antigen structures that are then assessed for interface quality and downstream paratope/epitope prediction.","source_ids":["antibody-flexibility-2025"],"source_locator":"Results/Fold quality affects predictive performance/Folding quality declines across models without MSAs or templates. (paragraph 3); Results/Fold quality affects predictive performance/Folding quality declines across models without MSAs or templates. (paragraph 2)"},{"title":"Underlying method and version boundaries","body":"The official Boltz repository publishes separate Boltz-1 and Boltz-2 models. Boltz-2 adds affinity prediction and other changes; these are not retroactively attributed to Boltz-1 evaluations.","source_ids":["evidence-reported-base-boltz-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies Boltz-1: Antibody–antigen interaction prediction using folded complexes. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["antibody-flexibility-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b3-038"}],"facts":[{"label":"Model type","value":"Biomolecular structure predictor; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-boltz-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"The generative co-folding predictor builds antibody–antigen structures that are then assessed for interface quality and downstream paratope/epitope prediction.","status":"source_checked","source_ids":["antibody-flexibility-2025"],"source_locator":"Results/Fold quality affects predictive performance/Folding quality declines across models without MSAs or templates. (paragraph 3); Results/Fold quality affects predictive performance/Folding quality declines across models without MSAs or templates. (paragraph 2)"},{"label":"Biological inputs","value":"Antibody and antigen sequences; MSA/template availability is explicitly configuration-dependent.","status":"source_checked","source_ids":["antibody-flexibility-2025"],"source_locator":"Results/Flexibility improves antibody-antigen interaction prediction/Binary flexibility patterns emerge despite linear flexibility modeling. (paragraph 3); Results/Flexibility improves antibody-antigen interaction prediction (paragraph 1)"},{"label":"Outputs","value":"Predicted antibody–antigen complex coordinates","status":"source_checked","source_ids":["antibody-flexibility-2025"],"source_locator":"Results/Fold quality affects predictive performance/Folding quality declines across models without MSAs or templates. (paragraph 1); Results/Fold quality affects predictive performance/AF3 and Chai-1 (no-MSA/template) effectively preserve prediction quality. (paragraph 3)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["antibody-flexibility-2025","evidence-reported-base-boltz-readme-md"],"source_locator":"Materials and methods; Materials and methods/Dataset; Materials and methods/Method; Materials and methods/Method/Data representation.; Materials and methods/Method/Flexibility score with ESMFold.; Materials and methods/Method/Model architecture.; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"Boltz-1 is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["antibody-flexibility-2025"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"The released source implementation supplies prefolded antibody–antigen complexes for the GEP test set, with neither MSAs nor templates. No retraining of the structure predictor is described; downstream interface prediction is a separate component.","status":"source_checked","source_ids":["antibody-flexibility-2025"],"source_locator":"Results / Fold quality affects predictive performance"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["antibody-flexibility-2025","evidence-reported-base-boltz-readme-md"],"source_locator":"Materials and methods; Materials and methods/Dataset; Materials and methods/Method; Materials and methods/Method/Data representation.; Materials and methods/Method/Flexibility score with ESMFold.; Materials and methods/Method/Model architecture.; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://github.com/jwohlwend/boltz/blob/b1ebfc46ecf57f5414e0d1a6f9027bbb122c53bc/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-boltz-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"MIT (upstream repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-base-boltz-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"MIT for all code and model weights, explicitly stated in the official Boltz README. Historical evaluated weight identity remains separately recorded.","status":"source_checked","source_ids":["evidence-reported-base-boltz-readme-md"],"source_locator":"README.md; license and model release introduction"}],"strengths":[{"text":"Examines how predicted structures affect downstream interface-site analysis.","source_ids":["antibody-flexibility-2025"],"source_locator":"Results/Fold quality affects predictive performance/Folding quality declines across models without MSAs or templates. (paragraph 3); Results/Fold quality affects predictive performance/AF3 and Chai-1 (no-MSA/template) effectively preserve prediction quality. (paragraph 4)"}],"limitations":[{"text":"The paper reports configuration-dependent differences and limited antibody–antigen accuracy; general complex-folding scores cannot substitute for this task.","source_ids":["antibody-flexibility-2025"],"source_locator":"Results/Flexibility improves antibody-antigen interaction prediction/Binary flexibility patterns emerge despite linear flexibility modeling. (paragraph 3); Results/Fold quality affects predictive performance (paragraph 1)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Antibody and antigen sequences","Boltz-1","Predicted antibody–antigen complex coordinates"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["antibody-flexibility-2025"],"source_locator":"Results/Fold quality affects predictive performance/Folding quality declines across models without MSAs or templates. (paragraph 3); Results/Fold quality affects predictive performance/Folding quality declines across models without MSAs or templates. (paragraph 2)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","Boltz-1 is the comparison-table label; that label does not specify an immutable weight revision.","A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-8861b9ad9b9c9b","kind":"model","name":"Kraken2","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["microbes-communities"]},"source_ids":["icctax-2025"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"Kraken2","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Kraken 2 is the reference-database taxonomic classifier used in this metagenomic comparison.","summary_source_ids":["icctax-2025"],"summary_source_locator":"3 Experiments/3.2 Experiments on ICCTax performance/3.2.2 Performance on the complete dataset (paragraph 1); Abstract (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"Sequence evidence is matched against a taxonomic reference index to assign reads or contigs. Database construction and the chosen confidence/settings remain part of the evaluated configuration.","source_ids":["icctax-2025"],"source_locator":"2 Methods/2.2 Complementary-view based hierarchical metric learning (paragraph 1); 3 Experiments/3.6 Taxonomic profiling on real metagenomic data (paragraph 1)"},{"title":"Underlying method and version boundaries","body":"Kraken 2 classifies sequences against a taxonomic reference index. The software version and the database used to construct that index are distinct reproducibility requirements.","source_ids":["evidence-reported-base-kraken2-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies Kraken2: Hierarchical metagenomic taxonomy classification. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["icctax-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b3-036"}],"facts":[{"label":"Model type","value":"Reference-index taxonomic classifier; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-kraken2-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"Sequence evidence is matched against a taxonomic reference index to assign reads or contigs. Database construction and the chosen confidence/settings remain part of the evaluated configuration.","status":"source_checked","source_ids":["icctax-2025"],"source_locator":"2 Methods/2.2 Complementary-view based hierarchical metric learning (paragraph 1); 3 Experiments/3.6 Taxonomic profiling on real metagenomic data (paragraph 1)"},{"label":"Biological inputs","value":"Metagenomic reads or contigs and a taxonomic reference database","status":"source_checked","source_ids":["icctax-2025"],"source_locator":"3 Experiments/3.6 Taxonomic profiling on real metagenomic data (paragraph 1); 3 Experiments/3.2 Experiments on ICCTax performance/3.2.2 Performance on the complete dataset (paragraph 1)"},{"label":"Outputs","value":"Taxonomic assignments","status":"source_checked","source_ids":["icctax-2025"],"source_locator":"2 Methods/2.3 Compactness loss (paragraph 7); 2 Methods/2.3 Compactness loss (paragraph 2)"},{"label":"Parameters","value":"Not applicable as a neural model size.","status":"inapplicable","source_ids":["icctax-2025"],"source_locator":"2 Methods/2.1 HyenaDNA (paragraph 6); 2 Methods/2.1 HyenaDNA (paragraph 5)"},{"label":"Known versions / configuration","value":"Kraken2 is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["icctax-2025"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"Reference-index construction rather than foundation-model pretraining.","status":"source_checked","source_ids":["icctax-2025"],"source_locator":"3 Experiments/3.2 Experiments on ICCTax performance/3.2.2 Performance on the complete dataset (paragraph 1); 3 Experiments/3.1 Datasets (paragraph 1)"},{"label":"Context limits","value":"Not applicable to a pretrained sequence-token window; read/contig lengths and the reference-database or comparison configuration determine the analysed input.","status":"inapplicable","source_ids":["icctax-2025"],"source_locator":"Table vbaf257-T1 (paragraph 1); 3 Experiments/3.2 Experiments on ICCTax performance/3.2.2 Performance on the complete dataset (paragraph 1)"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://github.com/DerrickWood/kraken2/blob/8c190b1b668825935dbf6dee5f969227dc8269bb/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-kraken2-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"MIT (upstream repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-base-kraken2-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"Not applicable: uses a reference index rather than a pretrained neural checkpoint.","status":"inapplicable","source_ids":["icctax-2025"],"source_locator":"2 Methods/2.1 HyenaDNA (paragraph 6); 2 Methods/2.1 HyenaDNA (paragraph 5)"}],"strengths":[{"text":"Provides an established reference-based comparator for learned or alignment-based taxonomic methods.","source_ids":["icctax-2025"],"source_locator":"3 Experiments/3.2 Experiments on ICCTax performance/3.2.2 Performance on the complete dataset (paragraph 1); 3 Experiments/3.2 Experiments on ICCTax performance (paragraph 1)"}],"limitations":[{"text":"Unrepresented organisms and reference contamination can alter assignments; results are specific to the database and query distribution.","source_ids":["icctax-2025"],"source_locator":"3 Experiments/3.2 Experiments on ICCTax performance/3.2.2 Performance on the complete dataset (paragraph 1); 2 Methods/2.2 Complementary-view based hierarchical metric learning (paragraph 3)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Metagenomic reads or contigs and a taxonomic reference database","Kraken2","Taxonomic assignments"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["icctax-2025"],"source_locator":"2 Methods/2.2 Complementary-view based hierarchical metric learning (paragraph 1); 3 Experiments/3.6 Taxonomic profiling on real metagenomic data (paragraph 1)"},"coverage":"limited","gaps":["Kraken2 is the comparison-table label; that label does not specify an immutable weight revision.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-89f5a8f309fa18","kind":"model","name":"scRegNet (scBERT backbone)","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["cells-tissues"]},"source_ids":["scregnet-2025"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"scRegNet (scBERT backbone)","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"scRegNet predicts gene-regulatory links using single-cell foundation-model features and graph learning.","summary_source_ids":["scregnet-2025"],"summary_source_locator":"Conclusions and Discussions (paragraph 1); Results/Performance on benchmark datasets (paragraph 2)","sections":[{"title":"How the evaluated method works","body":"A specified pretrained backbone supplies context-aware gene representations; a graph-based supervised predictor learns regulatory connections from known links. Geneformer and scBERT are separate backbone configurations.","source_ids":["scregnet-2025"],"source_locator":"Method/Graph-based learning with GNNs (paragraph 1); Method/Gene representations from foundation models (paragraph 1)"},{"title":"What was evaluated","body":"The linked evaluation record identifies scRegNet (scBERT backbone): Gene-regulatory link prediction. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["scregnet-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-032"}],"facts":[{"label":"Model type","value":"Graph-based predictive method; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["scregnet-2025"],"source_locator":"Method/Graph-based learning with GNNs (paragraph 1); Method/Gene representations from foundation models (paragraph 1)"},{"label":"Architecture / procedure","value":"A specified pretrained backbone supplies context-aware gene representations; a graph-based supervised predictor learns regulatory connections from known links. Geneformer and scBERT are separate backbone configurations.","status":"source_checked","source_ids":["scregnet-2025"],"source_locator":"Method/Graph-based learning with GNNs (paragraph 1); Method/Gene representations from foundation models (paragraph 1)"},{"label":"Biological inputs","value":"Single-cell expression data, pretrained gene embeddings and known regulatory links","status":"source_checked","source_ids":["scregnet-2025"],"source_locator":"Method/Gene representations from foundation models/Geneformer (paragraph 2); Method/Graph-based learning with GNNs (paragraph 1)"},{"label":"Outputs","value":"Predicted gene-regulatory connections","status":"source_checked","source_ids":["scregnet-2025"],"source_locator":"Method/Model training (paragraph 1); Method/Link prediction layer (paragraph 2)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["scregnet-2025","evidence-reported-scregnet-2025-readme-md"],"source_locator":"Method; Method/Gene representations from foundation models; Method/Gene representations from foundation models/scBERT; Method/Gene representations from foundation models/scFoundation; Method/Gene representations from foundation models/Geneformer; Method/Gene representations from foundation models/Mean pooling; Method/Graph-based learning with GNNs; Method/Unified gene representations; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"scRegNet (scBERT backbone) is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["scregnet-2025"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"Backbone pretraining is separate from supervised regulatory-link fitting; the study uses known TF–DNA binding information for the latter.","status":"source_checked","source_ids":["scregnet-2025"],"source_locator":"Experimental Setup/Datasets and data pre-processing (paragraph 2); Experimental Setup/Datasets and data pre-processing (paragraph 4)"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["scregnet-2025","evidence-reported-scregnet-2025-readme-md"],"source_locator":"Method; Method/Gene representations from foundation models; Method/Gene representations from foundation models/scBERT; Method/Gene representations from foundation models/scFoundation; Method/Gene representations from foundation models/Geneformer; Method/Gene representations from foundation models/Mean pooling; Method/Graph-based learning with GNNs; Method/Unified gene representations; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/sindhura-cs/scRegNet/blob/30d0215efd99c40ceaceb37d161fc0a86a236e0b/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-scregnet-2025-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","status":"unreported","source_ids":["evidence-reported-scregnet-2025-readme-md"],"source_locator":"README.md and repository-root licence-file search"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-scregnet-2025-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Tests whether pretrained gene representations help when experimentally supported regulatory links are limited.","source_ids":["scregnet-2025"],"source_locator":"Method/Graph-based learning with GNNs (paragraph 1); Introduction (paragraph 3)"}],"limitations":[{"text":"Supervised link prediction on known binding data does not establish causality or intervention effects for all inferred edges.","source_ids":["scregnet-2025"],"source_locator":"Abstract (paragraph 1); Method/Graph-based learning with GNNs (paragraph 1)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Single-cell expression data, pretrained gene embeddings and known regulatory links","scRegNet (scBERT backbone)","Predicted gene-regulatory connections"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["scregnet-2025"],"source_locator":"Method/Graph-based learning with GNNs (paragraph 1); Method/Gene representations from foundation models (paragraph 1)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","scRegNet (scBERT backbone) is the comparison-table label; that label does not specify an immutable weight revision.","A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-8ad3e0cefde796","kind":"model","name":"Vaxign-DL + ESM","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["proteins-complexes"]},"source_ids":["vaxign-esm-2024"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"Vaxign-DL + ESM","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Vaxign-DL plus ESM combines protein-language-model features with engineered features to predict bacterial protective antigens.","summary_source_ids":["vaxign-esm-2024"],"summary_source_locator":"Results/Performance analysis ESM enhance Vaxign-DL model (paragraph 3); Methods/ESM generation of new features based on protein sequences (paragraph 2)","sections":[{"title":"How the evaluated method works","body":"ESM-1b (33 layers, 650M parameters) supplies 1,280 sequence features, combined with 509 existing features in the deep protective-antigen classifier. This paper does not use ESM-2 for the reported configuration.","source_ids":["vaxign-esm-2024"],"source_locator":"Methods/ESM generation of new features based on protein sequences (paragraph 1); Results/Performance analysis ESM enhance Vaxign-DL model (paragraph 3)"},{"title":"What was evaluated","body":"The linked evaluation record identifies Vaxign-DL + ESM: vaccine-antigen candidate prediction. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["vaxign-esm-2024"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b4-012"}],"facts":[{"label":"Model type","value":"Study-specific predictive method; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["vaxign-esm-2024"],"source_locator":"Methods/ESM generation of new features based on protein sequences (paragraph 1); Results/Performance analysis ESM enhance Vaxign-DL model (paragraph 3)"},{"label":"Architecture / procedure","value":"ESM-1b (33 layers, 650M parameters) supplies 1,280 sequence features, combined with 509 existing features in the deep protective-antigen classifier. This paper does not use ESM-2 for the reported configuration.","status":"source_checked","source_ids":["vaxign-esm-2024"],"source_locator":"Methods/ESM generation of new features based on protein sequences (paragraph 1); Results/Performance analysis ESM enhance Vaxign-DL model (paragraph 3)"},{"label":"Biological inputs","value":"Protein sequence-derived learned and engineered features","status":"source_checked","source_ids":["vaxign-esm-2024"],"source_locator":"Results/Performance analysis ESM enhance Vaxign-DL model (paragraph 3); Results/Hyperparameter Optimization Study/Leave-one-pathogen-out Validation (paragraph 1)"},{"label":"Outputs","value":"Predicted protective-antigen labels","status":"source_checked","source_ids":["vaxign-esm-2024"],"source_locator":"Results/Performance analysis ESM enhance Vaxign-DL model (paragraph 3); Methods/Collection of Positive and Negative protein sequences (paragraph 1)"},{"label":"Parameters","value":"650-million-parameter ESM-1b feature extractor; the downstream classifier is additional.","status":"source_checked","source_ids":["vaxign-esm-2024"],"source_locator":"Results/Performance analysis ESM enhance Vaxign-DL model (paragraph 3); Methods/Deep learning pipeline (paragraph 1)"},{"label":"Known versions / configuration","value":"Vaxign-DL + ESM is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["vaxign-esm-2024"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"Protegen-derived positive antigens, with 397 positives after the described homology filtering and separately constructed negatives; leave-one-pathogen-out validation is included.","status":"source_checked","source_ids":["vaxign-esm-2024"],"source_locator":"Introduction (paragraph 4); Abstract (paragraph 1)"},{"label":"Context limits","value":"The study either skips proteins longer than 1,024 amino acids or truncates them to the first 1,024; Skip and Cut are separate evaluated settings.","status":"source_checked","source_ids":["vaxign-esm-2024"],"source_locator":"Results/Hyperparameter Optimization Study/Comparison of Two ESM methods for Processing Long Sequence Proteins (paragraph 1); Methods/ESM generation of new features based on protein sequences (paragraph 3)"},{"label":"Access","value":"The paper describes the ESM-1b extension but does not establish a separate released checkpoint for this fitted Vaxign-DL plus ESM pipeline.","status":"unreported","source_ids":["vaxign-esm-2024"],"source_locator":"Complete paper; ESM-1b feature extraction and neural-network fitting descriptions; exact-name implementation search"},{"label":"Code licence","value":"No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","status":"unreported","source_ids":["vaxign-esm-2024"],"source_locator":"Discussion (paragraph 6); Discussion (paragraph 5)"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["vaxign-esm-2024"],"source_locator":"Methods/Deep learning pipeline (paragraph 2); Discussion (paragraph 6)"}],"strengths":[{"text":"Tests whether pretrained features add information beyond an established engineered-feature pipeline.","source_ids":["vaxign-esm-2024"],"source_locator":"Methods/Deep learning pipeline (paragraph 1); Methods/ESM generation of new features based on protein sequences (paragraph 3)"}],"limitations":[{"text":"The source is a preprint; computational antigen prioritisation is not vaccine-efficacy or safety validation.","source_ids":["vaxign-esm-2024"],"source_locator":"Results/Performance analysis ESM enhance Vaxign-DL model (paragraph 3); Introduction (paragraph 4)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Protein sequence-derived learned and engineered features","Vaxign-DL + ESM","Predicted protective-antigen labels"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["vaxign-esm-2024"],"source_locator":"Methods/ESM generation of new features based on protein sequences (paragraph 1); Results/Performance analysis ESM enhance Vaxign-DL model (paragraph 3)"},"coverage":"limited","gaps":["Vaxign-DL + ESM is the comparison-table label; that label does not specify an immutable weight revision.","The paper describes the ESM-1b extension but does not establish a separate released checkpoint for this fitted Vaxign-DL plus ESM pipeline.","No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-8cb3dd4e9f5b10","kind":"model","name":"Lazypipe-nt","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["microbes-communities"]},"source_ids":["lazypipe-2020"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"Lazypipe-nt","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"LAZYPIPE is an automated assembly and taxonomic-profiling workflow for virus discovery in sequencing libraries.","summary_source_ids":["lazypipe-2020"],"summary_source_locator":"2. Materials and methods/2.2 Unix pipeline for assembly, taxonomic profiling and binning of NGS data (paragraph 1); Abstract (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"Unix-based C++, Perl and R components assemble reads and assign taxonomy. The nt-labelled configuration is the nucleotide-database variant, distinct from other database choices in the paper.","source_ids":["lazypipe-2020"],"source_locator":"2. Materials and methods/2.2 Unix pipeline for assembly, taxonomic profiling and binning of NGS data (paragraph 3); 2. Materials and methods/2.2 Unix pipeline for assembly, taxonomic profiling and binning of NGS data (paragraph 5)"},{"title":"What was evaluated","body":"The linked evaluation record identifies Lazypipe-nt: Simulated metagenome virus-taxon retrieval. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["lazypipe-2020"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b3-021"}],"facts":[{"label":"Model type","value":"Study-specific predictive method; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["lazypipe-2020"],"source_locator":"2. Materials and methods/2.2 Unix pipeline for assembly, taxonomic profiling and binning of NGS data (paragraph 3); 2. Materials and methods/2.2 Unix pipeline for assembly, taxonomic profiling and binning of NGS data (paragraph 5)"},{"label":"Architecture / procedure","value":"Unix-based C++, Perl and R components assemble reads and assign taxonomy. The nt-labelled configuration is the nucleotide-database variant, distinct from other database choices in the paper.","status":"source_checked","source_ids":["lazypipe-2020"],"source_locator":"2. Materials and methods/2.2 Unix pipeline for assembly, taxonomic profiling and binning of NGS data (paragraph 3); 2. Materials and methods/2.2 Unix pipeline for assembly, taxonomic profiling and binning of NGS data (paragraph 5)"},{"label":"Biological inputs","value":"NGS libraries from host-associated or environmental samples","status":"source_checked","source_ids":["lazypipe-2020"],"source_locator":"Abstract (paragraph 1); 3. Results/3.3 Novel virome sequences from mink faecal samples (paragraph 1)"},{"label":"Outputs","value":"Assembled contigs and taxonomic/viral profiles","status":"source_checked","source_ids":["lazypipe-2020"],"source_locator":"1. Introduction (paragraph 6); 3. Results/3.1 Excellent recall and precision for both simulated and real datasets (paragraph 3)"},{"label":"Parameters","value":"Not applicable to a neural parameter count: this pipeline assembles, aligns and bins reads/contigs.","status":"inapplicable","source_ids":["lazypipe-2020"],"source_locator":"2. Materials and methods/2.2 Unix pipeline for assembly, taxonomic profiling and binning of NGS data (paragraph 4); 2. Materials and methods/2.2 Unix pipeline for assembly, taxonomic profiling and binning of NGS data (paragraph 8)"},{"label":"Known versions / configuration","value":"Lazypipe-nt is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["lazypipe-2020"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"No task-specific neural fitting: assembled contigs are classified by Centrifuge against NCBI nt. The benchmark removes alignments with less than 60 nucleotides matched; a dated NCBI nt snapshot is not identified in this passage.","status":"source_checked","source_ids":["lazypipe-2020"],"source_locator":"Materials and methods / Unix pipeline; Benchmarking performance"},{"label":"Context limits","value":"Not applicable to a learned token window; paired-end reads and their assembled contigs are the analysed input.","status":"inapplicable","source_ids":["lazypipe-2020"],"source_locator":"1. Introduction (paragraph 6); 2. Materials and methods/2.2 Unix pipeline for assembly, taxonomic profiling and binning of NGS data (paragraph 4)"},{"label":"Access","value":"The paper provides the source and manual at https://bitbucket.org/plyusnin/lazypipe/ and https://www.helsinki.fi/en/projects/lazypipe. This records the published access route, not a live-service availability guarantee.","status":"source_checked","source_ids":["lazypipe-2020"],"source_locator":"Materials and methods / Unix pipeline for assembly, taxonomic profiling and binning of NGS data"},{"label":"Code licence","value":"No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","status":"unreported","source_ids":["lazypipe-2020"],"source_locator":"1. Introduction (paragraph 4); 4. Discussion (paragraph 1)"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["lazypipe-2020"],"source_locator":"1. Introduction (paragraph 4); 4. Discussion (paragraph 1)"}],"strengths":[{"text":"Combines assembly with annotation for inspection of known and potentially novel viral sequences.","source_ids":["lazypipe-2020"],"source_locator":"4. Discussion (paragraph 1); 4. Discussion (paragraph 4)"}],"limitations":[{"text":"Results depend on library preprocessing, assembly and reference databases; the pipeline is not a single learned model.","source_ids":["lazypipe-2020"],"source_locator":"2. Materials and methods/2.2 Unix pipeline for assembly, taxonomic profiling and binning of NGS data (paragraph 8); 2. Materials and methods/2.3 Benchmarking performance (paragraph 1)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["NGS libraries from host-associated or environmental samples","Lazypipe-nt","Assembled contigs and taxonomic/viral profiles"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["lazypipe-2020"],"source_locator":"2. Materials and methods/2.2 Unix pipeline for assembly, taxonomic profiling and binning of NGS data (paragraph 3); 2. Materials and methods/2.2 Unix pipeline for assembly, taxonomic profiling and binning of NGS data (paragraph 5)"},"coverage":"limited","gaps":["Lazypipe-nt is the comparison-table label; that label does not specify an immutable weight revision.","No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-8d2c291733dfe1","kind":"model","name":"DeepInterAware","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["molecular-interactions"]},"source_ids":["deepinteraware-2025"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"DeepInterAware","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"DeepInterAware predicts antigen–antibody interactions from sequence representations and learned interface features.","summary_source_ids":["deepinteraware-2025"],"summary_source_locator":"Experimental Section/DeepInterAware Architecture—Sequence Encoder (paragraph 2); Body (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"ESM-2 encodes antigens and AbLang encodes antibodies. An interaction-interface learner and a specificity-information learner feed a dynamic-confidence-fusion module for the final prediction.","source_ids":["deepinteraware-2025"],"source_locator":"Experimental Section/DeepInterAware Architecture (paragraph 1); Results/DeepInterAware Framework (paragraph 1)"},{"title":"What was evaluated","body":"The linked evaluation record identifies DeepInterAware: antigen-antibody HIV neutralization prediction. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["deepinteraware-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b4-022"}],"facts":[{"label":"Model type","value":"Study-specific predictive method; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["deepinteraware-2025"],"source_locator":"Experimental Section/DeepInterAware Architecture (paragraph 1); Results/DeepInterAware Framework (paragraph 1)"},{"label":"Architecture / procedure","value":"ESM-2 encodes antigens and AbLang encodes antibodies. An interaction-interface learner and a specificity-information learner feed a dynamic-confidence-fusion module for the final prediction.","status":"source_checked","source_ids":["deepinteraware-2025"],"source_locator":"Experimental Section/DeepInterAware Architecture (paragraph 1); Results/DeepInterAware Framework (paragraph 1)"},{"label":"Biological inputs","value":"Antigen and antibody amino-acid sequences","status":"source_checked","source_ids":["deepinteraware-2025"],"source_locator":"Experimental Section/DeepInterAware Architecture—Sequence Encoder (paragraph 2); Experimental Section/DeepInterAware Architecture—Interaction Interface‐Aware Learner (paragraph 4)"},{"label":"Outputs","value":"Interaction predictions, with additional binding-site and mutation analyses in the study","status":"source_checked","source_ids":["deepinteraware-2025"],"source_locator":"Experimental Section/DeepInterAware Architecture—Dynamic Confidence Fusion and Prediction (paragraph 1); Experimental Section/Experimental Setting—Baseline Methods (paragraph 1)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["deepinteraware-2025","evidence-reported-deepinteraware-2025-readme-md"],"source_locator":"Results/DeepInterAware Outperforms the State‐of‐the‐Art Methods in Antigen–Antibody Interaction Predction; Experimental Section/DeepInterAware Architecture; Experimental Section/DeepInterAware Architecture—Sequence Encoder; Experimental Section/DeepInterAware Architecture—Interaction Interface‐Aware Learner; Experimental Section/DeepInterAware Architecture—Specificity Information Learner; Experimental Section/DeepInterAware Architecture—Dynamic Confidence Fusion and Prediction; Experimental Section/DeepInterAware Architecture—Model Training; Experimental Section/Experimental Setting—Datasets; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"DeepInterAware is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["deepinteraware-2025"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"Antigen–antibody interaction datasets include HIV training and subsequent CoV-AbDab fine-tuning; splits separately hold out antibodies, antigens, or both.","status":"source_checked","source_ids":["deepinteraware-2025"],"source_locator":"Experimental Section/Problem Formulation (paragraph 1); Experimental Section/Problem Formulation (paragraph 2)"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["deepinteraware-2025","evidence-reported-deepinteraware-2025-readme-md"],"source_locator":"Results/DeepInterAware Outperforms the State‐of‐the‐Art Methods in Antigen–Antibody Interaction Predction; Experimental Section/DeepInterAware Architecture; Experimental Section/DeepInterAware Architecture—Sequence Encoder; Experimental Section/DeepInterAware Architecture—Interaction Interface‐Aware Learner; Experimental Section/DeepInterAware Architecture—Specificity Information Learner; Experimental Section/DeepInterAware Architecture—Dynamic Confidence Fusion and Prediction; Experimental Section/DeepInterAware Architecture—Model Training; Experimental Section/Experimental Setting—Datasets; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/BioMedicalBigDataMiningLab/DeepInterAware/blob/11a283264fb7177f842d56d5c6b49f9bb7f10abd/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-deepinteraware-2025-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"CC BY-NC 4.0 for the DeepInterAware content and derivatives, explicitly declared in the maintained official repository.","status":"source_checked","source_ids":["evidence-reported-deepinteraware-2025-readme-md"],"source_locator":"README.md; License"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-deepinteraware-2025-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"The evaluation separates unseen-antibody, unseen-antigen and jointly unseen settings.","source_ids":["deepinteraware-2025"],"source_locator":"Experimental Section/DeepInterAware Architecture—Sequence Encoder (paragraph 2); Experimental Section/Problem Formulation (paragraph 1)"}],"limitations":[{"text":"Sequence-derived interface features are predictions rather than experimentally resolved contacts; transfer depends on the sequence-distribution and split.","source_ids":["deepinteraware-2025"],"source_locator":"Experimental Section/DeepInterAware Architecture—Interaction Interface‐Aware Learner (paragraph 5); Experimental Section/DeepInterAware Architecture—Specificity Information Learner (paragraph 1)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Antigen and antibody amino-acid sequences","DeepInterAware","Interaction predictions, with additional binding-site and mutation analyses in the study"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["deepinteraware-2025"],"source_locator":"Experimental Section/DeepInterAware Architecture (paragraph 1); Results/DeepInterAware Framework (paragraph 1)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","DeepInterAware is the comparison-table label; that label does not specify an immutable weight revision.","A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-8db190bee6aae5","kind":"model","name":"ENBED","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["dna-genomes"]},"source_ids":["enbed-2024"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"ENBED","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"ENBED is a byte-level encoder–decoder transformer for genomic sequence representation and sequence-to-sequence tasks.","summary_source_ids":["enbed-2024"],"summary_source_locator":"2 Methods/2.5 Application domains (paragraph 1); 5 Discussion (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"Byte-level nucleotide inputs feed encoder and decoder transformer blocks with a subquadratic attention implementation. Masked-language pretraining precedes task-specific adaptation.","source_ids":["enbed-2024"],"source_locator":"1 Introduction/1.2 Our contributions (paragraph 1); 4 Ablation studies/4.1 Encoder–decoder architecture (paragraph 2)"},{"title":"What was evaluated","body":"The linked evaluation record identifies ENBED: Enhancer classification. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["enbed-2024"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-003"}],"facts":[{"label":"Model type","value":"Transformer representation pipeline; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["enbed-2024"],"source_locator":"1 Introduction/1.2 Our contributions (paragraph 1); 4 Ablation studies/4.1 Encoder–decoder architecture (paragraph 2)"},{"label":"Architecture / procedure","value":"Byte-level nucleotide inputs feed encoder and decoder transformer blocks with a subquadratic attention implementation. Masked-language pretraining precedes task-specific adaptation.","status":"source_checked","source_ids":["enbed-2024"],"source_locator":"1 Introduction/1.2 Our contributions (paragraph 1); 4 Ablation studies/4.1 Encoder–decoder architecture (paragraph 2)"},{"label":"Biological inputs","value":"DNA sequences at single-byte nucleotide resolution","status":"source_checked","source_ids":["enbed-2024"],"source_locator":"1 Introduction/1.2 Our contributions (paragraph 2); 5 Discussion (paragraph 1)"},{"label":"Outputs","value":"Task-specific classifications or generated DNA sequences","status":"source_checked","source_ids":["enbed-2024"],"source_locator":"2 Methods/2.4 Applications of foundation models using transfer learning/2.4.2 Fine-tuning for downstream tasks (paragraph 1); 2 Methods/2.5 Application domains/2.5.1 Genomic benchmarks (paragraph 1)"},{"label":"Parameters","value":"1.2 billion trainable parameters in the full encoder–decoder model.","status":"source_checked","source_ids":["enbed-2024"],"source_locator":"4 Ablation studies (paragraph 1); 4 Ablation studies/4.1 Encoder–decoder architecture (paragraph 1)"},{"label":"Known versions / configuration","value":"ENBED is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["enbed-2024"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"Reference-genome sequences; the GRCh38-labelled row is a distinct human-reference configuration.","status":"source_checked","source_ids":["enbed-2024"],"source_locator":"1 Introduction/1.2 Our contributions/1.2.1 Evaluation of performance on genomic benchmark datasets (paragraph 1); 3 Results/3.1 ENBED outperforms state-of-the-art models on GB datasets (paragraph 2)"},{"label":"Context limits","value":"16,384 input/output tokens using local sliding-window plus global attention. The 512-token value in Methods describes the dense-attention hardware baseline, not ENBED’s final context.","status":"source_checked","source_ids":["enbed-2024"],"source_locator":"2 Methods/2.3 Attention (paragraph 2); 2 Methods/2.3 Attention/2.3.1 Sliding-window attention (paragraph 1)"},{"label":"Access","value":"The authors provide implementation code at https://github.itap.purdue.edu/Clan-labs/ENBED and model weights through https://huggingface.co/malusare. A table-specific checkpoint hash is not supplied by these account-level links.","status":"source_checked","source_ids":["enbed-2024"],"source_locator":"Data availability"},{"label":"Code licence","value":"No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","status":"unreported","source_ids":["enbed-2024"],"source_locator":"Abstract (paragraph 1); 2 Methods/2.4 Applications of foundation models using transfer learning/2.4.1 Building the foundation model (paragraph 1)"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["enbed-2024"],"source_locator":"Abstract (paragraph 1); 2 Methods/2.1 Encoder–decoder model architecture (paragraph 1)"}],"strengths":[{"text":"Single-base input resolution retains substitutions and indels that can be obscured by multi-base token boundaries.","source_ids":["enbed-2024"],"source_locator":"1 Introduction/1.1 Limitations of previous work/1.1.2 Tokenization (paragraph 1); 3 Results/3.4 ENBED generates mutations using sequence-to-sequence transformation (paragraph 1)"}],"limitations":[{"text":"The evaluated downstream head and training corpus must be kept with each result; the GRCh38 configuration is not interchangeable with every ENBED model.","source_ids":["enbed-2024"],"source_locator":"2 Methods/2.4 Applications of foundation models using transfer learning/2.4.2 Fine-tuning for downstream tasks (paragraph 1); 3 Results/3.3 ENBED identifies biological function annotations (paragraph 1)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["DNA sequences at single-byte nucleotide resolution","ENBED","Task-specific classifications or generated DNA sequences"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["enbed-2024"],"source_locator":"1 Introduction/1.2 Our contributions (paragraph 1); 4 Ablation studies/4.1 Encoder–decoder architecture (paragraph 2)"},"coverage":"limited","gaps":["ENBED is the comparison-table label; that label does not specify an immutable weight revision.","No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-953007693fb72a","kind":"model","name":"HyenaDNA","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["dna-genomes"]},"source_ids":["polya-glm-2025"],"links":[],"attributes":{"entity_level":"method","version":"not stated in table","reported_name":"HyenaDNA","historical_missing_metadata":{"checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"This HyenaDNA pipeline predicts polyadenylation sites from genomic sequence.","summary_source_ids":["polya-glm-2025"],"summary_source_locator":"Methods/Genome-wide discovery of de novo poly(A) sites using fine-tuned genome language models (paragraph 1); Methods (paragraph 2)","sections":[{"title":"How the evaluated method works","body":"The study evaluates both few-shot classification and fine-tuning, and also builds a position-wise token classifier for extended gene regions.","source_ids":["polya-glm-2025"],"source_locator":"Abstract (paragraph 1); Results/Model performance: few-shot capabilities and fine-tuning improvements (paragraph 1)"},{"title":"Underlying method and version boundaries","body":"HyenaDNA models DNA at single-nucleotide resolution and publishes checkpoints with different context lengths. The family’s longest advertised context must not be assigned to a shorter-context checkpoint or to every downstream experiment.","source_ids":["evidence-reported-base-hyenadna-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies HyenaDNA: polyadenylation site detection. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["polya-glm-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-b2-polya-glm-2025"}],"facts":[{"label":"Model type","value":"Long-context DNA model; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-hyenadna-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"The study evaluates both few-shot classification and fine-tuning, and also builds a position-wise token classifier for extended gene regions.","status":"source_checked","source_ids":["polya-glm-2025"],"source_locator":"Abstract (paragraph 1); Results/Model performance: few-shot capabilities and fine-tuning improvements (paragraph 1)"},{"label":"Biological inputs","value":"DNA windows around candidate polyadenylation/cleavage sites","status":"source_checked","source_ids":["polya-glm-2025"],"source_locator":"Results/Genome-wide prediction of novel poly(A) sites (paragraph 4); Results/Genome-wide prediction of novel poly(A) sites (paragraph 3)"},{"label":"Outputs","value":"Sequence-level or position-wise poly(A)-site predictions","status":"source_checked","source_ids":["polya-glm-2025"],"source_locator":"Results/Model interpretability: perturbation and attention analysis validate biological learning (paragraph 1); Abstract (paragraph 1)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["polya-glm-2025","evidence-reported-base-hyenadna-readme-md"],"source_locator":"Methods; Methods/An end-to-end framework for poly(A) site classification with genome language models; Methods/Genome-wide discovery of de novo poly(A) sites using fine-tuned genome language models; Methods/Fine-grained poly(A) site identification through token-level modeling; Methods/Validating biological feature learning through interpretability analysis; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"HyenaDNA is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["polya-glm-2025"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"GENCODE v38 positive poly(A) annotations and motif-matched negative sequences form balanced Gene-Gene and Gene-Intergene datasets, each with 50,182 positive and 50,182 negative examples. Five folds use 60% training,20% validation and 20% test; few-shot prototypes and full fine-tuning are distinct settings.","status":"source_checked","source_ids":["polya-glm-2025"],"source_locator":"Methods/An end-to-end framework for poly(A) site classification with genome language models (paragraph 1); Results/Model performance: few-shot capabilities and fine-tuning improvements (paragraph 2)"},{"label":"Context limits","value":"101-bp windows, with position 51 explicitly set to adenine in both positive and negative examples.","status":"source_checked","source_ids":["polya-glm-2025"],"source_locator":"Methods/An end-to-end framework for poly(A) site classification with genome language models (paragraph 1); Methods/An end-to-end framework for poly(A) site classification with genome language models (paragraph 2)"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://github.com/HazyResearch/hyena-dna/blob/d553021b483b82980aa4b868b37ec2d4332e198a/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-hyenadna-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"Apache 2.0 (upstream repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-base-hyenadna-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-base-hyenadna-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Signal-perturbation experiments examine recognition of canonical polyadenylation motifs and their location relative to cleavage sites.","source_ids":["polya-glm-2025"],"source_locator":"Results/Model performance: few-shot capabilities and fine-tuning improvements (paragraph 2); Abstract (paragraph 1)"}],"limitations":[{"text":"Few-shot and fine-tuned results are separate configurations; motif sensitivity alone is not experimental confirmation of a cleavage event.","source_ids":["polya-glm-2025"],"source_locator":"Results/Genome-wide prediction of novel poly(A) sites (paragraph 3); Methods/Validating biological feature learning through interpretability analysis (paragraph 1)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["DNA windows around candidate polyadenylation/cleavage sites","HyenaDNA","Sequence-level or position-wise poly(A)-site predictions"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["polya-glm-2025"],"source_locator":"Abstract (paragraph 1); Results/Model performance: few-shot capabilities and fine-tuning improvements (paragraph 1)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","HyenaDNA is the comparison-table label; that label does not specify an immutable weight revision.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-9a200c55b0e03e","kind":"model","name":"iPro-MP","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["microbes-communities"]},"source_ids":["ipromp-2025"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"iPro-MP","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"iPro-MP is a transformer-based prokaryotic promoter classifier evaluated across 23 species.","summary_source_ids":["ipromp-2025"],"summary_source_locator":"Discussion (paragraph 1); Results/iPro-MP outperforms classical and deep learning baselines across species (paragraph 4)","sections":[{"title":"How the evaluated method works","body":"Multi-head attention models DNA-sequence patterns in a species-specific promoter-prediction framework.","source_ids":["ipromp-2025"],"source_locator":"Discussion (paragraph 1); Methods/DNABERT model (paragraph 3)"},{"title":"What was evaluated","body":"The linked evaluation record identifies iPro-MP: Multi-species prokaryotic promoter detection. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["ipromp-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b3-033"}],"facts":[{"label":"Model type","value":"Study-specific predictive method; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["ipromp-2025"],"source_locator":"Discussion (paragraph 1); Methods/DNABERT model (paragraph 3)"},{"label":"Architecture / procedure","value":"Multi-head attention models DNA-sequence patterns in a species-specific promoter-prediction framework.","status":"source_checked","source_ids":["ipromp-2025"],"source_locator":"Discussion (paragraph 1); Methods/DNABERT model (paragraph 3)"},{"label":"Biological inputs","value":"81-bp DNA windows from −60 to +20 relative to the transcription start site","status":"source_checked","source_ids":["ipromp-2025"],"source_locator":"Background (paragraph 1); Results/iPro-MP exhibits excellent performance and robustness in multi-species promoter prediction (paragraph 4)"},{"label":"Outputs","value":"Promoter versus non-promoter classification","status":"source_checked","source_ids":["ipromp-2025"],"source_locator":"Results/iPro-MP outperforms classical and deep learning baselines across species (paragraph 1); Results/iPro-MP outperforms existing tools in multi-species promoter prediction (paragraph 3)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["ipromp-2025","evidence-reported-ipromp-2025-readme-md"],"source_locator":"Methods; Methods/Data collection and preprocessing; Methods/DNABERT model; Methods/Fine-tuning of DNABERT model; Methods/Performance evaluation; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"iPro-MP is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["ipromp-2025"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"107,286 promoter sequences from the Prokaryotic Promoter Database, filtered at 0.8 sequence identity; species selection generally requires more than 1,000 validated promoters.","status":"source_checked","source_ids":["ipromp-2025"],"source_locator":"Methods/Data collection and preprocessing (paragraph 1); Results/iPro-MP reveals the species-specificity at the sequential level (paragraph 5)"},{"label":"Context limits","value":"81 nucleotides per example.","status":"source_checked","source_ids":["ipromp-2025"],"source_locator":"Results/iPro-MP outperforms classical and deep learning baselines across species (paragraph 3); Methods/DNABERT model (paragraph 4)"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/Jackie-Suv/iPro-MP/blob/4266b521bc6617db939c5871cb1b6850dff63fdb/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-ipromp-2025-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","status":"unreported","source_ids":["evidence-reported-ipromp-2025-readme-md"],"source_locator":"README.md and repository-root licence-file search"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-ipromp-2025-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Includes model and non-model prokaryotic species with explicit cross-species tests.","source_ids":["ipromp-2025"],"source_locator":"Results/iPro-MP reveals the species-specificity at the sequential level (paragraph 1); Results/iPro-MP outperforms classical and deep learning baselines across species (paragraph 1)"}],"limitations":[{"text":"Cross-species experiments motivate species-specific models; a strong within-species result is not evidence of universal promoter recognition.","source_ids":["ipromp-2025"],"source_locator":"Results/iPro-MP reveals the species-specificity at the sequential level (paragraph 5); Results/iPro-MP reveals the species-specificity at the sequential level (paragraph 1)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["81-bp DNA windows from −60 to +20 relative to the transcription start site","iPro-MP","Promoter versus non-promoter classification"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["ipromp-2025"],"source_locator":"Discussion (paragraph 1); Methods/DNABERT model (paragraph 3)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","iPro-MP is the comparison-table label; that label does not specify an immutable weight revision.","No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-9b3bc255532dd3","kind":"model","name":"HyenaDNA","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["dna-genomes"]},"source_ids":["dnalongbench-2025"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"HyenaDNA","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"This fine-tuned genomic model is evaluated on long-range DNA tasks in DNALongBench.","summary_source_ids":["dnalongbench-2025"],"summary_source_locator":"Proposed Dataset: DNALongBench (paragraph 3); Introduction (paragraph 5)","sections":[{"title":"How the evaluated method works","body":"The DNA backbone is fine-tuned for task-specific outputs, with expert models and a CNN as separate controls.","source_ids":["dnalongbench-2025"],"source_locator":"Experiments/Benchmarking Results/The Expert Model achieves the highest scores on all tasks. (paragraph 1); Conclusion (paragraph 1)"},{"title":"Underlying method and version boundaries","body":"HyenaDNA models DNA at single-nucleotide resolution and publishes checkpoints with different context lengths. The family’s longest advertised context must not be assigned to a shorter-context checkpoint or to every downstream experiment.","source_ids":["evidence-reported-base-hyenadna-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies HyenaDNA: Enhancer-target gene prediction. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["dnalongbench-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-007"}],"facts":[{"label":"Model type","value":"Long-context DNA model; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-hyenadna-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"The DNA backbone is fine-tuned for task-specific outputs, with expert models and a CNN as separate controls.","status":"source_checked","source_ids":["dnalongbench-2025"],"source_locator":"Experiments/Benchmarking Results/The Expert Model achieves the highest scores on all tasks. (paragraph 1); Conclusion (paragraph 1)"},{"label":"Biological inputs","value":"DNA windows for enhancer–gene, eQTL, chromatin organisation, activity or initiation tasks","status":"source_checked","source_ids":["dnalongbench-2025"],"source_locator":"Conclusion (paragraph 1); Abstract (paragraph 1)"},{"label":"Outputs","value":"Task-specific long-range genomic predictions","status":"source_checked","source_ids":["dnalongbench-2025"],"source_locator":"Introduction (paragraph 5); Proposed Dataset: DNALongBench (paragraph 1)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["dnalongbench-2025","evidence-reported-base-hyenadna-readme-md"],"source_locator":"Complete primary text and named comparison table; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"HyenaDNA is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["dnalongbench-2025"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"Task-specific fine-tuning on the DNALongBench splits, separate from original genome pretraining.","status":"source_checked","source_ids":["dnalongbench-2025"],"source_locator":"Proposed Dataset: DNALongBench (paragraph 4); Proposed Dataset: DNALongBench/3D Chromatin Contact Map Prediction (paragraph 1)"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["dnalongbench-2025","evidence-reported-base-hyenadna-readme-md"],"source_locator":"Complete primary text and named comparison table; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://github.com/HazyResearch/hyena-dna/blob/d553021b483b82980aa4b868b37ec2d4332e198a/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-hyenadna-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"Apache 2.0 (upstream repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-base-hyenadna-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-base-hyenadna-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Tests genomic tasks that require information over long distances.","source_ids":["dnalongbench-2025"],"source_locator":"Proposed Dataset: DNALongBench (paragraph 3); Introduction (paragraph 4)"}],"limitations":[{"text":"Nominal long input support does not prove effective long-range reasoning; specialist models outperform the tested foundation models in the reported benchmark.","source_ids":["dnalongbench-2025"],"source_locator":"Proposed Dataset: DNALongBench (paragraph 3); Introduction (paragraph 5)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["DNA windows for enhancer–gene, eQTL, chromatin organisation, activity or initiation tasks","HyenaDNA","Task-specific long-range genomic predictions"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["dnalongbench-2025"],"source_locator":"Experiments/Benchmarking Results/The Expert Model achieves the highest scores on all tasks. (paragraph 1); Conclusion (paragraph 1)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","HyenaDNA is the comparison-table label; that label does not specify an immutable weight revision.","A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-9c10fbec02a365","kind":"model","name":"FUJISAN","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["proteins-complexes"]},"source_ids":["fujisan-2024"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"FUJISAN","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"FUJISAN predicts whether two proteins catalyse the same enzymatic reaction using sequence, domain and pocket similarities.","summary_source_ids":["fujisan-2024"],"summary_source_locator":"Abstract (paragraph 1); Results and discussion/Limitations and perspectives (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"Features include full-sequence similarity, domain structural similarity and pocket similarity from AlphaFold2 structural models. A LightGBM classifier combines these comparisons.","source_ids":["fujisan-2024"],"source_locator":"Materials and methods/Feature engineering/Pocket similarity features (paragraph 1); Conclusions (paragraph 1)"},{"title":"What was evaluated","body":"The linked evaluation record identifies FUJISAN: Enzyme functional identity prediction. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["fujisan-2024"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-019"}],"facts":[{"label":"Model type","value":"Gradient-boosted-tree pipeline; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["fujisan-2024"],"source_locator":"Materials and methods/Feature engineering/Pocket similarity features (paragraph 1); Conclusions (paragraph 1)"},{"label":"Architecture / procedure","value":"Features include full-sequence similarity, domain structural similarity and pocket similarity from AlphaFold2 structural models. A LightGBM classifier combines these comparisons.","status":"source_checked","source_ids":["fujisan-2024"],"source_locator":"Materials and methods/Feature engineering/Pocket similarity features (paragraph 1); Conclusions (paragraph 1)"},{"label":"Biological inputs","value":"Pairs of protein sequences and predicted structures with domain and pocket decomposition","status":"source_checked","source_ids":["fujisan-2024"],"source_locator":"Materials and methods/Dataset construction (paragraph 1); Results and discussion/Limitations and perspectives (paragraph 1)"},{"label":"Outputs","value":"Same-reaction versus different-reaction classification","status":"source_checked","source_ids":["fujisan-2024"],"source_locator":"Materials and methods/Model training and hyperparameter optimization (paragraph 1); Materials and methods/Feature importance/Gini importance (paragraph 1)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["fujisan-2024","evidence-reported-fujisan-2024-readme-md"],"source_locator":"Materials and methods/Dataset construction; Materials and methods/Feature engineering/Full-length sequence similarity features; Materials and methods/Feature engineering/Domain structural similarity features; Materials and methods/Feature engineering/Pocket similarity features; Materials and methods/Model training and hyperparameter optimization; Materials and methods/Performance assessment; Materials and methods/Prediction with DeepFRI; Materials and methods/Prediction with ESM-2; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"FUJISAN is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["fujisan-2024"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"20,800 same-function and 20,800 different-function pairs; train/validation/test proportions are 56.25%, 18.75% and 25%.","status":"source_checked","source_ids":["fujisan-2024"],"source_locator":"Materials and methods/Model training and hyperparameter optimization (paragraph 1); Results and discussion/Performance on low sequence similarity dataset (paragraph 1)"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["fujisan-2024","evidence-reported-fujisan-2024-readme-md"],"source_locator":"Materials and methods/Dataset construction; Materials and methods/Feature engineering/Full-length sequence similarity features; Materials and methods/Feature engineering/Domain structural similarity features; Materials and methods/Feature engineering/Pocket similarity features; Materials and methods/Model training and hyperparameter optimization; Materials and methods/Performance assessment; Materials and methods/Prediction with DeepFRI; Materials and methods/Prediction with ESM-2; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/sfujita0601/FUJISAN/blob/588daf65810c49b021af9e47b11ec752986dcb94/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-fujisan-2024-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","status":"unreported","source_ids":["evidence-reported-fujisan-2024-readme-md"],"source_locator":"README.md and repository-root licence-file search"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-fujisan-2024-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Combines sequence and structure information and exposes feature importance for the trained classifier.","source_ids":["fujisan-2024"],"source_locator":"Materials and methods/Feature importance/Gini importance (paragraph 1); Materials and methods/Feature importance/Shapley additive explanations (paragraph 1)"}],"limitations":[{"text":"Pairs with uncomputable features are excluded, and the random-pair split must not be interpreted as a guaranteed unseen-protein split.","source_ids":["fujisan-2024"],"source_locator":"Materials and methods/Feature engineering/Full-length sequence similarity features (paragraph 1); Materials and methods/Model training and hyperparameter optimization (paragraph 1)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Pairs of protein sequences and predicted structures with domain and pocket decomposition","FUJISAN","Same-reaction versus different-reaction classification"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["fujisan-2024"],"source_locator":"Materials and methods/Feature engineering/Pocket similarity features (paragraph 1); Conclusions (paragraph 1)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","FUJISAN is the comparison-table label; that label does not specify an immutable weight revision.","A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-9f39de53f7a139","kind":"model","name":"scXDR","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["cells-tissues"]},"source_ids":["scxdr-2026"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"scXDR","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"scXDR predicts single-cell drug responses through heterogeneous-network transfer learning.","summary_source_ids":["scxdr-2026"],"summary_source_locator":"Abstract (paragraph 2); Methods/Comparison experiments at the individual cell level (paragraph 2)","sections":[{"title":"How the evaluated method works","body":"Drug, gene and cell nodes exchange messages; feature and structure alignment, reconstruction and drug–cell scoring support transfer between single-cell datasets.","source_ids":["scxdr-2026"],"source_locator":"Methods/Model construction (paragraph 4); Methods/Comparison experiments at the individual cell level (paragraph 2)"},{"title":"What was evaluated","body":"The linked evaluation record identifies scXDR: Cross-dataset single-cell drug response transfer. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["scxdr-2026"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b3-017"}],"facts":[{"label":"Model type","value":"Study-specific predictive method; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["scxdr-2026"],"source_locator":"Methods/Model construction (paragraph 4); Methods/Comparison experiments at the individual cell level (paragraph 2)"},{"label":"Architecture / procedure","value":"Drug, gene and cell nodes exchange messages; feature and structure alignment, reconstruction and drug–cell scoring support transfer between single-cell datasets.","status":"source_checked","source_ids":["scxdr-2026"],"source_locator":"Methods/Model construction (paragraph 4); Methods/Comparison experiments at the individual cell level (paragraph 2)"},{"label":"Biological inputs","value":"Single-cell expression profiles, drug/gene features and heterogeneous graph relationships","status":"source_checked","source_ids":["scxdr-2026"],"source_locator":"Abstract (paragraph 2); Methods/Comparison experiments at the individual cell level (paragraph 2)"},{"label":"Outputs","value":"Drug-response scores for cells","status":"source_checked","source_ids":["scxdr-2026"],"source_locator":"Methods/Case study (paragraph 4); Methods/Comparison experiments at the individual cell level (paragraph 3)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["scxdr-2026","evidence-reported-scxdr-2026-readme-md"],"source_locator":"Results/Superior performance at the individual cell level compared to various methods; Results/Superior performance at the cell group level compared to various methods; Results/Contribution of model components and architecture to performance; Methods/Data collection and processing; Methods/Model construction; Methods/Experiment setting; Methods/Comparison experiments at the individual cell level; Methods/Comparison experiments at the cell group level; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"scXDR is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["scxdr-2026"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"Twelve GEO scRNA-seq datasets across six tumour types and ten drugs support 20 cross-dataset prediction tasks. Each task transfers from its specified source dataset to a separate target dataset.","status":"source_checked","source_ids":["scxdr-2026"],"source_locator":"Introduction (paragraph 4); Methods/Data collection and processing (paragraph 1)"},{"label":"Context limits","value":"Cell features use 5,000 highly variable genes; drug and target features use molecular and protein descriptors.","status":"source_checked","source_ids":["scxdr-2026"],"source_locator":"Methods/Data collection and processing (paragraph 2); Abstract (paragraph 2)"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/QiGuan1920/scXDR2025/blob/5b39f37ba4df186eeea0881d59366458d2535db9/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-scxdr-2026-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","status":"unreported","source_ids":["evidence-reported-scxdr-2026-readme-md"],"source_locator":"README.md and repository-root licence-file search"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-scxdr-2026-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Models cross-dataset transfer using single-cell information rather than simply assigning a bulk cell-line label to every cell.","source_ids":["scxdr-2026"],"source_locator":"Methods/Comparison experiments at the individual cell level (paragraph 2); Introduction (paragraph 4)"}],"limitations":[{"text":"Response labels and cross-dataset assumptions remain important; a computational prediction is not an experimentally confirmed response or a clinical treatment recommendation.","source_ids":["scxdr-2026"],"source_locator":"Methods/Data collection and processing (paragraph 1); Methods/Case study (paragraph 5)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Single-cell expression profiles, drug/gene features and heterogeneous graph relationships","scXDR","Drug-response scores for cells"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["scxdr-2026"],"source_locator":"Methods/Model construction (paragraph 4); Methods/Comparison experiments at the individual cell level (paragraph 2)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","scXDR is the comparison-table label; that label does not specify an immutable weight revision.","No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-a0db32ae53e5ed","kind":"model","name":"scLLMDA","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["cells-tissues"]},"source_ids":["scatac-llmda-2026"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"scLLMDA","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"scLLMDA annotates scATAC-seq cells using peak-sequence embeddings and graph-based domain adaptation.","summary_source_ids":["scatac-llmda-2026"],"summary_source_locator":"Discussion (paragraph 1); Introduction (paragraph 4)","sections":[{"title":"How the evaluated method works","body":"A pretrained DNA model encodes peak sequences; accessibility values form cell representations. Source and target similarity graphs enter a graph neural network that aligns domains while retaining local neighbourhoods.","source_ids":["scatac-llmda-2026"],"source_locator":"Materials and methods/Cell type annotation via graph domain adaptation/Capture the local consistency relationship of each graph. (paragraph 4); Materials and methods/Loss function (paragraph 1)"},{"title":"What was evaluated","body":"The linked evaluation record identifies scLLMDA: Cross-platform scATAC cell-type annotation. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["scatac-llmda-2026"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b3-005"}],"facts":[{"label":"Model type","value":"Graph-based predictive method; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["scatac-llmda-2026"],"source_locator":"Materials and methods/Cell type annotation via graph domain adaptation/Capture the local consistency relationship of each graph. (paragraph 4); Materials and methods/Loss function (paragraph 1)"},{"label":"Architecture / procedure","value":"A pretrained DNA model encodes peak sequences; accessibility values form cell representations. Source and target similarity graphs enter a graph neural network that aligns domains while retaining local neighbourhoods.","status":"source_checked","source_ids":["scatac-llmda-2026"],"source_locator":"Materials and methods/Cell type annotation via graph domain adaptation/Capture the local consistency relationship of each graph. (paragraph 4); Materials and methods/Loss function (paragraph 1)"},{"label":"Biological inputs","value":"scATAC-seq accessibility values, peak DNA sequences and labelled reference cells","status":"source_checked","source_ids":["scatac-llmda-2026"],"source_locator":"Materials and methods/Feature extraction from genomic sequences (paragraph 5); Materials and methods/Feature extraction from genomic sequences (paragraph 4)"},{"label":"Outputs","value":"Cell-type annotations in a target scATAC-seq dataset","status":"source_checked","source_ids":["scatac-llmda-2026"],"source_locator":"Materials and methods/Problem definition (paragraph 1); Results/Effectiveness of GDA module (paragraph 3)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["scatac-llmda-2026","evidence-reported-scatac-llmda-2026-readme-md"],"source_locator":"Materials and methods/Benchmark datasets; Materials and methods/Benchmark methods; Materials and methods/Problem definition; Materials and methods/Feature extraction from genomic sequences; Materials and methods/Cell type annotation via graph domain adaptation/Graph construction.; Materials and methods/Cell type annotation via graph domain adaptation/Capture the local consistency relationship of each graph.; Materials and methods/Cell type annotation via graph domain adaptation/Capture the global consistency relationship of each graph.; Materials and methods/Cell type annotation via graph domain adaptation/Feature fusion via attention.; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"scLLMDA is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["scatac-llmda-2026"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"Cross-dataset cell annotation uses mouse snATAC-seq/ATAC-seq datasets GSE126724 and GSE111586 plus10xMouseBrain; assemblies differ (GRCm38/mm10 versus mm9) and remain dataset-specific.","status":"source_checked","source_ids":["scatac-llmda-2026"],"source_locator":"Materials and methods/Benchmark datasets (paragraph 1); Results/Cross-platform cell type annotation (paragraph 1)"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["scatac-llmda-2026","evidence-reported-scatac-llmda-2026-readme-md"],"source_locator":"Materials and methods/Benchmark datasets; Materials and methods/Benchmark methods; Materials and methods/Problem definition; Materials and methods/Feature extraction from genomic sequences; Materials and methods/Cell type annotation via graph domain adaptation/Graph construction.; Materials and methods/Cell type annotation via graph domain adaptation/Capture the local consistency relationship of each graph.; Materials and methods/Cell type annotation via graph domain adaptation/Capture the global consistency relationship of each graph.; Materials and methods/Cell type annotation via graph domain adaptation/Feature fusion via attention.; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/sheng-guan-2001/scLLMDA/blob/5e24025710bb068312d50a5749ef6bb838ef5a32/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-scatac-llmda-2026-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","status":"unreported","source_ids":["evidence-reported-scatac-llmda-2026-readme-md"],"source_locator":"README.md and repository-root licence-file search"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-scatac-llmda-2026-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Uses both sequence features and cell-neighbourhood context for transfer within the ATAC modality.","source_ids":["scatac-llmda-2026"],"source_locator":"Materials and methods/Cell type annotation via graph domain adaptation/Feature fusion via attention. (paragraph 6); Materials and methods/Cell type annotation via graph domain adaptation/Feature fusion via attention. (paragraph 1)"}],"limitations":[{"text":"Access to target-cell structure is part of the adaptation protocol; it is not equivalent to a purely inductive sequence-only test.","source_ids":["scatac-llmda-2026"],"source_locator":"Materials and methods/Parameter settings (paragraph 1); Materials and methods/Cell type annotation via graph domain adaptation/Feature fusion via attention. (paragraph 1)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["scATAC-seq accessibility values, peak DNA sequences and labelled reference cells","scLLMDA","Cell-type annotations in a target scATAC-seq dataset"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["scatac-llmda-2026"],"source_locator":"Materials and methods/Cell type annotation via graph domain adaptation/Capture the local consistency relationship of each graph. (paragraph 4); Materials and methods/Loss function (paragraph 1)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","scLLMDA is the comparison-table label; that label does not specify an immutable weight revision.","A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-a29203c09857ef","kind":"model","name":"MULAN-ESM2 S","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["proteins-complexes"]},"source_ids":["mulan-2025"],"links":[],"attributes":{"entity_level":"method","version":"small ESM2 backbone","reported_name":"MULAN-ESM2 S","historical_missing_metadata":{"checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"MULAN-ESM2 S augments a pretrained ESM-2 sequence encoder with a structure adapter.","summary_source_ids":["mulan-2025"],"summary_source_locator":"2 Methods/2.1 MULAN architecture/2.1.1 Structural information (paragraph 1); 3 Discussion/3.2 Ablation study/3.2.1 The importance of the Structure Adapter (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"A parameter-efficient adapter encodes angle-based structural information, which is fused with the pretrained sequence representation and trained jointly.","source_ids":["mulan-2025"],"source_locator":"2 Methods/2.2 Training procedure and structure masking (paragraph 1); 2 Methods/2.1 MULAN architecture/2.1.2 The Structure Adapter (paragraph 1)"},{"title":"Underlying method and version boundaries","body":"ESM-2 is a transformer protein language-model family. The official repository exposes residue embeddings, sequence-level pooling and models at several sizes; the study configuration determines which of these is evaluated.","source_ids":["evidence-reported-base-esm-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies MULAN-ESM2 S: human protein-protein interaction prediction. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["mulan-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-b2-mulan-2025"}],"facts":[{"label":"Model type","value":"Protein sequence transformer; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-esm-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"A parameter-efficient adapter encodes angle-based structural information, which is fused with the pretrained sequence representation and trained jointly.","status":"source_checked","source_ids":["mulan-2025"],"source_locator":"2 Methods/2.2 Training procedure and structure masking (paragraph 1); 2 Methods/2.1 MULAN architecture/2.1.2 The Structure Adapter (paragraph 1)"},{"label":"Biological inputs","value":"Protein amino-acid sequences and angle-based structural descriptors","status":"source_checked","source_ids":["mulan-2025"],"source_locator":"2 Methods/2.1 MULAN architecture/2.1.1 Structural information (paragraph 2); 3 Discussion/3.4 Related work/3.4.2 Structure-informed models (paragraph 1)"},{"label":"Outputs","value":"Structure-aware representations for downstream protein tasks","status":"source_checked","source_ids":["mulan-2025"],"source_locator":"3 Discussion/3.1 Results (paragraph 1); 4 Conclusion (paragraph 1)"},{"label":"Parameters","value":"MULAN-ESM2 S has 9M parameters, including its ESM2-8M backbone and structure adapter. The 37M and 652M models are separate variants.","status":"source_checked","source_ids":["mulan-2025"],"source_locator":"3 Discussion/3.2 Ablation study/3.2.1 The importance of the Structure Adapter (paragraph 1); 2 Methods/2.1 MULAN architecture/2.1.1 Structural information (paragraph 1)"},{"label":"Known versions / configuration","value":"small ESM2 backbone","status":"source_checked","source_ids":["mulan-2025"],"source_locator":"Table vbaf117-T4 (paragraph 1); Table vbaf117-T2 (paragraph 1)"},{"label":"Training data / fitting","value":"501,348 Swiss-Prot proteins with AlphaFold2 structures after filtering sequences shorter than 30 residues; 5,000 proteins are selected for validation. The pretrained backbone is further trained with joint sequence/angle masking.","status":"source_checked","source_ids":["mulan-2025"],"source_locator":"2 Methods/2.3 Training details (paragraph 1); 2 Methods/2.2 Training procedure and structure masking (paragraph 1)"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["mulan-2025","evidence-reported-base-esm-readme-md"],"source_locator":"2 Methods/2.1 MULAN architecture/2.1.1 Structural information; 2 Methods/2.1 MULAN architecture/2.1.2 The Structure Adapter; 2 Methods/2.2 Training procedure and structure masking; 2 Methods/2.3 Training details; 2 Methods/2.4 Downstream tasks; 2 Methods/2.5 Downstream task evaluation; 3 Discussion/3.2 Ablation study/3.2.4 Architecture ablation; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://github.com/facebookresearch/esm/blob/2b369911bb5b4b0dda914521b9475cad1656b2ac/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-esm-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"MIT (upstream repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-base-esm-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-base-esm-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Adds structural information by adapting existing language-model weights rather than training a new sequence model from scratch.","source_ids":["mulan-2025"],"source_locator":"3 Discussion/3.2 Ablation study/3.2.3 Starting from the pre-trained model is necessary (paragraph 1); 1 Introduction (paragraph 3)"}],"limitations":[{"text":"Requires structural information beyond sequence; comparisons to sequence-only encoders must preserve that input difference.","source_ids":["mulan-2025"],"source_locator":"1 Introduction (paragraph 3); 2 Methods/2.1 MULAN architecture/2.1.2 The Structure Adapter (paragraph 1)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Protein amino-acid sequences and angle-based structural descriptors","MULAN-ESM2 S","Structure-aware representations for downstream protein tasks"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["mulan-2025"],"source_locator":"2 Methods/2.2 Training procedure and structure masking (paragraph 1); 2 Methods/2.1 MULAN architecture/2.1.2 The Structure Adapter (paragraph 1)"},"coverage":"limited","gaps":["A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-a7cfacf25d97ad","kind":"model","name":"Caduceus-Ph","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["dna-genomes"]},"source_ids":["dnalongbench-2025"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"Caduceus-Ph","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"This fine-tuned genomic model is evaluated on long-range DNA tasks in DNALongBench.","summary_source_ids":["dnalongbench-2025"],"summary_source_locator":"Proposed Dataset: DNALongBench (paragraph 3); Introduction (paragraph 5)","sections":[{"title":"How the evaluated method works","body":"The DNA backbone is fine-tuned for task-specific outputs, with expert models and a CNN as separate controls.","source_ids":["dnalongbench-2025"],"source_locator":"Experiments/Benchmarking Results/The Expert Model achieves the highest scores on all tasks. (paragraph 1); Conclusion (paragraph 1)"},{"title":"Underlying method and version boundaries","body":"Caduceus exposes distinct Ph and PS configurations. The documented Ph-131k checkpoint uses 16 layers, width 256 and reverse-complement data augmentation; PS implements reverse-complement equivariance. These training choices are not interchangeable.","source_ids":["evidence-reported-base-caduceus-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies Caduceus-Ph: Enhancer-target gene prediction. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["dnalongbench-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-008"}],"facts":[{"label":"Model type","value":"DNA state-space model; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-caduceus-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"The DNA backbone is fine-tuned for task-specific outputs, with expert models and a CNN as separate controls.","status":"source_checked","source_ids":["dnalongbench-2025"],"source_locator":"Experiments/Benchmarking Results/The Expert Model achieves the highest scores on all tasks. (paragraph 1); Conclusion (paragraph 1)"},{"label":"Biological inputs","value":"DNA windows for enhancer–gene, eQTL, chromatin organisation, activity or initiation tasks","status":"source_checked","source_ids":["dnalongbench-2025"],"source_locator":"Conclusion (paragraph 1); Abstract (paragraph 1)"},{"label":"Outputs","value":"Task-specific long-range genomic predictions","status":"source_checked","source_ids":["dnalongbench-2025"],"source_locator":"Introduction (paragraph 5); Proposed Dataset: DNALongBench (paragraph 1)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["dnalongbench-2025","evidence-reported-base-caduceus-readme-md"],"source_locator":"Complete primary text and named comparison table; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"Caduceus-Ph is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["dnalongbench-2025"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"Task-specific fine-tuning on the DNALongBench splits, separate from original genome pretraining.","status":"source_checked","source_ids":["dnalongbench-2025"],"source_locator":"Proposed Dataset: DNALongBench (paragraph 4); Proposed Dataset: DNALongBench/3D Chromatin Contact Map Prediction (paragraph 1)"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["dnalongbench-2025","evidence-reported-base-caduceus-readme-md"],"source_locator":"Complete primary text and named comparison table; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://github.com/kuleshov-group/caduceus/blob/0060a6d8079b6a040fc55d505e15972a327b70a6/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-caduceus-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"Apache 2.0 (upstream repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-base-caduceus-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-base-caduceus-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Tests genomic tasks that require information over long distances.","source_ids":["dnalongbench-2025"],"source_locator":"Proposed Dataset: DNALongBench (paragraph 3); Introduction (paragraph 4)"}],"limitations":[{"text":"Nominal long input support does not prove effective long-range reasoning; specialist models outperform the tested foundation models in the reported benchmark.","source_ids":["dnalongbench-2025"],"source_locator":"Proposed Dataset: DNALongBench (paragraph 3); Introduction (paragraph 5)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["DNA windows for enhancer–gene, eQTL, chromatin organisation, activity or initiation tasks","Caduceus-Ph","Task-specific long-range genomic predictions"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["dnalongbench-2025"],"source_locator":"Experiments/Benchmarking Results/The Expert Model achieves the highest scores on all tasks. (paragraph 1); Conclusion (paragraph 1)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","Caduceus-Ph is the comparison-table label; that label does not specify an immutable weight revision.","A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-aa763db2cfdeff","kind":"model","name":"EVO2","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["microbes-communities"]},"source_ids":["lambda-prophage-2026"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"EVO2","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"This configuration is evaluated for phage–bacterium discrimination and genome-wide prophage detection in LAMBDA.","summary_source_ids":["lambda-prophage-2026"],"summary_source_locator":"Materials and Methods/LAMBDA Benchmark Dataset Construction. (paragraph 1); Abstract (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"The benchmark separates embedding probes, fine-tuning, diagnostic tests and genome-wide scans; the associated result identifies which procedure is used.","source_ids":["lambda-prophage-2026"],"source_locator":"Abstract (paragraph 1); Results/Genome-Wide Prophage Detection./Extracting Prophage Signal from Raw Model Predictions. (paragraph 3)"},{"title":"Underlying method and version boundaries","body":"Evo 2 uses the StripedHyena 2 architecture at single-nucleotide resolution. The official family documentation describes autoregressive training on 8.8 trillion OpenGenome2 tokens and context lengths up to one million bases; an exact evaluated checkpoint remains a separate identity.","source_ids":["evidence-reported-base-evo2-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies EVO2: Genome-wide prophage detection. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["lambda-prophage-2026"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-037"}],"facts":[{"label":"Model type","value":"Autoregressive DNA model; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-evo2-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"The benchmark separates embedding probes, fine-tuning, diagnostic tests and genome-wide scans; the associated result identifies which procedure is used.","status":"source_checked","source_ids":["lambda-prophage-2026"],"source_locator":"Abstract (paragraph 1); Results/Genome-Wide Prophage Detection./Extracting Prophage Signal from Raw Model Predictions. (paragraph 3)"},{"label":"Biological inputs","value":"Bacterial/phage DNA sequences and genome windows","status":"source_checked","source_ids":["lambda-prophage-2026"],"source_locator":"Materials and Methods/LAMBDA Benchmark Dataset Construction./Segment Subsampling. (paragraph 1); Materials and Methods/LAMBDA Benchmark Dataset Construction./Data Sources. (paragraph 1)"},{"label":"Outputs","value":"Phage classification scores or candidate prophage regions","status":"source_checked","source_ids":["lambda-prophage-2026"],"source_locator":"Results/Genome-Wide Prophage Detection./Scanning for Novel, Unannotated Prophage Regions. (paragraph 2); Materials and Methods/LAMBDA Benchmark Dataset Construction./Prophage Signal Extraction Algorithm. (paragraph 1)"},{"label":"Parameters","value":"7B parameters in the benchmark model inventory.","status":"source_checked","source_ids":["lambda-prophage-2026"],"source_locator":"Table1; EVO2 row"},{"label":"Known versions / configuration","value":"EVO2 7B as identified by Table 1; exact checkpoint hash is not supplied in that inventory.","status":"source_checked","source_ids":["lambda-prophage-2026"],"source_locator":"Results/Genome-Wide Prophage Detection./Probing EVO2 Latent Space with Sparse Autoencoders. (paragraph 1); Table T2 (paragraph 1)"},{"label":"Training data / fitting","value":"OpenGenome2 multispecies pretraining, as identified in Table 1; LAMBDA trains separate linear or three-layer neural probes on frozen embeddings.","status":"source_checked","source_ids":["lambda-prophage-2026"],"source_locator":"Results/Evaluating the Power of Pretrained Sequence Representations. (paragraph 2); Materials and Methods/LAMBDA Benchmark Dataset Construction./Comparison with Traditional and Protein-based Models. (paragraph 1)"},{"label":"Context limits","value":"The source lists a 1M-base model limit; the actual LAMBDA experiments evaluate 2-kb,4-kb and 8-kb windows.","status":"source_checked","source_ids":["lambda-prophage-2026"],"source_locator":"Table1 EVO2 row; Tables2–3 window lengths"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://github.com/ArcInstitute/evo2/blob/53f195997257c56c00e5ef8d33a54f5baad143a6/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-evo2-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"Apache 2.0 (upstream repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-base-evo2-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-base-evo2-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Extends genomic-model evaluation beyond eukaryotic regulatory-element labels.","source_ids":["lambda-prophage-2026"],"source_locator":"Abstract (paragraph 1); Background (paragraph 1)"}],"limitations":[{"text":"Candidate regions outside reference annotations are not automatically validated prophages; the paper uses additional gene-content screening.","source_ids":["lambda-prophage-2026"],"source_locator":"Materials and Methods/LAMBDA Benchmark Dataset Construction./Candidate Prophage Screening. (paragraph 1); Results/Genome-Wide Prophage Detection./Scanning for Novel, Unannotated Prophage Regions. (paragraph 2)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Bacterial/phage DNA sequences and genome windows","EVO2","Phage classification scores or candidate prophage regions"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["lambda-prophage-2026"],"source_locator":"Abstract (paragraph 1); Results/Genome-Wide Prophage Detection./Extracting Prophage Signal from Raw Model Predictions. (paragraph 3)"},"coverage":"limited","gaps":["The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-ab02228f50a37c","kind":"model","name":"C2S (GPT-2 Large)","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["cells-tissues"]},"source_ids":["cell2sentence-2024"],"links":[],"attributes":{"entity_level":"method","version":"GPT-2 Large","reported_name":"C2S (GPT-2 Large)","historical_missing_metadata":{"checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Cell2Sentence adapts GPT-2 to single-cell transcriptomics by writing cells as expression-ranked gene-name sequences.","summary_source_ids":["cell2sentence-2024"],"summary_source_locator":"Methods/Data transformation (paragraph 8); Introduction (paragraph 2)","sections":[{"title":"How the evaluated method works","body":"Genes are ordered by decreasing transcript abundance to create cell sentences. GPT-2 Large is fine-tuned on this representation for cell-type-conditioned generation and cell-type annotation.","source_ids":["cell2sentence-2024"],"source_locator":"Methods (paragraph 1); Experiments/Fine-Tuning Datasets (paragraph 3)"},{"title":"What was evaluated","body":"The linked evaluation record identifies C2S (GPT-2 Large): Combinatorial cell-label classification. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["cell2sentence-2024"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-027"}],"facts":[{"label":"Model type","value":"Study-specific predictive method; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["cell2sentence-2024"],"source_locator":"Methods (paragraph 1); Experiments/Fine-Tuning Datasets (paragraph 3)"},{"label":"Architecture / procedure","value":"Genes are ordered by decreasing transcript abundance to create cell sentences. GPT-2 Large is fine-tuned on this representation for cell-type-conditioned generation and cell-type annotation.","status":"source_checked","source_ids":["cell2sentence-2024"],"source_locator":"Methods (paragraph 1); Experiments/Fine-Tuning Datasets (paragraph 3)"},{"label":"Biological inputs","value":"Single-cell gene-expression profiles converted to ranked gene names","status":"source_checked","source_ids":["cell2sentence-2024"],"source_locator":"Methods/Data transformation (paragraph 8); Methods/Data transformation (paragraph 3)"},{"label":"Outputs","value":"Cell-type annotations or generated ranked gene sequences","status":"source_checked","source_ids":["cell2sentence-2024"],"source_locator":"Methods/Data transformation (paragraph 7); Inference Details (paragraph 2)"},{"label":"Parameters","value":"774,030,080 parameters for the model checkpoint used in the Cell2Sentence L1000 experiment; this is not a claim about every checkpoint in the family.","status":"source_checked","source_ids":["cell2sentence-2024"],"source_locator":"Table 5, Comparison of required compute on the L1000 dataset; # Parameters column and caption"},{"label":"Known versions / configuration","value":"GPT-2 Large","status":"source_checked","source_ids":["cell2sentence-2024"],"source_locator":"Experiments/Experiment 3: abstract summary generation/Objective: (paragraph 1); Experiments/Fine-Tuning Datasets (paragraph 2)"},{"label":"Training data / fitting","value":"Task-specific single-cell expression datasets described in the preprint; gene ranking replaces direct numeric expression input.","status":"source_checked","source_ids":["cell2sentence-2024"],"source_locator":"Methods/Data transformation (paragraph 8); Experimental Details/Evaluation Datasets (paragraph 1)"},{"label":"Context limits","value":"GPT-2 configurations use 1,024 tokens; the separately evaluated Pythia-160m configuration uses 9,200 tokens.","status":"source_checked","source_ids":["cell2sentence-2024"],"source_locator":"Training Details (paragraph 1); Experiments (paragraph 1)"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/vandijklab/cell2sentence/blob/a6efaf079f98491d4723ced44b929936b94368aa/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-cell2sentence-2024-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"Apache 2.0 (study repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-cell2sentence-2024-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-cell2sentence-2024-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Reuses text-language-model training tools for a biological representation that supports annotation and generation.","source_ids":["cell2sentence-2024"],"source_locator":"Experiments/Experiment 3: abstract summary generation/Objective: (paragraph 1); Abstract (paragraph 1)"}],"limitations":[{"text":"Rank ordering does not directly retain measured expression magnitudes; reconstruction and generated-cell plausibility require the paper’s separate validation.","source_ids":["cell2sentence-2024"],"source_locator":"Methods/Data transformation (paragraph 7); Inference Details (paragraph 2)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Single-cell gene-expression profiles converted to ranked gene names","C2S (GPT-2 Large)","Cell-type annotations or generated ranked gene sequences"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["cell2sentence-2024"],"source_locator":"Methods (paragraph 1); Experiments/Fine-Tuning Datasets (paragraph 3)"},"coverage":"limited","gaps":["The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-abc19288fe9009","kind":"model","name":"Caduceus","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["dna-genomes"]},"source_ids":["quadruplex-llm-benchmark-2025"],"links":[],"attributes":{"entity_level":"method","version":"8M","reported_name":"Caduceus","historical_missing_metadata":{"checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"This genomic model is adapted to predict G-quadruplex-associated sequence regions.","summary_source_ids":["quadruplex-llm-benchmark-2025"],"summary_source_locator":"Discussion and conclusions (paragraph 6); Results/LLM performance at the genome-wide level (paragraph 4)","sections":[{"title":"How the evaluated method works","body":"The study fine-tunes genomic sequence models for G4 classification and applies them to whole-genome annotation, retaining each model’s native tokenisation.","source_ids":["quadruplex-llm-benchmark-2025"],"source_locator":"Materials and methods/Data preparation (paragraph 2); Results/Interpretation of LLMs (paragraph 2)"},{"title":"Underlying method and version boundaries","body":"Caduceus exposes distinct Ph and PS configurations. The documented Ph-131k checkpoint uses 16 layers, width 256 and reverse-complement data augmentation; PS implements reverse-complement equivariance. These training choices are not interchangeable.","source_ids":["evidence-reported-base-caduceus-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies Caduceus: G-quadruplex classification. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["quadruplex-llm-benchmark-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-006"}],"facts":[{"label":"Model type","value":"DNA state-space model; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-caduceus-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"The study fine-tunes genomic sequence models for G4 classification and applies them to whole-genome annotation, retaining each model’s native tokenisation.","status":"source_checked","source_ids":["quadruplex-llm-benchmark-2025"],"source_locator":"Materials and methods/Data preparation (paragraph 2); Results/Interpretation of LLMs (paragraph 2)"},{"label":"Biological inputs","value":"DNA sequences for G-quadruplex classification and genomic scanning","status":"source_checked","source_ids":["quadruplex-llm-benchmark-2025"],"source_locator":"Results/LLM comparison on quadruplex datasets (paragraph 1); Materials and methods/Data preparation (paragraph 1)"},{"label":"Outputs","value":"G4-associated sequence predictions","status":"source_checked","source_ids":["quadruplex-llm-benchmark-2025"],"source_locator":"Introduction (paragraph 3); CRediT authorship contribution statement (paragraph 1)"},{"label":"Parameters","value":"8 million parameters, as identified for this row","status":"source_checked","source_ids":["quadruplex-llm-benchmark-2025"],"source_locator":"Materials and methods/Low Rank Adaptation (LoRA) (paragraph 1); Discussion and conclusions (paragraph 3)"},{"label":"Known versions / configuration","value":"8M","status":"source_checked","source_ids":["quadruplex-llm-benchmark-2025"],"source_locator":"Table tbl0030 (paragraph 1); Table tbl0025 (paragraph 1)"},{"label":"Training data / fitting","value":"Task-specific fine-tuning for approximately four to ten epochs, depending on model performance, with gradient accumulation.","status":"source_checked","source_ids":["quadruplex-llm-benchmark-2025"],"source_locator":"Results/LLM comparison on quadruplex datasets (paragraph 5); Materials and methods/Fine-tuning (paragraph 1)"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["quadruplex-llm-benchmark-2025","evidence-reported-base-caduceus-readme-md"],"source_locator":"Materials and methods/Data preparation; Materials and methods/Tokenization for G4s; Materials and methods/Metrics of evaluation; Materials and methods/Fine-tuning; Materials and methods/Low Rank Adaptation (LoRA); inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://github.com/kuleshov-group/caduceus/blob/0060a6d8079b6a040fc55d505e15972a327b70a6/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-caduceus-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"Apache 2.0 (upstream repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-base-caduceus-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-base-caduceus-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Compares transformer, long-convolution and state-space architectures on a specialised non-B-DNA task.","source_ids":["quadruplex-llm-benchmark-2025"],"source_locator":"Introduction (paragraph 1); Abstract (paragraph 1)"}],"limitations":[{"text":"Predicted G4 sequence potential is not a direct measurement of structure formation in a particular cellular context.","source_ids":["quadruplex-llm-benchmark-2025"],"source_locator":"Introduction (paragraph 6); Discussion and conclusions (paragraph 1)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["DNA sequences for G-quadruplex classification and genomic scanning","Caduceus","G4-associated sequence predictions"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["quadruplex-llm-benchmark-2025"],"source_locator":"Materials and methods/Data preparation (paragraph 2); Results/Interpretation of LLMs (paragraph 2)"},"coverage":"limited","gaps":["A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-ade36035f58f27","kind":"model","name":"DNABERT-2","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["dna-genomes"]},"source_ids":["quadruplex-llm-benchmark-2025"],"links":[],"attributes":{"entity_level":"method","version":"117M","reported_name":"DNABERT-2","historical_missing_metadata":{"checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"This genomic model is adapted to predict G-quadruplex-associated sequence regions.","summary_source_ids":["quadruplex-llm-benchmark-2025"],"summary_source_locator":"Discussion and conclusions (paragraph 6); Results/LLM performance at the genome-wide level (paragraph 4)","sections":[{"title":"How the evaluated method works","body":"The study fine-tunes genomic sequence models for G4 classification and applies them to whole-genome annotation, retaining each model’s native tokenisation.","source_ids":["quadruplex-llm-benchmark-2025"],"source_locator":"Materials and methods/Data preparation (paragraph 2); Results/Interpretation of LLMs (paragraph 2)"},{"title":"Underlying method and version boundaries","body":"DNABERT-2 replaces overlapping k-mer tokens with byte-pair encoding and uses ALiBi positional biases. The official 117M model produces 768-dimensional token representations; downstream classifiers and pooling choices are separate configuration details.","source_ids":["evidence-reported-base-dnabert2-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies DNABERT-2: G-quadruplex classification. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["quadruplex-llm-benchmark-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-005"}],"facts":[{"label":"Model type","value":"DNA sequence transformer; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-dnabert2-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"The study fine-tunes genomic sequence models for G4 classification and applies them to whole-genome annotation, retaining each model’s native tokenisation.","status":"source_checked","source_ids":["quadruplex-llm-benchmark-2025"],"source_locator":"Materials and methods/Data preparation (paragraph 2); Results/Interpretation of LLMs (paragraph 2)"},{"label":"Biological inputs","value":"DNA sequences for G-quadruplex classification and genomic scanning","status":"source_checked","source_ids":["quadruplex-llm-benchmark-2025"],"source_locator":"Results/LLM comparison on quadruplex datasets (paragraph 1); Materials and methods/Data preparation (paragraph 1)"},{"label":"Outputs","value":"G4-associated sequence predictions","status":"source_checked","source_ids":["quadruplex-llm-benchmark-2025"],"source_locator":"Introduction (paragraph 3); CRediT authorship contribution statement (paragraph 1)"},{"label":"Parameters","value":"117 million parameters, as identified for this row","status":"source_checked","source_ids":["quadruplex-llm-benchmark-2025"],"source_locator":"Materials and methods/Low Rank Adaptation (LoRA) (paragraph 1); Discussion and conclusions (paragraph 3)"},{"label":"Known versions / configuration","value":"117M","status":"source_checked","source_ids":["quadruplex-llm-benchmark-2025"],"source_locator":"Table tbl0030 (paragraph 1); Table tbl0025 (paragraph 1)"},{"label":"Training data / fitting","value":"Task-specific fine-tuning for approximately four to ten epochs, depending on model performance, with gradient accumulation.","status":"source_checked","source_ids":["quadruplex-llm-benchmark-2025"],"source_locator":"Results/LLM comparison on quadruplex datasets (paragraph 5); Materials and methods/Fine-tuning (paragraph 1)"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["quadruplex-llm-benchmark-2025","evidence-reported-base-dnabert2-readme-md"],"source_locator":"Materials and methods/Data preparation; Materials and methods/Tokenization for G4s; Materials and methods/Metrics of evaluation; Materials and methods/Fine-tuning; Materials and methods/Low Rank Adaptation (LoRA); inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://github.com/MAGICS-LAB/DNABERT_2/blob/f25bed9ee20db966dff39e5c1571249d04e36404/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-dnabert2-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"Apache 2.0 (upstream repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-base-dnabert2-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-base-dnabert2-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Compares transformer, long-convolution and state-space architectures on a specialised non-B-DNA task.","source_ids":["quadruplex-llm-benchmark-2025"],"source_locator":"Introduction (paragraph 1); Abstract (paragraph 1)"}],"limitations":[{"text":"Predicted G4 sequence potential is not a direct measurement of structure formation in a particular cellular context.","source_ids":["quadruplex-llm-benchmark-2025"],"source_locator":"Introduction (paragraph 6); Discussion and conclusions (paragraph 1)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["DNA sequences for G-quadruplex classification and genomic scanning","DNABERT-2","G4-associated sequence predictions"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["quadruplex-llm-benchmark-2025"],"source_locator":"Materials and methods/Data preparation (paragraph 2); Results/Interpretation of LLMs (paragraph 2)"},"coverage":"limited","gaps":["A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-b3fdf259d51533","kind":"model","name":"Chai-1","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["molecular-interactions"]},"source_ids":["antibody-flexibility-2025"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"Chai-1","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"This structure-prediction configuration is evaluated for antibody–antigen complex and interface quality.","summary_source_ids":["antibody-flexibility-2025"],"summary_source_locator":"Results/Fold quality affects predictive performance/Folding quality declines across models without MSAs or templates. (paragraph 3); Results/Fold quality affects predictive performance/AF3 and Chai-1 (no-MSA/template) effectively preserve prediction quality. (paragraph 3)","sections":[{"title":"How the evaluated method works","body":"The generative co-folding predictor builds antibody–antigen structures that are then assessed for interface quality and downstream paratope/epitope prediction.","source_ids":["antibody-flexibility-2025"],"source_locator":"Results/Fold quality affects predictive performance/Folding quality declines across models without MSAs or templates. (paragraph 3); Results/Fold quality affects predictive performance/Folding quality declines across models without MSAs or templates. (paragraph 2)"},{"title":"Underlying method and version boundaries","body":"Chai-1 predicts biomolecular structures through the released inference implementation. Its input options, MSA/template use and sampling budget are part of each evaluated configuration.","source_ids":["evidence-reported-base-chai-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies Chai-1: Antibody–antigen interaction prediction using folded complexes. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["antibody-flexibility-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b3-037"}],"facts":[{"label":"Model type","value":"Biomolecular structure predictor; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-chai-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"The generative co-folding predictor builds antibody–antigen structures that are then assessed for interface quality and downstream paratope/epitope prediction.","status":"source_checked","source_ids":["antibody-flexibility-2025"],"source_locator":"Results/Fold quality affects predictive performance/Folding quality declines across models without MSAs or templates. (paragraph 3); Results/Fold quality affects predictive performance/Folding quality declines across models without MSAs or templates. (paragraph 2)"},{"label":"Biological inputs","value":"Antibody and antigen sequences; MSA/template availability is explicitly configuration-dependent.","status":"source_checked","source_ids":["antibody-flexibility-2025"],"source_locator":"Results/Flexibility improves antibody-antigen interaction prediction/Binary flexibility patterns emerge despite linear flexibility modeling. (paragraph 3); Results/Flexibility improves antibody-antigen interaction prediction (paragraph 1)"},{"label":"Outputs","value":"Predicted antibody–antigen complex coordinates","status":"source_checked","source_ids":["antibody-flexibility-2025"],"source_locator":"Results/Fold quality affects predictive performance/Folding quality declines across models without MSAs or templates. (paragraph 1); Results/Fold quality affects predictive performance/AF3 and Chai-1 (no-MSA/template) effectively preserve prediction quality. (paragraph 3)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["antibody-flexibility-2025","evidence-reported-base-chai-readme-md"],"source_locator":"Materials and methods; Materials and methods/Dataset; Materials and methods/Method; Materials and methods/Method/Data representation.; Materials and methods/Method/Flexibility score with ESMFold.; Materials and methods/Method/Model architecture.; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"Chai-1 without MSA/template information in the identified comparison","status":"source_checked","source_ids":["antibody-flexibility-2025"],"source_locator":"Table pcbi.1013576.t005 (paragraph 1); Results/Fold quality affects predictive performance/Folding quality declines across models without MSAs or templates. (paragraph 1)"},{"label":"Training data / fitting","value":"The released source implementation supplies prefolded antibody–antigen complexes for the GEP test set, with neither MSAs nor templates. No retraining of the structure predictor is described; downstream interface prediction is a separate component.","status":"source_checked","source_ids":["antibody-flexibility-2025"],"source_locator":"Results / Fold quality affects predictive performance"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["antibody-flexibility-2025","evidence-reported-base-chai-readme-md"],"source_locator":"Materials and methods; Materials and methods/Dataset; Materials and methods/Method; Materials and methods/Method/Data representation.; Materials and methods/Method/Flexibility score with ESMFold.; Materials and methods/Method/Model architecture.; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://github.com/chaidiscovery/chai-lab/blob/66c38d1fe5c6756a89ff8596b1dea87d305ec06f/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-chai-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"Apache 2.0 (upstream repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-base-chai-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"Apache 2.0 for both Chai-1 code and model weights, explicitly stated in the official README; paper-specific derived artifacts are separate.","status":"source_checked","source_ids":["evidence-reported-base-chai-readme-md"],"source_locator":"README.md; Licence section"}],"strengths":[{"text":"Examines how predicted structures affect downstream interface-site analysis.","source_ids":["antibody-flexibility-2025"],"source_locator":"Results/Fold quality affects predictive performance/Folding quality declines across models without MSAs or templates. (paragraph 3); Results/Fold quality affects predictive performance/AF3 and Chai-1 (no-MSA/template) effectively preserve prediction quality. (paragraph 4)"}],"limitations":[{"text":"The paper reports configuration-dependent differences and limited antibody–antigen accuracy; general complex-folding scores cannot substitute for this task.","source_ids":["antibody-flexibility-2025"],"source_locator":"Results/Flexibility improves antibody-antigen interaction prediction/Binary flexibility patterns emerge despite linear flexibility modeling. (paragraph 3); Results/Fold quality affects predictive performance (paragraph 1)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Antibody and antigen sequences","Chai-1","Predicted antibody–antigen complex coordinates"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["antibody-flexibility-2025"],"source_locator":"Results/Fold quality affects predictive performance/Folding quality declines across models without MSAs or templates. (paragraph 3); Results/Fold quality affects predictive performance/Folding quality declines across models without MSAs or templates. (paragraph 2)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-b46ae14b9927ac","kind":"model","name":"Geneformer","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["cells-tissues"]},"source_ids":["single-cell-peft-2024"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"Geneformer","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"This single-cell model is evaluated in a study of parameter-efficient adaptation for cell-type identification.","summary_source_ids":["single-cell-peft-2024"],"summary_source_locator":"Introduction (paragraph 5); Results/Comparison of proposed PEFT strategies and other finetuning approaches (paragraph 2)","sections":[{"title":"How the evaluated method works","body":"The study compares ordinary fine-tuning with methods that retain original model parameters while learning additional tensors.","source_ids":["single-cell-peft-2024"],"source_locator":"Introduction (paragraph 4); Proposed PEFT strategies for scLLMs/Finetuning and evaluation settings (paragraph 1)"},{"title":"Underlying method and version boundaries","body":"Geneformer ranks genes by expression scaled against its pretraining corpus, then uses a transformer encoder with a masked-gene objective. V1 and V2 have different corpora, vocabularies, sizes and context limits, so a historical paper name is not replaced with today’s default checkpoint.","source_ids":["evidence-reported-base-geneformer-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies Geneformer: Cell-type identification. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["single-cell-peft-2024"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-026"}],"facts":[{"label":"Model type","value":"Single-cell transformer; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-geneformer-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"The study compares ordinary fine-tuning with methods that retain original model parameters while learning additional tensors.","status":"source_checked","source_ids":["single-cell-peft-2024"],"source_locator":"Introduction (paragraph 4); Proposed PEFT strategies for scLLMs/Finetuning and evaluation settings (paragraph 1)"},{"label":"Biological inputs","value":"Single-cell gene-expression profiles","status":"source_checked","source_ids":["single-cell-peft-2024"],"source_locator":"An overview of current scLLMs (paragraph 1); Data and code availability (paragraph 1)"},{"label":"Outputs","value":"Cell-type predictions","status":"source_checked","source_ids":["single-cell-peft-2024"],"source_locator":"Proposed PEFT strategies for scLLMs/Finetuning and evaluation settings (paragraph 2); Results/Comparison of native scLLMs on cell type identification (paragraph 4)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["single-cell-peft-2024","evidence-reported-base-geneformer-readme-md"],"source_locator":"An overview of current scLLMs/Pretrainer:; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"Geneformer is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["single-cell-peft-2024"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"Genecorpus-30M pretraining (29.9 million transcriptomes) followed by the paper’s task-specific adaptation comparisons on MS, Zheng68k, NSCLC and COVID-19 datasets.","status":"source_checked","source_ids":["single-cell-peft-2024"],"source_locator":"An overview of current scLLMs / Pretrainer; Proposed PEFT strategies / Data preparation"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["single-cell-peft-2024","evidence-reported-base-geneformer-readme-md"],"source_locator":"An overview of current scLLMs/Pretrainer:; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://huggingface.co/ctheodoris/Geneformer/blob/1f7fbae4e469a5f4f1af8c111a529cfe1b3829f5/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-geneformer-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","status":"unreported","source_ids":["evidence-reported-base-geneformer-readme-md"],"source_locator":"README.md and repository-root licence-file search"},{"label":"Weights licence","value":"Apache 2.0 is declared in the official Geneformer model-card metadata; the exact historical configuration still needs its checkpoint identity.","status":"source_checked","source_ids":["evidence-reported-base-geneformer-readme-md"],"source_locator":"README.md front matter, license field; model-version list"}],"strengths":[{"text":"Examines the trade-off between adaptation performance and the number of trainable parameters.","source_ids":["single-cell-peft-2024"],"source_locator":"Results/Comparison of proposed PEFT strategies and other finetuning approaches (paragraph 3); Results/Comparison of proposed PEFT strategies and other finetuning approaches (paragraph 2)"}],"limitations":[{"text":"The base-model label alone does not identify which adaptation tensors were trained; exact PEFT and full-fine-tuning rows must remain separate.","source_ids":["single-cell-peft-2024"],"source_locator":"Results/Comparison of proposed PEFT strategies and other finetuning approaches (paragraph 2); Introduction (paragraph 5)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Single-cell gene-expression profiles","Geneformer","Cell-type predictions"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["single-cell-peft-2024"],"source_locator":"Introduction (paragraph 4); Proposed PEFT strategies for scLLMs/Finetuning and evaluation settings (paragraph 1)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","Geneformer is the comparison-table label; that label does not specify an immutable weight revision.","A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-bdb1db16d3389d","kind":"model","name":"ENBED (GRCh38)","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["dna-genomes"]},"source_ids":["enbed-2024"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"ENBED (GRCh38)","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"ENBED is a byte-level encoder–decoder transformer for genomic sequence representation and sequence-to-sequence tasks.","summary_source_ids":["enbed-2024"],"summary_source_locator":"2 Methods/2.5 Application domains (paragraph 1); 5 Discussion (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"Byte-level nucleotide inputs feed encoder and decoder transformer blocks with a subquadratic attention implementation. Masked-language pretraining precedes task-specific adaptation.","source_ids":["enbed-2024"],"source_locator":"1 Introduction/1.2 Our contributions (paragraph 1); 4 Ablation studies/4.1 Encoder–decoder architecture (paragraph 2)"},{"title":"What was evaluated","body":"The linked evaluation record identifies ENBED (GRCh38): Enhancer classification. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["enbed-2024"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-004"}],"facts":[{"label":"Model type","value":"Transformer representation pipeline; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["enbed-2024"],"source_locator":"1 Introduction/1.2 Our contributions (paragraph 1); 4 Ablation studies/4.1 Encoder–decoder architecture (paragraph 2)"},{"label":"Architecture / procedure","value":"Byte-level nucleotide inputs feed encoder and decoder transformer blocks with a subquadratic attention implementation. Masked-language pretraining precedes task-specific adaptation.","status":"source_checked","source_ids":["enbed-2024"],"source_locator":"1 Introduction/1.2 Our contributions (paragraph 1); 4 Ablation studies/4.1 Encoder–decoder architecture (paragraph 2)"},{"label":"Biological inputs","value":"DNA sequences at single-byte nucleotide resolution","status":"source_checked","source_ids":["enbed-2024"],"source_locator":"1 Introduction/1.2 Our contributions (paragraph 2); 5 Discussion (paragraph 1)"},{"label":"Outputs","value":"Task-specific classifications or generated DNA sequences","status":"source_checked","source_ids":["enbed-2024"],"source_locator":"2 Methods/2.4 Applications of foundation models using transfer learning/2.4.2 Fine-tuning for downstream tasks (paragraph 1); 2 Methods/2.5 Application domains/2.5.1 Genomic benchmarks (paragraph 1)"},{"label":"Parameters","value":"1.2 billion trainable parameters in the full encoder–decoder model.","status":"source_checked","source_ids":["enbed-2024"],"source_locator":"4 Ablation studies (paragraph 1); 4 Ablation studies/4.1 Encoder–decoder architecture (paragraph 1)"},{"label":"Known versions / configuration","value":"ENBED (GRCh38) is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["enbed-2024"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"Reference-genome sequences; the GRCh38-labelled row is a distinct human-reference configuration.","status":"source_checked","source_ids":["enbed-2024"],"source_locator":"1 Introduction/1.2 Our contributions/1.2.1 Evaluation of performance on genomic benchmark datasets (paragraph 1); 3 Results/3.1 ENBED outperforms state-of-the-art models on GB datasets (paragraph 2)"},{"label":"Context limits","value":"16,384 input/output tokens using local sliding-window plus global attention. The 512-token value in Methods describes the dense-attention hardware baseline, not ENBED’s final context.","status":"source_checked","source_ids":["enbed-2024"],"source_locator":"2 Methods/2.3 Attention (paragraph 2); 2 Methods/2.3 Attention/2.3.1 Sliding-window attention (paragraph 1)"},{"label":"Access","value":"The authors provide implementation code at https://github.itap.purdue.edu/Clan-labs/ENBED and model weights through https://huggingface.co/malusare. A table-specific checkpoint hash is not supplied by these account-level links.","status":"source_checked","source_ids":["enbed-2024"],"source_locator":"Data availability"},{"label":"Code licence","value":"No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","status":"unreported","source_ids":["enbed-2024"],"source_locator":"Abstract (paragraph 1); 2 Methods/2.4 Applications of foundation models using transfer learning/2.4.1 Building the foundation model (paragraph 1)"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["enbed-2024"],"source_locator":"Abstract (paragraph 1); 2 Methods/2.1 Encoder–decoder model architecture (paragraph 1)"}],"strengths":[{"text":"Single-base input resolution retains substitutions and indels that can be obscured by multi-base token boundaries.","source_ids":["enbed-2024"],"source_locator":"1 Introduction/1.1 Limitations of previous work/1.1.2 Tokenization (paragraph 1); 3 Results/3.4 ENBED generates mutations using sequence-to-sequence transformation (paragraph 1)"}],"limitations":[{"text":"The evaluated downstream head and training corpus must be kept with each result; the GRCh38 configuration is not interchangeable with every ENBED model.","source_ids":["enbed-2024"],"source_locator":"2 Methods/2.4 Applications of foundation models using transfer learning/2.4.2 Fine-tuning for downstream tasks (paragraph 1); 3 Results/3.3 ENBED identifies biological function annotations (paragraph 1)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["DNA sequences at single-byte nucleotide resolution","ENBED (GRCh38)","Task-specific classifications or generated DNA sequences"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["enbed-2024"],"source_locator":"1 Introduction/1.2 Our contributions (paragraph 1); 4 Ablation studies/4.1 Encoder–decoder architecture (paragraph 2)"},"coverage":"limited","gaps":["ENBED (GRCh38) is the comparison-table label; that label does not specify an immutable weight revision.","No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-c464bface507ee","kind":"model","name":"BPfold","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["rna-transcriptomes"]},"source_ids":["bpfold-2025"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"BPfold","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"BPfold predicts RNA secondary structure by combining learned sequence features with base-pair motif energies.","summary_source_ids":["bpfold-2025"],"summary_source_locator":"Discussion (paragraph 5); Abstract (paragraph 2)","sections":[{"title":"How the evaluated method works","body":"Modified transformer blocks combine sequence attention with a hybrid convolutional branch operating on motif-energy maps. Squeeze-and-excitation recalibrates channels, while pair attention integrates thermodynamic priors into base-pair prediction.","source_ids":["bpfold-2025"],"source_locator":"Methods/Deep neural network with base pair attention (paragraph 1); Methods/Deep neural network with base pair attention (paragraph 4)"},{"title":"What was evaluated","body":"The linked evaluation record identifies BPfold: RNA secondary structure. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["bpfold-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-011"}],"facts":[{"label":"Model type","value":"Hybrid convolutional/transformer pipeline; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["bpfold-2025"],"source_locator":"Methods/Deep neural network with base pair attention (paragraph 1); Methods/Deep neural network with base pair attention (paragraph 4)"},{"label":"Architecture / procedure","value":"Modified transformer blocks combine sequence attention with a hybrid convolutional branch operating on motif-energy maps. Squeeze-and-excitation recalibrates channels, while pair attention integrates thermodynamic priors into base-pair prediction.","status":"source_checked","source_ids":["bpfold-2025"],"source_locator":"Methods/Deep neural network with base pair attention (paragraph 1); Methods/Deep neural network with base pair attention (paragraph 4)"},{"label":"Biological inputs","value":"RNA sequences and a precomputed base-pair motif-energy library","status":"source_checked","source_ids":["bpfold-2025"],"source_locator":"Methods/Base pair motif energy as thermodynamic prior (paragraph 4); Abstract (paragraph 2)"},{"label":"Outputs","value":"RNA secondary-structure base-pair predictions","status":"source_checked","source_ids":["bpfold-2025"],"source_locator":"Discussion (paragraph 5); Abstract (paragraph 2)"},{"label":"Parameters","value":"7,962,416 parameters in the published BPfold configuration.","status":"source_checked","source_ids":["evidence-reported-bpfold-supplement"],"source_locator":"Supplementary Table 6, BPfold row"},{"label":"Known versions / configuration","value":"BPfold is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["bpfold-2025"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"Sequence-wise and family-wise RNA structure datasets described in the paper; the motif library enumerates local base-pair neighbourhoods.","status":"source_checked","source_ids":["bpfold-2025"],"source_locator":"Results/Assessing the effectiveness of base pair motif energy (paragraph 1); Results/Evaluating BPfold on sequence-wise datasets (paragraph 3)"},{"label":"Context limits","value":"BPfold is trained on RNAs of at most 600 nucleotides from RNAStrAlign and bpRNA; family-wise tests use separate curated sets.","status":"source_checked","source_ids":["bpfold-2025"],"source_locator":"Table Tab2 (paragraph 1); Table Tab1 (paragraph 1)"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/heqin-zhu/BPfold/blob/d37d6aa10cbca13e590ff83917fc4d63fec2ddbc/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-bpfold-2025-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"MIT (study repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-bpfold-2025-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-bpfold-2025-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Family-wise evaluation explicitly examines transfer to RNA families not represented in training.","source_ids":["bpfold-2025"],"source_locator":"Results/Evaluating BPfold on family-wise datasets (paragraph 1); Results/Evaluating BPfold on sequence-wise datasets (paragraph 1)"}],"limitations":[{"text":"The thermodynamic motif library is a prior, not a guarantee of generalisation to every RNA family or in-vivo folding condition.","source_ids":["bpfold-2025"],"source_locator":"Abstract (paragraph 2); Discussion (paragraph 1)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["RNA sequences and a precomputed base-pair motif-energy library","BPfold","RNA secondary-structure base-pair predictions"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["bpfold-2025"],"source_locator":"Methods/Deep neural network with base pair attention (paragraph 1); Methods/Deep neural network with base pair attention (paragraph 4)"},"coverage":"limited","gaps":["BPfold is the comparison-table label; that label does not specify an immutable weight revision.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-cbbe04b826ceff","kind":"model","name":"PhyloGPN","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["dna-genomes"]},"source_ids":["phylogpn-2025"],"links":[],"attributes":{"entity_level":"method","version":"not stated in table","reported_name":"PhyloGPN","historical_missing_metadata":{"checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"PhyloGPN learns genomic representations with a phylogeny-aware objective and predicts from single sequences.","summary_source_ids":["phylogpn-2025"],"summary_source_locator":"Abstract (paragraph 1); Discussion and Conclusion (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"Multispecies whole-genome alignments and phylogenetic trees enter the training loss to model nucleotide evolution. They are not required as inputs for downstream single-sequence prediction.","source_ids":["phylogpn-2025"],"source_locator":"Abstract (paragraph 1); Discussion and Conclusion (paragraph 1)"},{"title":"What was evaluated","body":"The linked evaluation record identifies PhyloGPN: ClinVar 3-prime UTR variant classification. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["phylogpn-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-b2-phylogpn-2025"}],"facts":[{"label":"Model type","value":"Study-specific predictive method; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["phylogpn-2025"],"source_locator":"Abstract (paragraph 1); Discussion and Conclusion (paragraph 1)"},{"label":"Architecture / procedure","value":"Multispecies whole-genome alignments and phylogenetic trees enter the training loss to model nucleotide evolution. They are not required as inputs for downstream single-sequence prediction.","status":"source_checked","source_ids":["phylogpn-2025"],"source_locator":"Abstract (paragraph 1); Discussion and Conclusion (paragraph 1)"},{"label":"Biological inputs","value":"DNA sequence at inference; multispecies alignment and tree information during training","status":"source_checked","source_ids":["phylogpn-2025"],"source_locator":"Abstract (paragraph 1); Background and Related Work/Molecular Phylogenetics (paragraph 1)"},{"label":"Outputs","value":"Sequence representations and variant-effect scores","status":"source_checked","source_ids":["phylogpn-2025"],"source_locator":"Results/Embedding Evaluation (paragraph 1); Introduction (paragraph 3)"},{"label":"Parameters","value":"About 83M stored parameters; reverse-complement weight tying means the number of free parameters is roughly half the weights and biases, plus layer-normalisation parameters.","status":"source_checked","source_ids":["phylogpn-2025"],"source_locator":"Methods/Architecture and Training (paragraph 1); Methods/Loss Function (paragraph 1)"},{"label":"Known versions / configuration","value":"PhyloGPN is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["phylogpn-2025"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"Multispecies whole-genome alignment with explicit phylogenetic modelling.","status":"source_checked","source_ids":["phylogpn-2025"],"source_locator":"Methods/Training Data (paragraph 2); Methods/Training Data (paragraph 1)"},{"label":"Context limits","value":"A 481-bp receptive field from 40 dilated residual blocks; training consumes approximately 10-kb alignment blocks, a different quantity.","status":"source_checked","source_ids":["phylogpn-2025"],"source_locator":"Methods/Architecture and Training (paragraph 1); Methods/Architecture and Training (paragraph 2)"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/songlab-cal/gpn/blob/6f28c81bcbfe7d65cb6d8ece9ce88f87ca583791/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-phylogpn-2025-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"MIT (study repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-phylogpn-2025-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-phylogpn-2025-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Uses evolutionary information during training while retaining sequence-only inference.","source_ids":["phylogpn-2025"],"source_locator":"Background and Related Work/Transfer Learning and Genomic Language Models (gLMs) (paragraph 2); Results/Embedding Evaluation (paragraph 4)"}],"limitations":[{"text":"Sequence-only inference does not imply sequence-only training; alignment quality and phylogenetic assumptions remain relevant.","source_ids":["phylogpn-2025"],"source_locator":"Abstract (paragraph 1); Background and Related Work/Molecular Phylogenetics (paragraph 1)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["DNA sequence at inference","PhyloGPN","Sequence representations and variant-effect scores"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["phylogpn-2025"],"source_locator":"Abstract (paragraph 1); Discussion and Conclusion (paragraph 1)"},"coverage":"limited","gaps":["PhyloGPN is the comparison-table label; that label does not specify an immutable weight revision.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-ccd1160ad4ec27","kind":"model","name":"ESM2","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["proteins-complexes"]},"source_ids":["fujisan-2024"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"ESM2","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"This FUJISAN-study baseline compares proteins by cosine similarity between mean ESM-2 embeddings.","summary_source_ids":["fujisan-2024"],"summary_source_locator":"Materials and methods/Prediction with ESM-2 (paragraph 2); Results and discussion/Limitations and perspectives (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"esm2_t33_650M_UR50D emits 1,280-dimensional residue vectors. Mean pooling gives a protein vector; cosine similarity estimates whether two proteins share an enzymatic function.","source_ids":["fujisan-2024"],"source_locator":"Materials and methods/Prediction with ESM-2 (paragraph 2); Materials and methods/Prediction with DeepFRI (paragraph 1)"},{"title":"Underlying method and version boundaries","body":"ESM-2 is a transformer protein language-model family. The official repository exposes residue embeddings, sequence-level pooling and models at several sizes; the study configuration determines which of these is evaluated.","source_ids":["evidence-reported-base-esm-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies ESM2: Enzyme functional identity prediction. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["fujisan-2024"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-020"}],"facts":[{"label":"Model type","value":"Protein sequence transformer; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-esm-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"esm2_t33_650M_UR50D emits 1,280-dimensional residue vectors. Mean pooling gives a protein vector; cosine similarity estimates whether two proteins share an enzymatic function.","status":"source_checked","source_ids":["fujisan-2024"],"source_locator":"Materials and methods/Prediction with ESM-2 (paragraph 2); Materials and methods/Prediction with DeepFRI (paragraph 1)"},{"label":"Biological inputs","value":"Pairs of protein amino-acid sequences","status":"source_checked","source_ids":["fujisan-2024"],"source_locator":"Materials and methods/Prediction with ESM-2 (paragraph 1); Introduction (paragraph 4)"},{"label":"Outputs","value":"Pairwise cosine similarity used as a same-function score","status":"source_checked","source_ids":["fujisan-2024"],"source_locator":"Materials and methods/Prediction with ESM-2 (paragraph 2); Materials and methods/Prediction with DeepFRI (paragraph 1)"},{"label":"Parameters","value":"650 million backbone parameters; no FUJISAN LightGBM features are part of this comparator.","status":"source_checked","source_ids":["fujisan-2024"],"source_locator":"Results and discussion/Analysis of feature importance (paragraph 1); Materials and methods/Model training and hyperparameter optimization (paragraph 1)"},{"label":"Known versions / configuration","value":"esm2_t33_650M_UR50D","status":"unreported","source_ids":["fujisan-2024"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"The comparator uses pretrained ESM-2 sequence embeddings and cosine similarity; it does not fit FUJISAN’s LightGBM head or a task-specific interaction classifier.","status":"source_checked","source_ids":["fujisan-2024"],"source_locator":"ESM2 embedding comparator and cosine-similarity procedure"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["fujisan-2024","evidence-reported-base-esm-readme-md"],"source_locator":"Materials and methods/Dataset construction; Materials and methods/Feature engineering/Full-length sequence similarity features; Materials and methods/Feature engineering/Domain structural similarity features; Materials and methods/Feature engineering/Pocket similarity features; Materials and methods/Model training and hyperparameter optimization; Materials and methods/Performance assessment; Materials and methods/Prediction with DeepFRI; Materials and methods/Prediction with ESM-2; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://github.com/facebookresearch/esm/blob/2b369911bb5b4b0dda914521b9475cad1656b2ac/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-esm-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"MIT (upstream repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-base-esm-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-base-esm-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"A simple sequence-only representation baseline with an explicit pooling and scoring rule.","source_ids":["fujisan-2024"],"source_locator":"Materials and methods/Prediction with ESM-2 (paragraph 1); Materials and methods/Feature engineering/Pocket similarity features (paragraph 1)"}],"limitations":[{"text":"Cosine similarity is a proxy for functional similarity, not a calibrated catalytic-function probability.","source_ids":["fujisan-2024"],"source_locator":"Materials and methods/Prediction with ESM-2 (paragraph 2); Conclusions (paragraph 1)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Pairs of protein amino-acid sequences","ESM2","Pairwise cosine similarity used as a same-function score"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["fujisan-2024"],"source_locator":"Materials and methods/Prediction with ESM-2 (paragraph 2); Materials and methods/Prediction with DeepFRI (paragraph 1)"},"coverage":"limited","gaps":["esm2_t33_650M_UR50D","A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-cd246741c378db","kind":"model","name":"CodonBERT","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["rna-transcriptomes"]},"source_ids":["codonbert-vaccines-2024"],"links":[],"attributes":{"entity_level":"method","version":"not stated in table","reported_name":"CodonBERT","historical_missing_metadata":{"checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"CodonBERT learns mRNA representations using codon tokens and adapts them to mRNA-property prediction.","summary_source_ids":["codonbert-vaccines-2024"],"summary_source_locator":"Results (paragraph 1); Methods/Model architecture (paragraph 2)","sections":[{"title":"How the evaluated method works","body":"A BERT-style model treats codons as tokens. Pretraining includes masked-language modelling and an STP task described in the paper, followed by supervised downstream prediction.","source_ids":["codonbert-vaccines-2024"],"source_locator":"Results/Pretrained representation model (paragraph 1); Methods/Pretraining CodonBERT (paragraph 6)"},{"title":"What was evaluated","body":"The linked evaluation record identifies CodonBERT: flu-vaccine mRNA property prediction. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["codonbert-vaccines-2024"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-b2-codonbert-vaccines-2024"}],"facts":[{"label":"Model type","value":"Study-specific predictive method; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["codonbert-vaccines-2024"],"source_locator":"Results/Pretrained representation model (paragraph 1); Methods/Pretraining CodonBERT (paragraph 6)"},{"label":"Architecture / procedure","value":"A BERT-style model treats codons as tokens. Pretraining includes masked-language modelling and an STP task described in the paper, followed by supervised downstream prediction.","status":"source_checked","source_ids":["codonbert-vaccines-2024"],"source_locator":"Results/Pretrained representation model (paragraph 1); Methods/Pretraining CodonBERT (paragraph 6)"},{"label":"Biological inputs","value":"Coding mRNA sequences represented as codons","status":"source_checked","source_ids":["codonbert-vaccines-2024"],"source_locator":"Results (paragraph 1); Methods/Assembly of mRNA sequences for pretraining (paragraph 2)"},{"label":"Outputs","value":"mRNA embeddings and task-specific property predictions","status":"source_checked","source_ids":["codonbert-vaccines-2024"],"source_locator":"Methods/Pretraining CodonBERT (paragraph 6); Results (paragraph 2)"},{"label":"Parameters","value":"Approximately 110 million parameters; 12 layers, 12 attention heads and hidden size 768.","status":"source_checked","source_ids":["codonbert-vaccines-2024"],"source_locator":"Methods/Pretraining CodonBERT (paragraph 5); Methods/Model architecture (paragraph 3)"},{"label":"Known versions / configuration","value":"CodonBERT is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["codonbert-vaccines-2024"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"More than ten million mRNA sequences from multiple organisms; 1% of each category is held out during pretraining.","status":"source_checked","source_ids":["codonbert-vaccines-2024"],"source_locator":"Methods/Pretraining CodonBERT (paragraph 5); Methods/Assembly of mRNA sequences for pretraining (paragraph 1)"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["codonbert-vaccines-2024","evidence-reported-codonbert-vaccines-2024-readme-md"],"source_locator":"Results/Pretrained representation model; Results/Evaluating CodonBERT and comparison to prior methods on supervised learning tasks; Methods/Assembly of mRNA sequences for pretraining; Methods/Model architecture; Methods/Pretraining CodonBERT; Methods/Pretraining Codon2vec; Methods/In vitro transcription, cell culture, and transfections; Methods/Comparisons to other methods; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/Sanofi-Public/CodonBert/blob/451a1b167c06028dfbf2ff7aa2cfdea46fbcc4f4/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-codonbert-vaccines-2024-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","status":"unreported","source_ids":["evidence-reported-codonbert-vaccines-2024-readme-md"],"source_locator":"README.md and repository-root licence-file search"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-codonbert-vaccines-2024-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Codon-level inputs expose coding structure directly instead of requiring the model to recover reading-frame units from single bases.","source_ids":["codonbert-vaccines-2024"],"source_locator":"Results (paragraph 2); Methods/Assembly of mRNA sequences for pretraining (paragraph 2)"}],"limitations":[{"text":"The learned representation and task-specific property heads must be distinguished; performance depends on the mRNA assay and downstream training data.","source_ids":["codonbert-vaccines-2024"],"source_locator":"Methods/Pretraining CodonBERT (paragraph 6); Results/Evaluating CodonBERT and comparison to prior methods on supervised learning tasks (paragraph 3)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Coding mRNA sequences represented as codons","CodonBERT","mRNA embeddings and task-specific property predictions"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["codonbert-vaccines-2024"],"source_locator":"Results/Pretrained representation model (paragraph 1); Methods/Pretraining CodonBERT (paragraph 6)"},"coverage":"limited","gaps":["CodonBERT is the comparison-table label; that label does not specify an immutable weight revision.","A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-cdc9aabf4efc04","kind":"model","name":"Boltz-2","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["molecular-interactions"]},"source_ids":["mpro-pose-affinity-2025"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"Boltz-2","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"This configuration predicts ligand poses for the SARS-CoV-2 and MERS-CoV main proteases in the ASAP challenge setting.","summary_source_ids":["mpro-pose-affinity-2025"],"summary_source_locator":"Conclusions (paragraph 1); Introduction (paragraph 2)","sections":[{"title":"How the evaluated method works","body":"Boltz-2 is the co-folding pose-generation comparator; downstream LRIP-SF affinity fitting is separate from this row.","source_ids":["mpro-pose-affinity-2025"],"source_locator":"Methodologies/Potency Prediction/LRIP-SF/Pose Generation (paragraph 1); Results/Potency Prediction (paragraph 3)"},{"title":"Underlying method and version boundaries","body":"The official Boltz repository publishes separate Boltz-1 and Boltz-2 models. Boltz-2 adds affinity prediction and other changes; these are not retroactively attributed to Boltz-1 evaluations.","source_ids":["evidence-reported-base-boltz-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies Boltz-2: Ligand potency prediction using generated poses. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["mpro-pose-affinity-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-047"}],"facts":[{"label":"Model type","value":"Biomolecular structure predictor; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-boltz-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"Boltz-2 is the co-folding pose-generation comparator; downstream LRIP-SF affinity fitting is separate from this row.","status":"source_checked","source_ids":["mpro-pose-affinity-2025"],"source_locator":"Methodologies/Potency Prediction/LRIP-SF/Pose Generation (paragraph 1); Results/Potency Prediction (paragraph 3)"},{"label":"Biological inputs","value":"Main-protease protein information and ligand structures/SMILES","status":"source_checked","source_ids":["mpro-pose-affinity-2025"],"source_locator":"Methodologies/Pose Prediction/Deep Learning-Based Modeling with DiffDock (paragraph 1); Methodologies/Pose Prediction/Deep\nLearning-Based Modeling with Boltz-2 (paragraph 1)"},{"label":"Outputs","value":"Predicted ligand binding poses","status":"source_checked","source_ids":["mpro-pose-affinity-2025"],"source_locator":"Methodologies/Pose Prediction/Deep Learning-Based Modeling with Gnina (paragraph 2); Methodologies/Potency Prediction/LRIP-SF/Pose Generation (paragraph 2)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["mpro-pose-affinity-2025","evidence-reported-base-boltz-readme-md"],"source_locator":"Methodologies/Pose Prediction/Data Preparation; Methodologies/Pose Prediction/Molecular Docking with\nGlide; Methodologies/Pose Prediction/Molecular Docking with AutoDock Vina; Methodologies/Pose Prediction/Flexible\nLigand Superposition with FlexS; Methodologies/Pose Prediction/Deep Learning-Based Modeling with AlphaFold3; Methodologies/Pose Prediction/Deep Learning-Based Modeling with DiffDock; Methodologies/Pose Prediction/Deep\nLearning-Based Modeling with Boltz-2; Methodologies/Pose Prediction/Deep Learning-Based Modeling with Gnina; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"Boltz-2 is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["mpro-pose-affinity-2025"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"The study contains 770 SARS-CoV-2 training complexes and test sets of 98 SARS-CoV-2 and 97 MERS-CoV complexes; these are study partitions, not necessarily upstream-model training corpora.","status":"source_checked","source_ids":["mpro-pose-affinity-2025"],"source_locator":"Methodologies/Pose Prediction/Data Preparation (paragraph 2); Results/Potency Prediction (paragraph 5)"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["mpro-pose-affinity-2025","evidence-reported-base-boltz-readme-md"],"source_locator":"Methodologies/Pose Prediction/Data Preparation; Methodologies/Pose Prediction/Molecular Docking with\nGlide; Methodologies/Pose Prediction/Molecular Docking with AutoDock Vina; Methodologies/Pose Prediction/Flexible\nLigand Superposition with FlexS; Methodologies/Pose Prediction/Deep Learning-Based Modeling with AlphaFold3; Methodologies/Pose Prediction/Deep Learning-Based Modeling with DiffDock; Methodologies/Pose Prediction/Deep\nLearning-Based Modeling with Boltz-2; Methodologies/Pose Prediction/Deep Learning-Based Modeling with Gnina; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://github.com/jwohlwend/boltz/blob/b1ebfc46ecf57f5414e0d1a6f9027bbb122c53bc/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-boltz-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"MIT (upstream repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-base-boltz-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"MIT for all code and model weights, explicitly stated in the official Boltz README. Historical evaluated weight identity remains separately recorded.","status":"source_checked","source_ids":["evidence-reported-base-boltz-readme-md"],"source_locator":"README.md; license and model release introduction"}],"strengths":[{"text":"Evaluates pose generation and downstream affinity estimation separately.","source_ids":["mpro-pose-affinity-2025"],"source_locator":"Conclusions (paragraph 3); Discussion (paragraph 10)"}],"limitations":[{"text":"An affinity result obtained by LRIP-SF on a predicted pose must not be assigned to the pose generator alone.","source_ids":["mpro-pose-affinity-2025"],"source_locator":"Methodologies/Potency Prediction/LRIP-SF/Pose Generation (paragraph 1); Results/Potency Prediction (paragraph 3)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Main-protease protein information and ligand structures/SMILES","Boltz-2","Predicted ligand binding poses"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["mpro-pose-affinity-2025"],"source_locator":"Methodologies/Potency Prediction/LRIP-SF/Pose Generation (paragraph 1); Results/Potency Prediction (paragraph 3)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","Boltz-2 is the comparison-table label; that label does not specify an immutable weight revision.","A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-d023fbe78bc4df","kind":"model","name":"ERNIE-RNA","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["rna-transcriptomes"]},"source_ids":["ernie-rna-2025"],"links":[],"attributes":{"entity_level":"method","version":"86M","reported_name":"ERNIE-RNA","historical_missing_metadata":{"checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"ERNIE-RNA is an RNA transformer that introduces base-pairing information into its attention mechanism.","summary_source_ids":["ernie-rna-2025"],"summary_source_locator":"Results/ERNIE-RNA learns functional and structural information through pre-training (paragraph 1); Methods/Model architecture (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"A modified BERT uses 12 transformer blocks, 12 attention heads and 768-dimensional token vectors. Base-pairing priors guide attention; downstream tasks use attention-derived structure estimates or adapted prediction heads.","source_ids":["ernie-rna-2025"],"source_locator":"Methods/Model architecture (paragraph 1); Results/The architecture and pre-training of ERNIE-RNA (paragraph 1)"},{"title":"What was evaluated","body":"The linked evaluation record identifies ERNIE-RNA: RNA secondary-structure prediction. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["ernie-rna-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-b2-ernie-rna-2025"}],"facts":[{"label":"Model type","value":"Transformer representation pipeline; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["ernie-rna-2025"],"source_locator":"Methods/Model architecture (paragraph 1); Results/The architecture and pre-training of ERNIE-RNA (paragraph 1)"},{"label":"Architecture / procedure","value":"A modified BERT uses 12 transformer blocks, 12 attention heads and 768-dimensional token vectors. Base-pairing priors guide attention; downstream tasks use attention-derived structure estimates or adapted prediction heads.","status":"source_checked","source_ids":["ernie-rna-2025"],"source_locator":"Methods/Model architecture (paragraph 1); Results/The architecture and pre-training of ERNIE-RNA (paragraph 1)"},{"label":"Biological inputs","value":"RNA nucleotide sequences","status":"source_checked","source_ids":["ernie-rna-2025"],"source_locator":"Methods/Training details (paragraph 1); Methods/RNA 3D closeness prediction (paragraph 1)"},{"label":"Outputs","value":"RNA representations, attention-derived pairing information and downstream task predictions","status":"source_checked","source_ids":["ernie-rna-2025"],"source_locator":"Introduction (paragraph 6); Abstract (paragraph 1)"},{"label":"Parameters","value":"Approximately 86 million parameters","status":"source_checked","source_ids":["ernie-rna-2025"],"source_locator":"Methods/Model architecture (paragraph 1); Results/ERNIE-RNA learns functional and structural information through pre-training (paragraph 3)"},{"label":"Known versions / configuration","value":"86M","status":"source_checked","source_ids":["ernie-rna-2025"],"source_locator":"Table Tab2 (paragraph 1); Table Tab1 (paragraph 1)"},{"label":"Training data / fitting","value":"20.4 million RNAcentral sequences after filtering and redundancy removal from an initial 34-million-sequence collection.","status":"source_checked","source_ids":["ernie-rna-2025"],"source_locator":"Results/The architecture and pre-training of ERNIE-RNA (paragraph 2); Methods/Training dataset (paragraph 1)"},{"label":"Context limits","value":"Pretraining sequences are filtered to no more than 1,022 nucleotides, leaving room for special tokens.","status":"source_checked","source_ids":["ernie-rna-2025"],"source_locator":"Methods/RNA 3D closeness prediction (paragraph 2); Results/RNA contact map prediction (paragraph 1)"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/Bruce-ywj/ERNIE-RNA/blob/43bc06de1088ed03ffd7de918ad4b2c2a3346a43/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-ernie-rna-2025-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"MIT (study repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-ernie-rna-2025-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-ernie-rna-2025-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Evaluates structural information in attention maps before task-specific fine-tuning.","source_ids":["ernie-rna-2025"],"source_locator":"Discussions (paragraph 1); Methods/Downstream tasks/RNA secondary structure prediction/Zero-shot Prediction: (paragraph 1)"}],"limitations":[{"text":"The pretraining corpus is rich in rRNA and tRNA, and the paper explicitly examines composition bias; zero-shot and fine-tuned results represent different settings.","source_ids":["ernie-rna-2025"],"source_locator":"Methods/Downstream tasks/RNA secondary structure prediction (paragraph 1); Results/The architecture and pre-training of ERNIE-RNA (paragraph 3)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["RNA nucleotide sequences","ERNIE-RNA","RNA representations, attention-derived pairing information and downstream task predictions"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["ernie-rna-2025"],"source_locator":"Methods/Model architecture (paragraph 1); Results/The architecture and pre-training of ERNIE-RNA (paragraph 1)"},"coverage":"limited","gaps":["The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-d0677d52d2b9fd","kind":"model","name":"geNomad","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["microbes-communities"]},"source_ids":["lambda-prophage-2026"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"geNomad","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"This configuration is evaluated for phage–bacterium discrimination and genome-wide prophage detection in LAMBDA.","summary_source_ids":["lambda-prophage-2026"],"summary_source_locator":"Materials and Methods/LAMBDA Benchmark Dataset Construction. (paragraph 1); Abstract (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"The benchmark separates embedding probes, fine-tuning, diagnostic tests and genome-wide scans; the associated result identifies which procedure is used.","source_ids":["lambda-prophage-2026"],"source_locator":"Abstract (paragraph 1); Results/Genome-Wide Prophage Detection./Extracting Prophage Signal from Raw Model Predictions. (paragraph 3)"},{"title":"Underlying method and version boundaries","body":"geNomad identifies viral and plasmid sequences, including proviruses, and supplies viral taxonomy and protein-function annotations. These separate outputs have different evaluation definitions.","source_ids":["evidence-reported-base-genomad-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies geNomad: Genome-wide prophage detection. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["lambda-prophage-2026"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-038"}],"facts":[{"label":"Model type","value":"Mobile-genetic-element detector; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-genomad-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"The benchmark separates embedding probes, fine-tuning, diagnostic tests and genome-wide scans; the associated result identifies which procedure is used.","status":"source_checked","source_ids":["lambda-prophage-2026"],"source_locator":"Abstract (paragraph 1); Results/Genome-Wide Prophage Detection./Extracting Prophage Signal from Raw Model Predictions. (paragraph 3)"},{"label":"Biological inputs","value":"Bacterial/phage DNA sequences and genome windows","status":"source_checked","source_ids":["lambda-prophage-2026"],"source_locator":"Materials and Methods/LAMBDA Benchmark Dataset Construction./Segment Subsampling. (paragraph 1); Materials and Methods/LAMBDA Benchmark Dataset Construction./Data Sources. (paragraph 1)"},{"label":"Outputs","value":"Phage classification scores or candidate prophage regions","status":"source_checked","source_ids":["lambda-prophage-2026"],"source_locator":"Results/Genome-Wide Prophage Detection./Scanning for Novel, Unannotated Prophage Regions. (paragraph 2); Materials and Methods/LAMBDA Benchmark Dataset Construction./Prophage Signal Extraction Algorithm. (paragraph 1)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["lambda-prophage-2026","evidence-reported-base-genomad-readme-md"],"source_locator":"Results/Evaluating the Power of Pretrained Sequence Representations.; Materials and Methods/LAMBDA Benchmark Dataset Construction.; Materials and Methods/LAMBDA Benchmark Dataset Construction./Data Sources.; Materials and Methods/LAMBDA Benchmark Dataset Construction./Bacterial contamination filtering via BLAST.; Materials and Methods/LAMBDA Benchmark Dataset Construction./Segment Subsampling.; Materials and Methods/LAMBDA Benchmark Dataset Construction./Data Splitting and Data Leakage Prevention.; Materials and Methods/LAMBDA Benchmark Dataset Construction./GC-Control Dataset.; Materials and Methods/LAMBDA Benchmark Dataset Construction./Class Prediction Control Datasets.; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"geNomad is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["lambda-prophage-2026"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"The benchmark runs the existing geNomad detector as a comparator to the fitted genomic-language-model probes. Its exact upstream training/reference-data release is not identified in the comparison methods.","status":"source_checked","source_ids":["lambda-prophage-2026"],"source_locator":"Materials and Methods / LAMBDA Benchmark Dataset Construction / Comparison with Traditional and Protein-based Models"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["lambda-prophage-2026","evidence-reported-base-genomad-readme-md"],"source_locator":"Results/Evaluating the Power of Pretrained Sequence Representations.; Materials and Methods/LAMBDA Benchmark Dataset Construction.; Materials and Methods/LAMBDA Benchmark Dataset Construction./Data Sources.; Materials and Methods/LAMBDA Benchmark Dataset Construction./Bacterial contamination filtering via BLAST.; Materials and Methods/LAMBDA Benchmark Dataset Construction./Segment Subsampling.; Materials and Methods/LAMBDA Benchmark Dataset Construction./Data Splitting and Data Leakage Prevention.; Materials and Methods/LAMBDA Benchmark Dataset Construction./GC-Control Dataset.; Materials and Methods/LAMBDA Benchmark Dataset Construction./Class Prediction Control Datasets.; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://github.com/apcamargo/genomad/blob/8c5fd0d1722d458a3e8ff50278cdc00ed4a514fc/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-genomad-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"Repository-specific licence text is available; a standard SPDX label has not been established. (upstream repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-base-genomad-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-base-genomad-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Extends genomic-model evaluation beyond eukaryotic regulatory-element labels.","source_ids":["lambda-prophage-2026"],"source_locator":"Abstract (paragraph 1); Background (paragraph 1)"}],"limitations":[{"text":"Candidate regions outside reference annotations are not automatically validated prophages; the paper uses additional gene-content screening.","source_ids":["lambda-prophage-2026"],"source_locator":"Materials and Methods/LAMBDA Benchmark Dataset Construction./Candidate Prophage Screening. (paragraph 1); Results/Genome-Wide Prophage Detection./Scanning for Novel, Unannotated Prophage Regions. (paragraph 2)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Bacterial/phage DNA sequences and genome windows","geNomad","Phage classification scores or candidate prophage regions"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["lambda-prophage-2026"],"source_locator":"Abstract (paragraph 1); Results/Genome-Wide Prophage Detection./Extracting Prophage Signal from Raw Model Predictions. (paragraph 3)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","geNomad is the comparison-table label; that label does not specify an immutable weight revision.","A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-d0d5df2beb02b2","kind":"model","name":"ESM-2","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["proteins-complexes"]},"source_ids":["prime-2026"],"links":[],"attributes":{"entity_level":"method","version":"8M","reported_name":"ESM-2","historical_missing_metadata":{"checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"This protein-language-model configuration is evaluated for viral RBD mutation properties under PRIME.","summary_source_ids":["prime-2026"],"summary_source_locator":"Methods/Model architecture and scale (paragraph 1); Discussion (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"Protein embeddings feed downstream prediction heads, with frozen and fine-tuned encoder conditions explicitly separated.","source_ids":["prime-2026"],"source_locator":"Results/Optimization of neural architectures for real-time deployment (paragraph 1); Methods/Model architecture and scale (paragraph 1)"},{"title":"Underlying method and version boundaries","body":"ESM-2 is a transformer protein language-model family. The official repository exposes residue embeddings, sequence-level pooling and models at several sizes; the study configuration determines which of these is evaluated.","source_ids":["evidence-reported-base-esm-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies ESM-2: Mutated RBD binding prediction. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["prime-2026"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-021"}],"facts":[{"label":"Model type","value":"Protein sequence transformer; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-esm-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"Protein embeddings feed downstream prediction heads, with frozen and fine-tuned encoder conditions explicitly separated.","status":"source_checked","source_ids":["prime-2026"],"source_locator":"Results/Optimization of neural architectures for real-time deployment (paragraph 1); Methods/Model architecture and scale (paragraph 1)"},{"label":"Biological inputs","value":"SARS-CoV-2 receptor-binding-domain sequences","status":"source_checked","source_ids":["prime-2026"],"source_locator":"Results/Benchmark data collection for viral phenotype prediction (paragraph 3); Discussion (paragraph 7)"},{"label":"Outputs","value":"Binding-affinity and expression predictions","status":"source_checked","source_ids":["prime-2026"],"source_locator":"Methods/Computational resources and reproducibility (paragraph 2); Results/Optimization of neural architectures for real-time deployment (paragraph 2)"},{"label":"Parameters","value":"8 million parameters, as identified for this row","status":"source_checked","source_ids":["prime-2026"],"source_locator":"Abstract (paragraph 1); Methods/Clustering and phylogenetic analysis/Clustering (paragraph 1)"},{"label":"Known versions / configuration","value":"8M","status":"source_checked","source_ids":["prime-2026"],"source_locator":"Table Tab1 (paragraph 1); Declarations/Consent for publication (paragraph 1)"},{"label":"Training data / fitting","value":"347,432 RBD sequences in the study; position-stratified partitions test unseen mutational sites.","status":"source_checked","source_ids":["prime-2026"],"source_locator":"Methods/Position-stratified validation protocol/Leakage control (paragraph 1); Background (paragraph 5)"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["prime-2026","evidence-reported-base-esm-readme-md"],"source_locator":"Results/Optimization of neural architectures for real-time deployment; Methods/Data collection and curation; Methods/Data collection and curation/Outbreak dataset; Methods/Data collection and curation/DMS dataset; Methods/Data collection and curation/BetaCov dataset; Methods/Model architecture and scale; Methods/Model architecture and scale/ESM-2 family; Methods/Model architecture and scale/ESM-C family; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://github.com/facebookresearch/esm/blob/2b369911bb5b4b0dda914521b9475cad1656b2ac/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-esm-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"MIT (upstream repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-base-esm-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-base-esm-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Contrasts random splitting with a position-stratified protocol that exposes leakage from closely related variants.","source_ids":["prime-2026"],"source_locator":"Methods/Position-stratified validation protocol (paragraph 1); Methods/Position-stratified validation protocol/Leakage control (paragraph 1)"}],"limitations":[{"text":"Random-split performance is not evidence of generalisation to unseen mutation sites; the row’s model size and adaptation must stay attached to its score.","source_ids":["prime-2026"],"source_locator":"Background (paragraph 5); Discussion (paragraph 2)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["SARS-CoV-2 receptor-binding-domain sequences","ESM-2","Binding-affinity and expression predictions"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["prime-2026"],"source_locator":"Results/Optimization of neural architectures for real-time deployment (paragraph 1); Methods/Model architecture and scale (paragraph 1)"},"coverage":"limited","gaps":["A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-d0e594ec3c0430","kind":"model","name":"DNABERT2-Enhancer","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["dna-genomes"]},"source_ids":["dnabert2-enhancer-2025"],"links":[],"attributes":{"entity_level":"method","version":"not stated in table","reported_name":"DNABERT2-Enhancer","historical_missing_metadata":{"checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"DNABERT2-Enhancer combines a fine-tuned DNA transformer with a CNN to classify enhancers and their activity.","summary_source_ids":["dnabert2-enhancer-2025"],"summary_source_locator":"Abstract (paragraph 1); Results/Comparison of the proposed model with existing methods (paragraph 4)","sections":[{"title":"How the evaluated method works","body":"DNABERT-2 initialises a BERT feature extractor. Transfer learning adapts it to enhancer data, and a convolutional network classifies the resulting sequence features.","source_ids":["dnabert2-enhancer-2025"],"source_locator":"Materials and methods/DNABERT2-enhancer model (paragraph 1); Materials and methods/DNABERT-2 model (paragraph 4)"},{"title":"Underlying method and version boundaries","body":"DNABERT-2 replaces overlapping k-mer tokens with byte-pair encoding and uses ALiBi positional biases. The official 117M model produces 768-dimensional token representations; downstream classifiers and pooling choices are separate configuration details.","source_ids":["evidence-reported-base-dnabert2-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies DNABERT2-Enhancer: enhancer recognition. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["dnabert2-enhancer-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-b2-dnabert2-enhancer-2025"}],"facts":[{"label":"Model type","value":"DNA sequence transformer; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-dnabert2-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"DNABERT-2 initialises a BERT feature extractor. Transfer learning adapts it to enhancer data, and a convolutional network classifies the resulting sequence features.","status":"source_checked","source_ids":["dnabert2-enhancer-2025"],"source_locator":"Materials and methods/DNABERT2-enhancer model (paragraph 1); Materials and methods/DNABERT-2 model (paragraph 4)"},{"label":"Biological inputs","value":"DNA sequence windows","status":"source_checked","source_ids":["dnabert2-enhancer-2025"],"source_locator":"Conclusion (paragraph 1); Discussion (paragraph 3)"},{"label":"Outputs","value":"Enhancer/non-enhancer labels and strong/weak enhancer classification","status":"source_checked","source_ids":["dnabert2-enhancer-2025"],"source_locator":"Introduction (paragraph 7); Materials and methods/Benchmark dataset (paragraph 4)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["dnabert2-enhancer-2025","evidence-reported-base-dnabert2-readme-md"],"source_locator":"Materials and methods/Benchmark dataset; Materials and methods/DNABERT2-enhancer model; Materials and methods/DNABERT-2 model; Materials and methods/CNN model; Results/Comparison of the proposed model with existing methods; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"DNABERT2-Enhancer is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["dnabert2-enhancer-2025"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"Liu and Basith enhancer datasets; Basith includes eight cell-line subsets, 204–2,000-bp sequences and a 60% CD-HIT redundancy threshold.","status":"source_checked","source_ids":["dnabert2-enhancer-2025"],"source_locator":"Materials and methods/Benchmark dataset (paragraph 3); Materials and methods/Benchmark dataset (paragraph 1)"},{"label":"Context limits","value":"The Basith benchmark includes 204–2,000-bp sequences; this is benchmark input coverage, not a validated universal model limit.","status":"source_checked","source_ids":["dnabert2-enhancer-2025"],"source_locator":"Results (paragraph 3); Materials and methods/DNABERT2-enhancer model (paragraph 1)"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://github.com/MAGICS-LAB/DNABERT_2/blob/f25bed9ee20db966dff39e5c1571249d04e36404/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-dnabert2-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"Apache 2.0 (upstream repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-base-dnabert2-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-base-dnabert2-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Explicitly separates enhancer identification from enhancer-strength classification.","source_ids":["dnabert2-enhancer-2025"],"source_locator":"Results (paragraph 4); Introduction (paragraph 7)"}],"limitations":[{"text":"The balanced training sets and cell-line-specific labels limit direct extrapolation to genome-wide enhancer prevalence.","source_ids":["dnabert2-enhancer-2025"],"source_locator":"Results/Comparison of the proposed model with existing methods (paragraph 4); Materials and methods/Benchmark dataset (paragraph 3)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["DNA sequence windows","DNABERT2-Enhancer","Enhancer/non-enhancer labels and strong/weak enhancer classification"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["dnabert2-enhancer-2025"],"source_locator":"Materials and methods/DNABERT2-enhancer model (paragraph 1); Materials and methods/DNABERT-2 model (paragraph 4)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","DNABERT2-Enhancer is the comparison-table label; that label does not specify an immutable weight revision.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-d1cd9a425f9bbd","kind":"model","name":"TU-Fold (aug)","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["rna-transcriptomes"]},"source_ids":["tu-fold-2025"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"TU-Fold (aug)","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"TU-Fold predicts RNA secondary structure as a set of multiclass decisions; this row includes augmentation.","summary_source_ids":["tu-fold-2025"],"summary_source_locator":"Materials and methods/Training method (paragraph 3); Results/Knowledge merge alleviates the performance drop in the cross-RNA-family evaluation (paragraph 2)","sections":[{"title":"How the evaluated method works","body":"Attention and convolutional components generate a matrix-based secondary-structure prediction, with multiclass training designed to reduce the need for complex validity post-processing.","source_ids":["tu-fold-2025"],"source_locator":"Materials and methods/Training method (paragraph 1); Abstract (paragraph 3)"},{"title":"What was evaluated","body":"The linked evaluation record identifies TU-Fold (aug): RNA secondary structure. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["tu-fold-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-013"}],"facts":[{"label":"Model type","value":"Convolutional neural network; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["tu-fold-2025"],"source_locator":"Materials and methods/Training method (paragraph 1); Abstract (paragraph 3)"},{"label":"Architecture / procedure","value":"Attention and convolutional components generate a matrix-based secondary-structure prediction, with multiclass training designed to reduce the need for complex validity post-processing.","status":"source_checked","source_ids":["tu-fold-2025"],"source_locator":"Materials and methods/Training method (paragraph 1); Abstract (paragraph 3)"},{"label":"Biological inputs","value":"RNA nucleotide sequences","status":"source_checked","source_ids":["tu-fold-2025"],"source_locator":"CRediT authorship contribution statement (paragraph 1); Discussion (paragraph 3)"},{"label":"Outputs","value":"RNA secondary-structure pairings","status":"source_checked","source_ids":["tu-fold-2025"],"source_locator":"Materials and methods/Training method (paragraph 3); Materials and methods/Experimental settings (paragraph 6)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["tu-fold-2025","evidence-reported-tu-fold-2025-readme-md"],"source_locator":"Materials and methods/Training method; Materials and methods/Evaluation method; Materials and methods/Data augmentation; Materials and methods/Knowledge merge; Materials and methods/Experimental settings; Results/TU-fold outperforms existing methods without using post-processing steps; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"TU-Fold (aug) is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["tu-fold-2025"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"The paper’s within-family and cross-family RNA-structure partitions; augmentation is part of this named configuration.","status":"source_checked","source_ids":["tu-fold-2025"],"source_locator":"Materials and methods/Data augmentation (paragraph 1); Materials and methods/Training method (paragraph 3)"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["tu-fold-2025","evidence-reported-tu-fold-2025-readme-md"],"source_locator":"Materials and methods/Training method; Materials and methods/Evaluation method; Materials and methods/Data augmentation; Materials and methods/Knowledge merge; Materials and methods/Experimental settings; Results/TU-fold outperforms existing methods without using post-processing steps; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/ygjiyn/tu_fold/blob/f0532b6bf38b2f57baf0ba6afce7766bfc64899b/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-tu-fold-2025-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"MIT (study repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-tu-fold-2025-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-tu-fold-2025-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Separately examines structural validity and generalisation across RNA families.","source_ids":["tu-fold-2025"],"source_locator":"Results/Data augmentation further improves the performance (paragraph 2); Materials and methods/Experimental settings (paragraph 1)"}],"limitations":[{"text":"The augmented setting is distinct from the unaugmented model and from the separate strategy proposed for cross-family transfer.","source_ids":["tu-fold-2025"],"source_locator":"Materials and methods/Knowledge merge (paragraph 2); Discussion (paragraph 2)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["RNA nucleotide sequences","TU-Fold (aug)","RNA secondary-structure pairings"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["tu-fold-2025"],"source_locator":"Materials and methods/Training method (paragraph 1); Abstract (paragraph 3)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","TU-Fold (aug) is the comparison-table label; that label does not specify an immutable weight revision.","A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-d25dab1a9c4fff","kind":"model","name":"ESM-2","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["proteins-complexes"]},"source_ids":["pst-2025"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"ESM-2","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"ESM-2 is the sequence-only baseline in the Protein Structure Transformer study.","summary_source_ids":["pst-2025"],"summary_source_locator":"1 Introduction (paragraph 5); 2 Related work (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"A pretrained protein transformer produces sequence representations without the structural adapters added by PST.","source_ids":["pst-2025"],"source_locator":"4 Experiments/4.1 Experimental setup/4.1.5 Task-specific models (paragraph 1); 2 Related work/2.1 Sequence-based models (paragraph 1)"},{"title":"Underlying method and version boundaries","body":"ESM-2 is a transformer protein language-model family. The official repository exposes residue embeddings, sequence-level pooling and models at several sizes; the study configuration determines which of these is evaluated.","source_ids":["evidence-reported-base-esm-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies ESM-2: Zero-shot variant effect prediction. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["pst-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-024"}],"facts":[{"label":"Model type","value":"Protein sequence transformer; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-esm-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"A pretrained protein transformer produces sequence representations without the structural adapters added by PST.","status":"source_checked","source_ids":["pst-2025"],"source_locator":"4 Experiments/4.1 Experimental setup/4.1.5 Task-specific models (paragraph 1); 2 Related work/2.1 Sequence-based models (paragraph 1)"},{"label":"Biological inputs","value":"Protein amino-acid sequences","status":"source_checked","source_ids":["pst-2025"],"source_locator":"3 Methods/3.1 Evolutionary scale modeling (paragraph 3); 2 Related work (paragraph 1)"},{"label":"Outputs","value":"Protein embeddings and task-specific predictions","status":"source_checked","source_ids":["pst-2025"],"source_locator":"4 Experiments/4.1 Experimental setup/4.1.5 Task-specific models (paragraph 1); 1 Introduction (paragraph 1)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["pst-2025","evidence-reported-base-esm-readme-md"],"source_locator":"3 Methods; 3 Methods/3.1 Evolutionary scale modeling; 3 Methods/3.1 Evolutionary scale modeling/3.1.1 ESM-2 model architecture; 3 Methods/3.2 Protein structure transformer; 3 Methods/3.2 Protein structure transformer/3.2.1 Protein graph representation; 3 Methods/3.2 Protein structure transformer/3.2.2 Protein structure transformer construction; 3 Methods/3.2 Protein structure transformer/3.2.3 Pretraining the PST; 4 Experiments/4.1 Experimental setup/4.1.4 Pretraining; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"ESM-2 is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["pst-2025"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"Masked-language-model pretraining samples approximately 43 million UniRef50 clusters from 138 million UniRef90 sequences, exposing ESM-2 to about 65 million distinct sequences. The sequence-only comparator remains distinct from PST’s structure-adaptation stage.","status":"source_checked","source_ids":["pst-2025"],"source_locator":"Methods / Evolutionary scale modeling"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["pst-2025","evidence-reported-base-esm-readme-md"],"source_locator":"3 Methods; 3 Methods/3.1 Evolutionary scale modeling; 3 Methods/3.1 Evolutionary scale modeling/3.1.1 ESM-2 model architecture; 3 Methods/3.2 Protein structure transformer; 3 Methods/3.2 Protein structure transformer/3.2.1 Protein graph representation; 3 Methods/3.2 Protein structure transformer/3.2.2 Protein structure transformer construction; 3 Methods/3.2 Protein structure transformer/3.2.3 Pretraining the PST; 4 Experiments/4.1 Experimental setup/4.1.4 Pretraining; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://github.com/facebookresearch/esm/blob/2b369911bb5b4b0dda914521b9475cad1656b2ac/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-esm-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"MIT (upstream repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-base-esm-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-base-esm-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Controls for the value of the pretrained sequence representation when testing added structural information.","source_ids":["pst-2025"],"source_locator":"1 Introduction (paragraph 7); 4 Experiments/4.4 Ablation studies (paragraph 1)"}],"limitations":[{"text":"PST variants use additional structure data and adapters; their scores and parameter overhead must not be attributed to this unmodified baseline.","source_ids":["pst-2025"],"source_locator":"4 Experiments/4.4 Ablation studies/4.4.3 Pretraining only structure extractors almost matches full-model pretraining performance (paragraph 2); 4 Experiments/4.4 Ablation studies/4.4.4 PST gains are due to structural information (paragraph 1)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Protein amino-acid sequences","ESM-2","Protein embeddings and task-specific predictions"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["pst-2025"],"source_locator":"4 Experiments/4.1 Experimental setup/4.1.5 Task-specific models (paragraph 1); 2 Related work/2.1 Sequence-based models (paragraph 1)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","ESM-2 is the comparison-table label; that label does not specify an immutable weight revision.","A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-d2c81acf1c42c4","kind":"model","name":"position-aware CNN","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["dna-genomes"]},"source_ids":["enhancer-position-encoding-2024"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"position-aware CNN","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"PDCNN is a position-aware convolutional enhancer classifier built from nucleotide distribution features.","summary_source_ids":["enhancer-position-encoding-2024"],"summary_source_locator":"STAR★Methods/Method details/A model of convolutional neural network based on position awareness of positively and negatively modified nucleotide classes (paragraph 4); STAR★Methods/Method details/A model of convolutional neural network based on position awareness of positively and negatively modified nucleotide classes (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"The POCD-ND encoding measures positional nucleotide patterns; a CNN with two convolutional and fully connected stages is trained with cross-entropy.","source_ids":["enhancer-position-encoding-2024"],"source_locator":"STAR★Methods/Method details/The position-aware encoding of positive and negative modification classes of nucleotides (paragraph 1); STAR★Methods/Method details/A model of convolutional neural network based on position awareness of positively and negatively modified nucleotide classes (paragraph 1)"},{"title":"What was evaluated","body":"The linked evaluation record identifies position-aware CNN: enhancer prediction. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["enhancer-position-encoding-2024"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b4-004"}],"facts":[{"label":"Model type","value":"Convolutional neural network; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["enhancer-position-encoding-2024"],"source_locator":"STAR★Methods/Method details/The position-aware encoding of positive and negative modification classes of nucleotides (paragraph 1); STAR★Methods/Method details/A model of convolutional neural network based on position awareness of positively and negatively modified nucleotide classes (paragraph 1)"},{"label":"Architecture / procedure","value":"The POCD-ND encoding measures positional nucleotide patterns; a CNN with two convolutional and fully connected stages is trained with cross-entropy.","status":"source_checked","source_ids":["enhancer-position-encoding-2024"],"source_locator":"STAR★Methods/Method details/The position-aware encoding of positive and negative modification classes of nucleotides (paragraph 1); STAR★Methods/Method details/A model of convolutional neural network based on position awareness of positively and negatively modified nucleotide classes (paragraph 1)"},{"label":"Biological inputs","value":"DNA windows encoded as positional nucleotide-distribution matrices","status":"source_checked","source_ids":["enhancer-position-encoding-2024"],"source_locator":"STAR★Methods/Method details/The position-aware encoding of positive and negative modification classes of nucleotides (paragraph 5); STAR★Methods/Method details/The position-aware encoding of positive and negative modification classes of nucleotides (paragraph 6)"},{"label":"Outputs","value":"Enhancer classification","status":"source_checked","source_ids":["enhancer-position-encoding-2024"],"source_locator":"STAR★Methods/Quantification and statistical analysis/Model performance evaluation metrics (paragraph 1); Results and discussion/Comparison with existing DNA enhancer predictors (paragraph 2)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["enhancer-position-encoding-2024","evidence-reported-enhancer-position-encoding-2024-readme-md"],"source_locator":"Results and discussion/Impact of different feature encoding methods on model performance; Results and discussion/Comparison with classical machine learning methods; STAR★Methods/Resource availability/Lead contact; STAR★Methods/Resource availability/Materials availability; STAR★Methods/Method details/Datasets; STAR★Methods/Method details/The position-aware encoding of positive and negative modification classes of nucleotides; STAR★Methods/Method details/A model of convolutional neural network based on position awareness of positively and negatively modified nucleotide classes; STAR★Methods/Quantification and statistical analysis/Model performance evaluation metrics; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"position-aware CNN is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["enhancer-position-encoding-2024"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"Human and mouse enhancer datasets described in the paper; sequence lengths and k-mer settings are tested separately.","status":"source_checked","source_ids":["enhancer-position-encoding-2024"],"source_locator":"STAR★Methods/Method details/Datasets (paragraph 1); Results and discussion/Impact of different feature encoding methods on model performance (paragraph 1)"},{"label":"Context limits","value":"The study selects 200-bp windows after comparing 50,100,150,200,250 and 300-bp inputs.","status":"source_checked","source_ids":["enhancer-position-encoding-2024"],"source_locator":"Results and discussion/Comparison with classical machine learning methods (paragraph 2); STAR★Methods/Method details/The position-aware encoding of positive and negative modification classes of nucleotides (paragraph 5)"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/xing1999/PDCNN/blob/ff302344eb1a03bca4802408d03651cb4d2abc99/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-enhancer-position-encoding-2024-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","status":"unreported","source_ids":["evidence-reported-enhancer-position-encoding-2024-readme-md"],"source_locator":"README.md and repository-root licence-file search"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-enhancer-position-encoding-2024-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Provides a conventional learned comparator with explicit position-dependent features rather than foundation-model pretraining.","source_ids":["enhancer-position-encoding-2024"],"source_locator":"STAR★Methods/Method details/The position-aware encoding of positive and negative modification classes of nucleotides (paragraph 6); Introduction (paragraph 5)"}],"limitations":[{"text":"Feature encoding, sequence length and k-mer selection are part of the fitted pipeline and must match the evaluated setting.","source_ids":["enhancer-position-encoding-2024"],"source_locator":"STAR★Methods/Method details/The position-aware encoding of positive and negative modification classes of nucleotides (paragraph 9); STAR★Methods/Method details/A model of convolutional neural network based on position awareness of positively and negatively modified nucleotide classes (paragraph 3)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["DNA windows encoded as positional nucleotide-distribution matrices","position-aware CNN","Enhancer classification"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["enhancer-position-encoding-2024"],"source_locator":"STAR★Methods/Method details/The position-aware encoding of positive and negative modification classes of nucleotides (paragraph 1); STAR★Methods/Method details/A model of convolutional neural network based on position awareness of positively and negatively modified nucleotide classes (paragraph 1)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","position-aware CNN is the comparison-table label; that label does not specify an immutable weight revision.","No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-d326e3c4e3ba20","kind":"model","name":"ESM-2","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["proteins-complexes"]},"source_ids":["proteingym-2023"],"links":[],"attributes":{"entity_level":"method","version":"15B","reported_name":"ESM-2","historical_missing_metadata":{"checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"This protein model is evaluated as a zero-shot mutation-effect scorer in ProteinGym.","summary_source_ids":["proteingym-2023"],"summary_source_locator":"ProteinGym benchmarks (paragraph 1); Introduction (paragraph 4)","sections":[{"title":"How the evaluated method works","body":"The benchmark converts model sequence or structure-conditioned scores into variant rankings and compares them with experimental deep-mutational-scanning measurements.","source_ids":["proteingym-2023"],"source_locator":"ProteinGym benchmarks/Dataset types (paragraph 1); Appendix/A Appendix/A.2 Limitations/Deep mutational scans (paragraph 1)"},{"title":"Underlying method and version boundaries","body":"ESM-2 is a transformer protein language-model family. The official repository exposes residue embeddings, sequence-level pooling and models at several sizes; the study configuration determines which of these is evaluated.","source_ids":["evidence-reported-base-esm-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies ESM-2: Zero-shot substitution mutation effects: stability. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["proteingym-2023"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-017"}],"facts":[{"label":"Model type","value":"Protein sequence transformer; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-esm-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"The benchmark converts model sequence or structure-conditioned scores into variant rankings and compares them with experimental deep-mutational-scanning measurements.","status":"source_checked","source_ids":["proteingym-2023"],"source_locator":"ProteinGym benchmarks/Dataset types (paragraph 1); Appendix/A Appendix/A.2 Limitations/Deep mutational scans (paragraph 1)"},{"label":"Biological inputs","value":"Variant protein sequences; inverse-folding methods additionally require structures","status":"source_checked","source_ids":["proteingym-2023"],"source_locator":"Appendix/A Appendix/A.4 Baselines/A.4.1 Zero-shot baselines/Inverse Folding models (paragraph 1); Appendix/A Appendix/A.1 Social Impact (paragraph 2)"},{"label":"Outputs","value":"Mutation-effect scores and ranked variants","status":"source_checked","source_ids":["proteingym-2023"],"source_locator":"Appendix/A Appendix/A.4 Baselines/A.4.3 Clinical baselines (paragraph 1); Appendix/A Appendix/A.3 Datasets/A.3.2 Clinical datasets (paragraph 1)"},{"label":"Parameters","value":"15 billion parameters, as identified for this row","status":"source_checked","source_ids":["proteingym-2023"],"source_locator":"Appendix/A Appendix/A.5 Detailed performance results/A.5.4 Clinical indel benchmarks (paragraph 1); Appendix/A Appendix/A.5 Detailed performance results/A.5.3 Clinical substitution benchmarks (paragraph 1)"},{"label":"Known versions / configuration","value":"15B","status":"source_checked","source_ids":["proteingym-2023"],"source_locator":"Appendix/A Appendix/A.5 Detailed performance results/A.5.4 Clinical indel benchmarks (paragraph 1); Appendix/A Appendix/A.5 Detailed performance results/A.5.3 Clinical substitution benchmarks (paragraph 1)"},{"label":"Training data / fitting","value":"The linked ProteinGym row is a zero-shot evaluation of a pretrained protein language model: no DMS phenotype labels are used to fit this score. Protein-family-specific alignment training and supervised assay predictors are separate benchmark settings.","status":"source_checked","source_ids":["proteingym-2023"],"source_locator":"Evaluation framework / Zero-shot benchmarks / Baselines; Appendix A.4.1"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["proteingym-2023","evidence-reported-base-esm-readme-md"],"source_locator":"ProteinGym benchmarks/Model training regime; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://github.com/facebookresearch/esm/blob/2b369911bb5b4b0dda914521b9475cad1656b2ac/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-esm-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"MIT (upstream repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-base-esm-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-base-esm-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Uses standardised evaluation across diverse experimental assays.","source_ids":["proteingym-2023"],"source_locator":"Introduction (paragraph 2); Abstract (paragraph 1)"}],"limitations":[{"text":"Different models receive different evolutionary/structural information; assay fitness, organismal fitness and clinical pathogenicity are not interchangeable endpoints.","source_ids":["proteingym-2023"],"source_locator":"ProteinGym benchmarks (paragraph 1); Appendix/A Appendix/A.3 Datasets/A.3.1 DMS assays/Processing of large thermostability dataset (paragraph 2)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Variant protein sequences","ESM-2","Mutation-effect scores and ranked variants"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["proteingym-2023"],"source_locator":"ProteinGym benchmarks/Dataset types (paragraph 1); Appendix/A Appendix/A.2 Limitations/Deep mutational scans (paragraph 1)"},"coverage":"limited","gaps":["A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-d3fd83835a2d44","kind":"model","name":"BiRNA-BERT","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["rna-transcriptomes"]},"source_ids":["birna-bert-2025"],"links":[],"attributes":{"entity_level":"method","version":"not stated in table","reported_name":"BiRNA-BERT","historical_missing_metadata":{"checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"BiRNA-BERT is an RNA transformer that combines nucleotide and byte-pair tokenisation for structural and longer-sequence tasks.","summary_source_ids":["birna-bert-2025"],"summary_source_locator":"Abstract (paragraph 1); Discussion (paragraph 2)","sections":[{"title":"How the evaluated method works","body":"An ALiBi-equipped transformer encoder is pretrained with two tokenisation schemes. Nucleotide tokens retain position resolution; BPE compresses longer sequences. Tokenisation is selected according to input length.","source_ids":["birna-bert-2025"],"source_locator":"Introduction (paragraph 4); Methods/Adaptive tokenization and dual pretraining (paragraph 1)"},{"title":"What was evaluated","body":"The linked evaluation record identifies BiRNA-BERT: extremely long RNA species classification. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["birna-bert-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-b2-birna-bert-2025"}],"facts":[{"label":"Model type","value":"Transformer representation pipeline; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["birna-bert-2025"],"source_locator":"Introduction (paragraph 4); Methods/Adaptive tokenization and dual pretraining (paragraph 1)"},{"label":"Architecture / procedure","value":"An ALiBi-equipped transformer encoder is pretrained with two tokenisation schemes. Nucleotide tokens retain position resolution; BPE compresses longer sequences. Tokenisation is selected according to input length.","status":"source_checked","source_ids":["birna-bert-2025"],"source_locator":"Introduction (paragraph 4); Methods/Adaptive tokenization and dual pretraining (paragraph 1)"},{"label":"Biological inputs","value":"RNA nucleotide sequences","status":"source_checked","source_ids":["birna-bert-2025"],"source_locator":"Methods/Adaptive tokenization and dual pretraining (paragraph 3); Methods/Adaptive tokenization and dual pretraining (paragraph 5)"},{"label":"Outputs","value":"RNA sequence or nucleotide representations for downstream prediction heads","status":"source_checked","source_ids":["birna-bert-2025"],"source_locator":"Results (paragraph 4); Methods/Statistics and reproducibility (paragraph 1)"},{"label":"Parameters","value":"The paper is internally inconsistent: model descriptions use 117M, while Results / Computational resources for pretraining / BiRNA-BERT calls it 116M. An exact reconciled count is unreported.","status":"unreported","source_ids":["birna-bert-2025"],"source_locator":"Results (paragraph 2); Results/Comparison of computational efficiency/Computational resources for pretraining/BiRNA-BERT (paragraph 2)"},{"label":"Known versions / configuration","value":"BiRNA-BERT is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["birna-bert-2025"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"Approximately 36 million RNAcentral ncRNA sequences; the Methods describe two epochs and 26.42 billion training tokens. Hardware and elapsed-time statements differ across sections and are not treated as a reproducible compute budget.","status":"source_checked","source_ids":["birna-bert-2025"],"source_locator":"Methods/Pretraining dataset (paragraph 1); Methods/Pretraining configuration (paragraph 1)"},{"label":"Context limits","value":"For the long-sequence comparison, nucleotide-token inputs are truncated to 1,022 tokens for comparability; BPE inputs are not truncated and reach 807 tokens. ALiBi support does not establish an unlimited practical context.","status":"source_checked","source_ids":["birna-bert-2025"],"source_locator":"Results/Impact of dual tokenization is more significant in long sequences (paragraph 2); Results/Empirical perplexity analysis of different RNA language models (paragraph 6)"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/buetnlpbio/BiRNA-BERT/blob/14dc86b1b44c266f01025fc425103f2878646b39/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-birna-bert-2025-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","status":"unreported","source_ids":["evidence-reported-birna-bert-2025-readme-md"],"source_locator":"README.md and repository-root licence-file search"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-birna-bert-2025-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Supports nucleotide-level tasks and longer-context tasks within a dual-tokenisation framework.","source_ids":["birna-bert-2025"],"source_locator":"Results/Nucleotide-level task (paragraph 1); Results/NUC tokenization is preferable over BPE if memory allows (paragraph 4)"}],"limitations":[{"text":"Relative positional encoding and compressed tokens do not prove unlimited useful biological context; practical memory and task-specific validation still constrain use.","source_ids":["birna-bert-2025"],"source_locator":"Methods/Tokenization strategies for biological foundational models (paragraph 1); Methods/Positional encoding in the transformer architecture (paragraph 1)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["RNA nucleotide sequences","BiRNA-BERT","RNA sequence or nucleotide representations for downstream prediction heads"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["birna-bert-2025"],"source_locator":"Introduction (paragraph 4); Methods/Adaptive tokenization and dual pretraining (paragraph 1)"},"coverage":"limited","gaps":["The paper is internally inconsistent: model descriptions use 117M, while Results / Computational resources for pretraining / BiRNA-BERT calls it 116M. An exact reconciled count is unreported.","BiRNA-BERT is the comparison-table label; that label does not specify an immutable weight revision.","No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-d5bc536ca6f0d3","kind":"model","name":"Caduceus (character tokens)","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["dna-genomes"]},"source_ids":["genomic-tokenizer-selection-2025"],"links":[],"attributes":{"entity_level":"method","version":"3.9M parameter variant","reported_name":"Caduceus (character tokens)","historical_missing_metadata":{"checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"This Caduceus configuration is a character-token baseline in a study of genomic tokenisation.","summary_source_ids":["genomic-tokenizer-selection-2025"],"summary_source_locator":"1 Introduction (paragraph 9); 4 Discussion (paragraph 4)","sections":[{"title":"How the evaluated method works","body":"A state-space genomic sequence model processes individual nucleotide tokens. The study compares character, k-mer and byte-pair schemes across downstream classifiers.","source_ids":["genomic-tokenizer-selection-2025"],"source_locator":"1 Introduction (paragraph 9); Abstract (paragraph 1)"},{"title":"Underlying method and version boundaries","body":"Caduceus exposes distinct Ph and PS configurations. The documented Ph-131k checkpoint uses 16 layers, width 256 and reverse-complement data augmentation; PS implements reverse-complement equivariance. These training choices are not interchangeable.","source_ids":["evidence-reported-base-caduceus-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies Caduceus (character tokens): regulatory sequence classification. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["genomic-tokenizer-selection-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-b2-genomic-tokenizer-selection-2025"}],"facts":[{"label":"Model type","value":"DNA state-space model; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-caduceus-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"A state-space genomic sequence model processes individual nucleotide tokens. The study compares character, k-mer and byte-pair schemes across downstream classifiers.","status":"source_checked","source_ids":["genomic-tokenizer-selection-2025"],"source_locator":"1 Introduction (paragraph 9); Abstract (paragraph 1)"},{"label":"Biological inputs","value":"DNA sequences represented by individual nucleotide characters","status":"source_checked","source_ids":["genomic-tokenizer-selection-2025"],"source_locator":"1 Introduction (paragraph 6); 1 Introduction (paragraph 3)"},{"label":"Outputs","value":"Task-specific genomic classification predictions","status":"source_checked","source_ids":["genomic-tokenizer-selection-2025"],"source_locator":"2 Materials and methods/2.6 Genomic tasks (paragraph 4); 2 Materials and methods/2.6 Genomic tasks (paragraph 6)"},{"label":"Parameters","value":"3.9-million-parameter variant as reported for this row","status":"source_checked","source_ids":["genomic-tokenizer-selection-2025"],"source_locator":"2 Materials and methods/2.3 Metrics (paragraph 1); 2 Materials and methods/2.6 Genomic tasks (paragraph 6)"},{"label":"Known versions / configuration","value":"3.9M parameter variant","status":"source_checked","source_ids":["genomic-tokenizer-selection-2025"],"source_locator":"2 Materials and methods/2.6 Genomic tasks (paragraph 6); 5 Conclusions (paragraph 1)"},{"label":"Training data / fitting","value":"The tokenisation study’s model pretraining and 44 classification fine-tuning tasks; each task retains its own adaptation and split.","status":"source_checked","source_ids":["genomic-tokenizer-selection-2025"],"source_locator":"2 Materials and methods/2.3 Metrics (paragraph 1); 4 Discussion (paragraph 5)"},{"label":"Context limits","value":"The paper uses a 4,000-nucleotide context for the compared genomic tokenisation models.","status":"source_checked","source_ids":["genomic-tokenizer-selection-2025"],"source_locator":"1 Introduction (paragraph 5); 1 Introduction (paragraph 9)"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://github.com/kuleshov-group/caduceus/blob/0060a6d8079b6a040fc55d505e15972a327b70a6/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-caduceus-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"Apache 2.0 (upstream repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-base-caduceus-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-base-caduceus-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Single-base tokens preserve resolution needed for position-sensitive tasks such as splice-site prediction.","source_ids":["genomic-tokenizer-selection-2025"],"source_locator":"4 Discussion (paragraph 4); 4 Discussion (paragraph 2)"}],"limitations":[{"text":"Tokenisation benefits are task-dependent; the paper reports different behaviour for SARS-CoV-2 classification and does not establish a universal best tokenizer.","source_ids":["genomic-tokenizer-selection-2025"],"source_locator":"4 Discussion (paragraph 5); 3 Results/3.1 Tokenization differentially impacts performance of gLMs on specific benchmark tasks (paragraph 1)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["DNA sequences represented by individual nucleotide characters","Caduceus (character tokens)","Task-specific genomic classification predictions"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["genomic-tokenizer-selection-2025"],"source_locator":"1 Introduction (paragraph 9); Abstract (paragraph 1)"},"coverage":"limited","gaps":["The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-d60f505aabb19c","kind":"model","name":"Geneformer","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["cells-tissues"]},"source_ids":["scelmo-2025"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"Geneformer","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"This single-cell foundation-model configuration is a comparator in the scELMo study.","summary_source_ids":["scelmo-2025"],"summary_source_locator":"Results/scELMo for clustering and batch effect correction. (paragraph 1); Methods/Data pre-processing. (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"Pretrained cell/gene representations are evaluated on the paper’s clustering, integration, annotation or perturbation tasks, with the associated evaluation retaining the task-specific procedure.","source_ids":["scelmo-2025"],"source_locator":"Methods/scELMo under the fine-tuning framework. (paragraph 1); Methods/Data pre-processing./Metrics. (paragraph 12)"},{"title":"Underlying method and version boundaries","body":"Geneformer ranks genes by expression scaled against its pretraining corpus, then uses a transformer encoder with a masked-gene objective. V1 and V2 have different corpora, vocabularies, sizes and context limits, so a historical paper name is not replaced with today’s default checkpoint.","source_ids":["evidence-reported-base-geneformer-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies Geneformer: Cell-type annotation. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["scelmo-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-030"}],"facts":[{"label":"Model type","value":"Single-cell transformer; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-geneformer-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"Pretrained cell/gene representations are evaluated on the paper’s clustering, integration, annotation or perturbation tasks, with the associated evaluation retaining the task-specific procedure.","status":"source_checked","source_ids":["scelmo-2025"],"source_locator":"Methods/scELMo under the fine-tuning framework. (paragraph 1); Methods/Data pre-processing./Metrics. (paragraph 12)"},{"label":"Biological inputs","value":"Single-cell gene-expression measurements","status":"source_checked","source_ids":["scelmo-2025"],"source_locator":"Methods/Data pre-processing./Metrics. (paragraph 15); Methods/Data pre-processing./Metrics. (paragraph 12)"},{"label":"Outputs","value":"Cell embeddings or adapted task predictions","status":"source_checked","source_ids":["scelmo-2025"],"source_locator":"Methods/Problem definition. (paragraph 4); Results/scELMo for clustering and batch effect correction. (paragraph 1)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["scelmo-2025","evidence-reported-base-geneformer-readme-md"],"source_locator":"Methods/Problem definition.; Methods/Method explanation.; Methods/scELMo under the zero-shot learning framework.; Methods/scELMo under the fine-tuning framework.; Methods/Data pre-processing.; Methods/Data pre-processing./Metrics.; Methods/Explanations of baseline models.; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"Geneformer is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["scelmo-2025"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"The comparison includes frozen and fine-tuned embedding settings followed by k-nearest-neighbour classification. Table 1 notes that some results were taken from GenePT; the exact origin and adaptation must therefore remain attached to each row.","status":"source_checked","source_ids":["scelmo-2025"],"source_locator":"Methods / Explanations of baseline models; Table 1 caption"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["scelmo-2025","evidence-reported-base-geneformer-readme-md"],"source_locator":"Methods/Problem definition.; Methods/Method explanation.; Methods/scELMo under the zero-shot learning framework.; Methods/scELMo under the fine-tuning framework.; Methods/Data pre-processing.; Methods/Data pre-processing./Metrics.; Methods/Explanations of baseline models.; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://huggingface.co/ctheodoris/Geneformer/blob/1f7fbae4e469a5f4f1af8c111a529cfe1b3829f5/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-geneformer-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","status":"unreported","source_ids":["evidence-reported-base-geneformer-readme-md"],"source_locator":"README.md and repository-root licence-file search"},{"label":"Weights licence","value":"Apache 2.0 is declared in the official Geneformer model-card metadata; the exact historical configuration still needs its checkpoint identity.","status":"source_checked","source_ids":["evidence-reported-base-geneformer-readme-md"],"source_locator":"README.md front matter, license field; model-version list"}],"strengths":[{"text":"Provides a learned-expression reference for comparison with text-derived biological embeddings.","source_ids":["scelmo-2025"],"source_locator":"Methods/scELMo under the fine-tuning framework. (paragraph 4); Discussion (paragraph 2)"}],"limitations":[{"text":"Different downstream tasks and adaptation settings cannot be combined into a single model-quality claim.","source_ids":["scelmo-2025"],"source_locator":"Introduction (paragraph 3); Results/Overview of scELMo. (paragraph 2)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Single-cell gene-expression measurements","Geneformer","Cell embeddings or adapted task predictions"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["scelmo-2025"],"source_locator":"Methods/scELMo under the fine-tuning framework. (paragraph 1); Methods/Data pre-processing./Metrics. (paragraph 12)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","Geneformer is the comparison-table label; that label does not specify an immutable weight revision.","A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-d6a7fa854437e8","kind":"model","name":"scGPT + residual geometry","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["cells-tissues"]},"source_ids":["single-cell-residual-geometry-2026"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"scGPT + residual geometry","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"This pipeline tests whether scGPT residual-stream geometry adds signal for gene-regulatory links beyond expression features.","summary_source_ids":["single-cell-residual-geometry-2026"],"summary_source_locator":"Results/Residual-stream geometry carries regulatory signal beyond expression confounds (paragraph 1); Background (paragraph 7)","sections":[{"title":"How the evaluated method works","body":"Gene vectors extracted from frozen representations yield cosine, centred-cosine, PCA-projected or multilayer geometry features. Supervised comparison with expression controls tests incremental regulatory-link prediction.","source_ids":["single-cell-residual-geometry-2026"],"source_locator":"Methods/Geometric feature extraction (paragraph 2); Results/Residual-stream geometry carries regulatory signal beyond expression confounds (paragraph 1)"},{"title":"Underlying method and version boundaries","body":"The official scGPT implementation supplies pretrained checkpoints and separate workflows for embedding extraction, cell annotation, integration and perturbation modelling. The checkpoint and adaptation procedure must be identified separately for each result.","source_ids":["evidence-reported-base-scgpt-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies scGPT + residual geometry: gene-regulatory signal prediction. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["single-cell-residual-geometry-2026"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b4-013"}],"facts":[{"label":"Model type","value":"Single-cell foundation model; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-scgpt-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"Gene vectors extracted from frozen representations yield cosine, centred-cosine, PCA-projected or multilayer geometry features. Supervised comparison with expression controls tests incremental regulatory-link prediction.","status":"source_checked","source_ids":["single-cell-residual-geometry-2026"],"source_locator":"Methods/Geometric feature extraction (paragraph 2); Results/Residual-stream geometry carries regulatory signal beyond expression confounds (paragraph 1)"},{"label":"Biological inputs","value":"Tabula Sapiens single-cell expression and curated TRRUST TF–target edges","status":"source_checked","source_ids":["single-cell-residual-geometry-2026"],"source_locator":"Methods/Data sources and preprocessing (paragraph 1); Results/Residual-stream geometry carries regulatory signal beyond expression confounds (paragraph 1)"},{"label":"Outputs","value":"Regulatory-edge prediction scores and incremental performance over expression controls","status":"source_checked","source_ids":["single-cell-residual-geometry-2026"],"source_locator":"Methods/Ensembling with expression-based GRN inference (paragraph 1); Results/Residual-stream geometry carries regulatory signal beyond expression confounds (paragraph 1)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["single-cell-residual-geometry-2026","evidence-reported-base-scgpt-readme-md"],"source_locator":"Results/Geometric signal is tissue-dependent but recoverable through methodological refinement; Discussion/Complementarity between model architectures; Methods/Data sources and preprocessing; Methods/Foundation models; Methods/Geometric feature extraction; Methods/Edge classification framework; Methods/Evaluation protocol; Methods/Null controls; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"scGPT + residual geometry is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["single-cell-residual-geometry-2026"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"Four tissue contexts with repeated cross-validation, leave-TF-out/target-out tests and matched-negative controls.","status":"source_checked","source_ids":["single-cell-residual-geometry-2026"],"source_locator":"Conclusions (paragraph 1); Methods/Edge classification framework (paragraph 1)"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["single-cell-residual-geometry-2026","evidence-reported-base-scgpt-readme-md"],"source_locator":"Results/Geometric signal is tissue-dependent but recoverable through methodological refinement; Discussion/Complementarity between model architectures; Methods/Data sources and preprocessing; Methods/Foundation models; Methods/Geometric feature extraction; Methods/Edge classification framework; Methods/Evaluation protocol; Methods/Null controls; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://github.com/bowang-lab/scGPT/blob/cebd6fae655b9c585a4807daa3ac31bb764f06b4/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-scgpt-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"MIT (upstream repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-base-scgpt-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-base-scgpt-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Includes label-permutation and geometry-shuffle controls to test whether added geometric features carry incremental signal.","source_ids":["single-cell-residual-geometry-2026"],"source_locator":"Results/Geometric signal is tissue-dependent but recoverable through methodological refinement (paragraph 2); Methods/Null controls (paragraph 2)"}],"limitations":[{"text":"Predictive association with curated links is not proof of causality, and tissue-specific effects must not be universalised.","source_ids":["single-cell-residual-geometry-2026"],"source_locator":"Methods/Edge classification framework (paragraph 1); Results/Residual-stream geometry carries regulatory signal beyond expression confounds (paragraph 1)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Tabula Sapiens single-cell expression and curated TRRUST TF–target edges","scGPT + residual geometry","Regulatory-edge prediction scores and incremental performance over expression controls"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["single-cell-residual-geometry-2026"],"source_locator":"Methods/Geometric feature extraction (paragraph 2); Results/Residual-stream geometry carries regulatory signal beyond expression confounds (paragraph 1)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","scGPT + residual geometry is the comparison-table label; that label does not specify an immutable weight revision.","A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-d9a06805b36b8a","kind":"model","name":"Boltz-1","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["molecular-interactions"]},"source_ids":["boltz-stereochemistry-2025"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"Boltz-1","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"This protein–ligand pose predictor is evaluated for stereochemical validity as well as pose accuracy.","summary_source_ids":["boltz-stereochemistry-2025"],"summary_source_locator":"Results/Benchmark of Protein–Ligand\nComplex Structure Prediction (paragraph 6); Methods/Benchmark Dataset (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"The study compares generated ligand poses and protein–ligand complexes under a shared geometry assessment. The unrestrained baseline is distinct from the proposed restraint-guided Boltz inference.","source_ids":["boltz-stereochemistry-2025"],"source_locator":"Results/Case Studies for Bond and Angle Geometries of Ligands (paragraph 3); Results/Structure Prediction Using\nRestraint-Guided Inference (paragraph 1)"},{"title":"Underlying method and version boundaries","body":"The official Boltz repository publishes separate Boltz-1 and Boltz-2 models. Boltz-2 adds affinity prediction and other changes; these are not retroactively attributed to Boltz-1 evaluations.","source_ids":["evidence-reported-base-boltz-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies Boltz-1: Protein–ligand pose prediction. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["boltz-stereochemistry-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-045"}],"facts":[{"label":"Model type","value":"Biomolecular structure predictor; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-boltz-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"The study compares generated ligand poses and protein–ligand complexes under a shared geometry assessment. The unrestrained baseline is distinct from the proposed restraint-guided Boltz inference.","status":"source_checked","source_ids":["boltz-stereochemistry-2025"],"source_locator":"Results/Case Studies for Bond and Angle Geometries of Ligands (paragraph 3); Results/Structure Prediction Using\nRestraint-Guided Inference (paragraph 1)"},{"label":"Biological inputs","value":"Protein and ligand information under the method-specific docking or co-folding protocol","status":"source_checked","source_ids":["boltz-stereochemistry-2025"],"source_locator":"Results/Benchmark of Protein–Ligand\nComplex Structure Prediction (paragraph 1); Results/Case Studies for Bond and Angle Geometries of Ligands (paragraph 1)"},{"label":"Outputs","value":"Predicted protein–ligand binding poses","status":"source_checked","source_ids":["boltz-stereochemistry-2025"],"source_locator":"Results/Case Studies for Bond and Angle Geometries of Ligands (paragraph 2); Methods/Implementation of the Restraints (paragraph 1)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["boltz-stereochemistry-2025","evidence-reported-base-boltz-readme-md"],"source_locator":"Methods/Benchmark Dataset; Methods/Evaluation Metrics; Methods/Improving Geometry through Restraints; Methods/Implementation of the Restraints; Methods/Multiple Sequence Alignment and Structure Prediction; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"Boltz-1 is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["boltz-stereochemistry-2025"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"The evaluation separates structures before and after training-date cutoffs and adds chemical/protein similarity filters.","status":"source_checked","source_ids":["boltz-stereochemistry-2025"],"source_locator":"Methods/Benchmark Dataset (paragraph 3); Methods/Benchmark Dataset (paragraph 4)"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["boltz-stereochemistry-2025","evidence-reported-base-boltz-readme-md"],"source_locator":"Methods/Benchmark Dataset; Methods/Evaluation Metrics; Methods/Improving Geometry through Restraints; Methods/Implementation of the Restraints; Methods/Multiple Sequence Alignment and Structure Prediction; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://github.com/jwohlwend/boltz/blob/b1ebfc46ecf57f5414e0d1a6f9027bbb122c53bc/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-boltz-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"MIT (upstream repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-base-boltz-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"MIT for all code and model weights, explicitly stated in the official Boltz README. Historical evaluated weight identity remains separately recorded.","status":"source_checked","source_ids":["evidence-reported-base-boltz-readme-md"],"source_locator":"README.md; license and model release introduction"}],"strengths":[{"text":"Reports chirality, bond geometry and steric clashes alongside RMSD.","source_ids":["boltz-stereochemistry-2025"],"source_locator":"Results/Benchmark of Protein–Ligand\nComplex Structure Prediction (paragraph 6); Results/Structure Prediction Using\nRestraint-Guided Inference (paragraph 4)"}],"limitations":[{"text":"Pose accuracy deteriorates on filtered post-cutoff data, and low RMSD alone does not imply physically valid ligand geometry.","source_ids":["boltz-stereochemistry-2025"],"source_locator":"Results/Benchmark of Protein–Ligand\nComplex Structure Prediction (paragraph 6); Methods/Evaluation Metrics (paragraph 3)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Protein and ligand information under the method-specific docking or co-folding protocol","Boltz-1","Predicted protein–ligand binding poses"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["boltz-stereochemistry-2025"],"source_locator":"Results/Case Studies for Bond and Angle Geometries of Ligands (paragraph 3); Results/Structure Prediction Using\nRestraint-Guided Inference (paragraph 1)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","Boltz-1 is the comparison-table label; that label does not specify an immutable weight revision.","A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-de89576d8d316b","kind":"model","name":"AK-score-single","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["molecular-interactions"]},"source_ids":["akscore-2020"],"links":[],"attributes":{"entity_level":"method","version":"single; learning rate 0.0007","reported_name":"AK-score-single","historical_missing_metadata":{"checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"AK-score predicts protein–ligand binding affinity from a three-dimensional complex.","summary_source_ids":["akscore-2020"],"summary_source_locator":"3. Methods/3.5. Performance Assessment (paragraph 8); 3. Methods/3.2. Convolutional Neural Network (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"The binding pocket and ligand are voxelised on a 30 Å cube with 1 Å spacing. Multichannel 3D convolutional networks learn interaction patterns; the ensemble averages independently trained models, while the single configuration uses one network.","source_ids":["akscore-2020"],"source_locator":"3. Methods/3.2. Convolutional Neural Network (paragraph 1); 3. Methods/3.4. Ensemble Prediction (paragraph 1)"},{"title":"What was evaluated","body":"The linked evaluation record identifies AK-score-single: Protein–ligand binding affinity scoring. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["akscore-2020"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b3-049"}],"facts":[{"label":"Model type","value":"Convolutional neural network; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["akscore-2020"],"source_locator":"3. Methods/3.2. Convolutional Neural Network (paragraph 1); 3. Methods/3.4. Ensemble Prediction (paragraph 1)"},{"label":"Architecture / procedure","value":"The binding pocket and ligand are voxelised on a 30 Å cube with 1 Å spacing. Multichannel 3D convolutional networks learn interaction patterns; the ensemble averages independently trained models, while the single configuration uses one network.","status":"source_checked","source_ids":["akscore-2020"],"source_locator":"3. Methods/3.2. Convolutional Neural Network (paragraph 1); 3. Methods/3.4. Ensemble Prediction (paragraph 1)"},{"label":"Biological inputs","value":"Protein–ligand complex coordinates represented as atomic-density grids","status":"source_checked","source_ids":["akscore-2020"],"source_locator":"3. Methods/3.2. Convolutional Neural Network (paragraph 1); 3. Methods/3.2. Convolutional Neural Network (paragraph 3)"},{"label":"Outputs","value":"Predicted binding affinity","status":"source_checked","source_ids":["akscore-2020"],"source_locator":"3. Methods/3.5. Performance Assessment (paragraph 8); 3. Methods/3.5. Performance Assessment (paragraph 10)"},{"label":"Parameters","value":"Each network has 1,294,925 parameters:1,293,447 trainable and 1,478 non-trainable. The ensemble uses separately trained networks; this is the per-network count.","status":"source_checked","source_ids":["akscore-2020"],"source_locator":"Methods3.3 Network Architecture; total-parameter paragraph"},{"label":"Known versions / configuration","value":"single; learning rate 0.0007","status":"source_checked","source_ids":["akscore-2020"],"source_locator":"Table ijms-21-08424-t002 (paragraph 1); Table ijms-21-08424-t001 (paragraph 1)"},{"label":"Training data / fitting","value":"3,772 PDBbind-2016 refined-set complexes for training; the 285-complex core set is held out for testing.","status":"source_checked","source_ids":["akscore-2020"],"source_locator":"2. Results and Discussion/2.4. Assessment with an Additional Dataset (paragraph 1); 3. Methods/3.5. Performance Assessment (paragraph 14)"},{"label":"Context limits","value":"30 × 30 × 30 spatial grid at 1 Å spacing","status":"source_checked","source_ids":["akscore-2020"],"source_locator":"3. Methods/3.2. Convolutional Neural Network (paragraph 1); 3. Methods/3.2. Convolutional Neural Network (paragraph 4)"},{"label":"Access","value":"A public release of the original 2020 AK-score model was not established from the paper or targeted official-repository search. AK-Score2 is a separate later method and is not substituted.","status":"unreported","source_ids":["akscore-2020"],"source_locator":"Complete 2020 paper, including Methods and Supplementary Materials statement; exact-name repository discovery"},{"label":"Code licence","value":"No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","status":"unreported","source_ids":["akscore-2020"],"source_locator":"4. Conclusions (paragraph 1); 3. Methods/3.5. Performance Assessment (paragraph 14)"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["akscore-2020"],"source_locator":"2. Results and Discussion/2.5. Identifying Hot Spots for Binding Affinity Determination Using Grad-CAM (paragraph 1); 4. Conclusions (paragraph 1)"}],"strengths":[{"text":"The same framework supports affinity regression and relative affinity ranking within the evaluated structural benchmark.","source_ids":["akscore-2020"],"source_locator":"3. Methods/3.5. Performance Assessment (paragraph 11); 3. Methods/3.5. Performance Assessment (paragraph 3)"}],"limitations":[{"text":"Requires a complex structure and inherits the chemical and structural coverage of PDBbind; a favourable core-set result does not establish prospective docking accuracy.","source_ids":["akscore-2020"],"source_locator":"Abstract (paragraph 1); 2. Results and Discussion/2.5. Identifying Hot Spots for Binding Affinity Determination Using Grad-CAM (paragraph 2)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Protein–ligand complex coordinates represented as atomic-density grids","AK-score-single","Predicted binding affinity"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["akscore-2020"],"source_locator":"3. Methods/3.2. Convolutional Neural Network (paragraph 1); 3. Methods/3.4. Ensemble Prediction (paragraph 1)"},"coverage":"limited","gaps":["A public release of the original 2020 AK-score model was not established from the paper or targeted official-repository search. AK-Score2 is a separate later method and is not substituted.","No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-df0efcc0346224","kind":"model","name":"binding-affinity meta-model","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["molecular-interactions"]},"source_ids":["ligand-affinity-meta-model-2024"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"binding-affinity meta-model","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"This binding-affinity meta-model combines predictions from docking and sequence-based learning models.","summary_source_ids":["ligand-affinity-meta-model-2024"],"summary_source_locator":"Results/Docking Tools (paragraph 4); Results/Meta-models of Docking and DL Tools (paragraph 3)","sections":[{"title":"How the evaluated method works","body":"Force-field-derived empirical docking scores and sequence-based deep-learning estimates feed a learned meta-model, optionally with physicochemical or molecular descriptors.","source_ids":["ligand-affinity-meta-model-2024"],"source_locator":"Abstract (paragraph 1); Results/Meta-models of Docking and DL Tools (paragraph 1)"},{"title":"What was evaluated","body":"The linked evaluation record identifies binding-affinity meta-model: protein-ligand binding affinity prediction. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["ligand-affinity-meta-model-2024"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b4-021"}],"facts":[{"label":"Model type","value":"Molecular docking procedure; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["ligand-affinity-meta-model-2024"],"source_locator":"Abstract (paragraph 1); Results/Meta-models of Docking and DL Tools (paragraph 1)"},{"label":"Architecture / procedure","value":"Force-field-derived empirical docking scores and sequence-based deep-learning estimates feed a learned meta-model, optionally with physicochemical or molecular descriptors.","status":"source_checked","source_ids":["ligand-affinity-meta-model-2024"],"source_locator":"Abstract (paragraph 1); Results/Meta-models of Docking and DL Tools (paragraph 1)"},{"label":"Biological inputs","value":"Protein/ligand representations and predictions of component affinity models","status":"source_checked","source_ids":["ligand-affinity-meta-model-2024"],"source_locator":"Methods/Dataset Selection/BindingDB (paragraph 1); Results/Meta-models of Docking and DL Tools (paragraph 7)"},{"label":"Outputs","value":"Predicted binding affinity","status":"source_checked","source_ids":["ligand-affinity-meta-model-2024"],"source_locator":"Results/Docking Tools (paragraph 4); Methods/GeneralSet Benchmark (paragraph 1)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["ligand-affinity-meta-model-2024","evidence-reported-ligand-affinity-meta-model-2024-readme-md"],"source_locator":"Methods/Dataset Selection/BindingDB; Methods/Dataset Selection/PDBbind; Methods/Molecular Docking/Implementation of Ligand Docking; Methods/Molecular Docking/Processing of Docking Scores; Methods/Deep Learning/DeepPurpose Library; Methods/Deep Learning/Training of 12 DL Models from the DeepPurpose Library Using\nBindingDB; Methods/Deep Learning/Training of DL Models Using PDBbind; Methods/Deep Learning/Fine-Tuning of the BDB-Trained Models Using PDBbind; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"binding-affinity meta-model is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["ligand-affinity-meta-model-2024"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"BindingDB version 2020m2 supplies 66,444 filtered ligand–protein complexes for deep-learning model training; the paper explores multiple training databases and model combinations.","status":"source_checked","source_ids":["ligand-affinity-meta-model-2024"],"source_locator":"Methods/Dataset Selection/BindingDB (paragraph 1); Methods/Deep Learning/DeepPurpose Library (paragraph 1)"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["ligand-affinity-meta-model-2024","evidence-reported-ligand-affinity-meta-model-2024-readme-md"],"source_locator":"Methods/Dataset Selection/BindingDB; Methods/Dataset Selection/PDBbind; Methods/Molecular Docking/Implementation of Ligand Docking; Methods/Molecular Docking/Processing of Docking Scores; Methods/Deep Learning/DeepPurpose Library; Methods/Deep Learning/Training of 12 DL Models from the DeepPurpose Library Using\nBindingDB; Methods/Deep Learning/Training of DL Models Using PDBbind; Methods/Deep Learning/Fine-Tuning of the BDB-Trained Models Using PDBbind; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"The authors provide code and pretrained models at https://github.com/Lee1701/Lee2023a, explicitly describing the release as partial because of a pending patent.","status":"source_checked","source_ids":["ligand-affinity-meta-model-2024"],"source_locator":"Data and Software Availability"},{"label":"Code licence","value":"GNU GPL version 3 (study repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-ligand-affinity-meta-model-2024-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-ligand-affinity-meta-model-2024-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Tests whether combining models reduces method-specific prediction errors.","source_ids":["ligand-affinity-meta-model-2024"],"source_locator":"Results/Prediction Synergy and Complementarity of Docking and DL Tools (paragraph 1); Introduction (paragraph 2)"}],"limitations":[{"text":"The exact base-model combination and training database define each meta-model; a generic ensemble name does not identify a reproducible checkpoint.","source_ids":["ligand-affinity-meta-model-2024"],"source_locator":"Discussion (paragraph 2); Results/Prediction Synergy and Complementarity of Docking and DL Tools (paragraph 2)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Protein/ligand representations and predictions of component affinity models","binding-affinity meta-model","Predicted binding affinity"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["ligand-affinity-meta-model-2024"],"source_locator":"Abstract (paragraph 1); Results/Meta-models of Docking and DL Tools (paragraph 1)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","binding-affinity meta-model is the comparison-table label; that label does not specify an immutable weight revision.","A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-df4084611520b7","kind":"model","name":"MetaPhlAn3","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["microbes-communities"]},"source_ids":["nabas-plus-2025"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"MetaPhlAn3","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"MetaPhlAn3 is the marker-based metagenomic comparator in the NABAS+ study.","summary_source_ids":["nabas-plus-2025"],"summary_source_locator":"Results/Creating a novel alignment-based metagenomic tool to improve classification accuracy and reduce false positives (paragraph 4); Introduction (paragraph 4)","sections":[{"title":"How the evaluated method works","body":"Taxonomic profiling uses clade-specific marker information to estimate microbial composition from shotgun reads.","source_ids":["nabas-plus-2025"],"source_locator":"Materials and methods/Datasets and reference databases/Consent for sample collection (paragraph 1); Results/Creating a novel alignment-based metagenomic tool to improve classification accuracy and reduce false positives (paragraph 6)"},{"title":"Underlying method and version boundaries","body":"MetaPhlAn profiles taxonomic composition using a marker reference. The repository’s current release is not the historical MetaPhlAn3 configuration; software and database versions require separate identification.","source_ids":["evidence-reported-base-metaphlan-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies MetaPhlAn3: Metagenomic taxonomic classification. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["nabas-plus-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-040"}],"facts":[{"label":"Model type","value":"Marker-based taxonomic profiler; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-metaphlan-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"Taxonomic profiling uses clade-specific marker information to estimate microbial composition from shotgun reads.","status":"source_checked","source_ids":["nabas-plus-2025"],"source_locator":"Materials and methods/Datasets and reference databases/Consent for sample collection (paragraph 1); Results/Creating a novel alignment-based metagenomic tool to improve classification accuracy and reduce false positives (paragraph 6)"},{"label":"Biological inputs","value":"Shotgun metagenomic reads and a versioned marker database","status":"source_checked","source_ids":["nabas-plus-2025"],"source_locator":"Discussion (paragraph 6); Introduction (paragraph 3)"},{"label":"Outputs","value":"Taxonomic abundance profiles","status":"source_checked","source_ids":["nabas-plus-2025"],"source_locator":"Results/Creating a novel alignment-based metagenomic tool to improve classification accuracy and reduce false positives (paragraph 8); Materials and methods/Datasets and reference databases/Recreating the CAMI sample19 (paragraph 1)"},{"label":"Parameters","value":"Not applicable: taxonomic marker profiling does not have a pretrained neural parameter total.","status":"inapplicable","source_ids":["nabas-plus-2025"],"source_locator":"Materials and methods/Selecting and setting up reference classifiers for NABAS+ benchmarking (paragraph 2); Materials and methods/Running the classifiers (paragraph 2)"},{"label":"Known versions / configuration","value":"MetaPhlAn3, as named in the paper; no exact package/database hash is supplied by the catalogue row.","status":"unreported","source_ids":["nabas-plus-2025"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"Reference-marker classification without task-specific neural fitting. CAMI comparisons use mpa_v31_CHOCOPhlAn_201901; a newer Zymo comparison uses mpa_vOct22_CHOCOPhlAnSGB_202212. These database conditions are not pooled.","status":"source_checked","source_ids":["nabas-plus-2025"],"source_locator":"Materials and methods / Datasets and reference databases / CAMI samples and Zymo community standards, database lists"},{"label":"Context limits","value":"Not applicable to a pretrained sequence-token window; read/contig lengths and the reference-database or comparison configuration determine the analysed input.","status":"inapplicable","source_ids":["nabas-plus-2025"],"source_locator":"Materials and methods/Datasets and reference databases/CAMI samples (paragraph 3); Discussion (paragraph 6)"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://github.com/biobakery/MetaPhlAn/blob/424f3e6e30618266404353e1083c6405a9f02f48/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-metaphlan-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"Repository-specific licence text is available; a standard SPDX label has not been established. (upstream repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-base-metaphlan-license-txt"],"source_locator":"license.txt; complete licence text"},{"label":"Weights licence","value":"Not applicable to neural weights; the software and reference database have separate reuse terms.","status":"inapplicable","source_ids":["nabas-plus-2025"],"source_locator":"Materials and methods/Datasets and reference databases/CAMI samples (paragraph 3); Materials and methods/Datasets and reference databases/CAMI samples (paragraph 2)"}],"strengths":[{"text":"Provides a marker-based reference against whole-genome alignment and k-mer classification.","source_ids":["nabas-plus-2025"],"source_locator":"Materials and methods/Selecting and setting up reference classifiers for NABAS+ benchmarking (paragraph 1); Discussion (paragraph 3)"}],"limitations":[{"text":"The current MetaPhlAn release is not the version in this row; database and parameter choices are required for reproduction.","source_ids":["nabas-plus-2025"],"source_locator":"Materials and methods/Datasets and reference databases/Zymo community standards (paragraph 3); Materials and methods/Statistical comparison and visualization (paragraph 10)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Shotgun metagenomic reads and a versioned marker database","MetaPhlAn3","Taxonomic abundance profiles"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["nabas-plus-2025"],"source_locator":"Materials and methods/Datasets and reference databases/Consent for sample collection (paragraph 1); Results/Creating a novel alignment-based metagenomic tool to improve classification accuracy and reduce false positives (paragraph 6)"},"coverage":"limited","gaps":["MetaPhlAn3, as named in the paper; no exact package/database hash is supplied by the catalogue row.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-e0443048c6e110","kind":"model","name":"ProteinMPNN","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["proteins-complexes"]},"source_ids":["proteingym-2023"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"ProteinMPNN","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"This protein model is evaluated as a zero-shot mutation-effect scorer in ProteinGym.","summary_source_ids":["proteingym-2023"],"summary_source_locator":"ProteinGym benchmarks (paragraph 1); Introduction (paragraph 4)","sections":[{"title":"How the evaluated method works","body":"ProteinMPNN scores sequences conditional on an input protein backbone. ProteinGym treats it as an inverse-folding baseline, distinct from sequence-only language models.","source_ids":["proteingym-2023"],"source_locator":"Appendix/A Appendix/A.4 Baselines/A.4.1 Zero-shot baselines/Inverse Folding models (paragraph 1); Appendix/A Appendix/A.4 Baselines/A.4.1 Zero-shot baselines/Protein language models (paragraph 1)"},{"title":"Underlying method and version boundaries","body":"ProteinMPNN exposes full-backbone and Cα-only models with separate weights. It can score a sequence conditional on structure or generate sequences; these modes and backbone variants must remain explicit.","source_ids":["evidence-reported-base-proteinmpnn-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies ProteinMPNN: Zero-shot substitution mutation effects: stability. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["proteingym-2023"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-018"}],"facts":[{"label":"Model type","value":"Structure-conditioned sequence model; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-proteinmpnn-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"ProteinMPNN scores sequences conditional on an input protein backbone. ProteinGym treats it as an inverse-folding baseline, distinct from sequence-only language models.","status":"source_checked","source_ids":["proteingym-2023"],"source_locator":"Appendix/A Appendix/A.4 Baselines/A.4.1 Zero-shot baselines/Inverse Folding models (paragraph 1); Appendix/A Appendix/A.4 Baselines/A.4.1 Zero-shot baselines/Protein language models (paragraph 1)"},{"label":"Biological inputs","value":"Variant sequences and protein backbone structure","status":"source_checked","source_ids":["proteingym-2023"],"source_locator":"ProteinGym benchmarks/Mutation types (paragraph 1); ProteinGym benchmarks/Mutation types/Indels (paragraph 1)"},{"label":"Outputs","value":"Mutation-effect scores and ranked variants","status":"source_checked","source_ids":["proteingym-2023"],"source_locator":"Appendix/A Appendix/A.4 Baselines/A.4.3 Clinical baselines (paragraph 1); Appendix/A Appendix/A.3 Datasets/A.3.2 Clinical datasets (paragraph 1)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["proteingym-2023","evidence-reported-base-proteinmpnn-readme-md"],"source_locator":"ProteinGym benchmarks/Model training regime; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"ProteinMPNN is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["proteingym-2023"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"PDB structure training, as identified in ProteinGym’s inverse-folding baseline description.","status":"source_checked","source_ids":["proteingym-2023"],"source_locator":"Appendix/A Appendix/A.3 Datasets/A.3.1 DMS assays/Selection and processing (paragraph 1); Appendix/A Appendix/A.3 Datasets/A.3.3 Access (paragraph 3)"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["proteingym-2023","evidence-reported-base-proteinmpnn-readme-md"],"source_locator":"ProteinGym benchmarks/Model training regime; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://github.com/dauparas/ProteinMPNN/blob/8907e6671bfbfc92303b5f79c4b5e6ce47cdef57/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-proteinmpnn-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"MIT (upstream repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-base-proteinmpnn-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-base-proteinmpnn-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Uses standardised evaluation across diverse experimental assays.","source_ids":["proteingym-2023"],"source_locator":"Introduction (paragraph 2); Abstract (paragraph 1)"}],"limitations":[{"text":"Different models receive different evolutionary/structural information; assay fitness, organismal fitness and clinical pathogenicity are not interchangeable endpoints.","source_ids":["proteingym-2023"],"source_locator":"ProteinGym benchmarks (paragraph 1); Appendix/A Appendix/A.3 Datasets/A.3.1 DMS assays/Processing of large thermostability dataset (paragraph 2)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Variant sequences and protein backbone structure","ProteinMPNN","Mutation-effect scores and ranked variants"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["proteingym-2023"],"source_locator":"Appendix/A Appendix/A.4 Baselines/A.4.1 Zero-shot baselines/Inverse Folding models (paragraph 1); Appendix/A Appendix/A.4 Baselines/A.4.1 Zero-shot baselines/Protein language models (paragraph 1)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","ProteinMPNN is the comparison-table label; that label does not specify an immutable weight revision.","A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-e2f2f0d4830bb0","kind":"model","name":"Mouse-Geneformer","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["cells-tissues"]},"source_ids":["mouse-geneformer-2025"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"Mouse-Geneformer","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Mouse-Geneformer adapts the Geneformer transformer design to mouse single-cell transcriptomes.","summary_source_ids":["mouse-geneformer-2025"],"summary_source_locator":"Results/In silico perturbation experiments/In silico perturbation experiment in COP1 KO microglia. (paragraph 2); Results/Cross-species application of mouse-Geneformer through orthologous gene name conversion/Human cell type classification using mouse-Geneformer. (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"Rank Value Encoding orders genes by expression-derived values for a transformer encoder. The mouse model follows the human Geneformer design with minor modifications and mouse-specific pretraining.","source_ids":["mouse-geneformer-2025"],"source_locator":"Materials and methods/Construction of the mouse-Geneformer/Pretraining of the mouse-Geneformer. (paragraph 1); Materials and methods/Construction of the mouse-Geneformer/Architecture. (paragraph 1)"},{"title":"What was evaluated","body":"The linked evaluation record identifies Mouse-Geneformer: Human thymus cell-type classification. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["mouse-geneformer-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b3-003"}],"facts":[{"label":"Model type","value":"Transformer representation pipeline; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["mouse-geneformer-2025"],"source_locator":"Materials and methods/Construction of the mouse-Geneformer/Pretraining of the mouse-Geneformer. (paragraph 1); Materials and methods/Construction of the mouse-Geneformer/Architecture. (paragraph 1)"},{"label":"Architecture / procedure","value":"Rank Value Encoding orders genes by expression-derived values for a transformer encoder. The mouse model follows the human Geneformer design with minor modifications and mouse-specific pretraining.","status":"source_checked","source_ids":["mouse-geneformer-2025"],"source_locator":"Materials and methods/Construction of the mouse-Geneformer/Pretraining of the mouse-Geneformer. (paragraph 1); Materials and methods/Construction of the mouse-Geneformer/Architecture. (paragraph 1)"},{"label":"Biological inputs","value":"Rank-encoded mouse single-cell transcriptomes","status":"source_checked","source_ids":["mouse-geneformer-2025"],"source_locator":"Results/In silico perturbation experiments/In silico perturbation experiment in COP1 KO microglia. (paragraph 2); Results/Cross-species application of mouse-Geneformer through orthologous gene name conversion/Human cell type classification using mouse-Geneformer. (paragraph 1)"},{"label":"Outputs","value":"Gene/cell representations and adapted cell-type or perturbation analyses","status":"source_checked","source_ids":["mouse-geneformer-2025"],"source_locator":"Introduction (paragraph 4); Results/Cross-species application of mouse-Geneformer through orthologous gene name conversion/Human cell type classification using mouse-Geneformer. (paragraph 1)"},{"label":"Parameters","value":"Six transformer blocks, four heads and 256-dimensional embeddings are specified; a total parameter count is not reported in Table 1.","status":"unreported","source_ids":["mouse-geneformer-2025"],"source_locator":"Table1 Summary of Experimental conditions"},{"label":"Known versions / configuration","value":"Mouse-Geneformer is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["mouse-geneformer-2025"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"Approximately 21 million mouse single-cell RNA-seq profiles; the human comparator was pretrained separately on human cells.","status":"source_checked","source_ids":["mouse-geneformer-2025"],"source_locator":"Results/Development of mouse-Geneformer (paragraph 1); Results/Cell type classification: pretraining is effective (paragraph 1)"},{"label":"Context limits","value":"2,048 gene tokens (Table 1); six transformer blocks, four attention heads and 256-dimensional embeddings.","status":"source_checked","source_ids":["mouse-geneformer-2025"],"source_locator":"Table1; Input max dimensions row"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/machine-perception-robotics-group/Mouse-Geneformer/blob/ed17d455193ed4c4d93230a8f5f2de349cf20c81/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-mouse-geneformer-2025-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","status":"unreported","source_ids":["evidence-reported-mouse-geneformer-2025-readme-md"],"source_locator":"README.md and repository-root licence-file search"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-mouse-geneformer-2025-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Tests species-specific pretraining instead of assuming a human representation transfers unchanged to mouse.","source_ids":["mouse-geneformer-2025"],"source_locator":"Results/Cross-species application of mouse-Geneformer through orthologous gene name conversion (paragraph 2); Discussion (paragraph 1)"}],"limitations":[{"text":"In-silico perturbation scores are model predictions; validation of selected genes does not validate every predicted regulatory relationship.","source_ids":["mouse-geneformer-2025"],"source_locator":"Results/Cross-species application of mouse-Geneformer through orthologous gene name conversion/In silico perturbation of human data using mouse-Geneformer 2: COVID-19 human blood. (paragraph 1); Results/Cross-species application of mouse-Geneformer through orthologous gene name conversion (paragraph 2)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Rank-encoded mouse single-cell transcriptomes","Mouse-Geneformer","Gene/cell representations and adapted cell-type or perturbation analyses"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["mouse-geneformer-2025"],"source_locator":"Materials and methods/Construction of the mouse-Geneformer/Pretraining of the mouse-Geneformer. (paragraph 1); Materials and methods/Construction of the mouse-Geneformer/Architecture. (paragraph 1)"},"coverage":"limited","gaps":["Six transformer blocks, four heads and 256-dimensional embeddings are specified; a total parameter count is not reported in Table 1.","Mouse-Geneformer is the comparison-table label; that label does not specify an immutable weight revision.","No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-e3abb0b9a2ec79","kind":"model","name":"UFold","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["rna-transcriptomes"]},"source_ids":["tu-fold-2025"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"UFold","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"UFold is a learned RNA-secondary-structure comparator in the TU-Fold study.","summary_source_ids":["tu-fold-2025"],"summary_source_locator":"Materials and methods/Training method (paragraph 3); Results/TU-fold outperforms existing methods without using post-processing steps (paragraph 2)","sections":[{"title":"How the evaluated method works","body":"RNA sequences become an image-like pairwise representation processed by a fully convolutional network, with base-pairing rules supporting structure prediction.","source_ids":["tu-fold-2025"],"source_locator":"Materials and methods/Training method (paragraph 1); Materials and methods/Training method (paragraph 3)"},{"title":"Underlying method and version boundaries","body":"UFold converts RNA sequences into an image-like representation and uses a fully convolutional network, trained on annotated structures and pairing rules, to predict secondary structure.","source_ids":["evidence-reported-base-ufold-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies UFold: RNA secondary structure. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["tu-fold-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-014"}],"facts":[{"label":"Model type","value":"Convolutional RNA folding model; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-ufold-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"RNA sequences become an image-like pairwise representation processed by a fully convolutional network, with base-pairing rules supporting structure prediction.","status":"source_checked","source_ids":["tu-fold-2025"],"source_locator":"Materials and methods/Training method (paragraph 1); Materials and methods/Training method (paragraph 3)"},{"label":"Biological inputs","value":"RNA nucleotide sequences","status":"source_checked","source_ids":["tu-fold-2025"],"source_locator":"CRediT authorship contribution statement (paragraph 1); Discussion (paragraph 3)"},{"label":"Outputs","value":"RNA secondary-structure base pairs","status":"source_checked","source_ids":["tu-fold-2025"],"source_locator":"Materials and methods/Training method (paragraph 3); Materials and methods/Data augmentation (paragraph 2)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["tu-fold-2025","evidence-reported-base-ufold-readme-md"],"source_locator":"Materials and methods/Training method; Materials and methods/Evaluation method; Materials and methods/Data augmentation; Materials and methods/Knowledge merge; Materials and methods/Experimental settings; Results/TU-fold outperforms existing methods without using post-processing steps; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"UFold is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["tu-fold-2025"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"The comparison evaluates UFold on the RNA8F collection under three random train/validation/test partitions. RNA8F combines RNAStrAlign, ArchiveII, bpRNA and CRW2, with CD-HIT filtering at 0.9. Exact UFold optimiser/checkpoint details are not stated in the comparison passage.","status":"source_checked","source_ids":["tu-fold-2025"],"source_locator":"Materials and methods / Experimental settings; Table 2"},{"label":"Context limits","value":"The evaluated RNA8F collection contains sequences of at most 500 nucleotides. This is a dataset restriction, not a universal UFold inference limit.","status":"source_checked","source_ids":["tu-fold-2025"],"source_locator":"Materials and methods / Experimental settings"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://github.com/uci-cbcl/UFold/blob/75bd9acc83826059682dfca9d3659df66b132cd1/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-ufold-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"MIT (upstream repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-base-ufold-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-base-ufold-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Provides an established convolutional comparator for the new multiclass formulation.","source_ids":["tu-fold-2025"],"source_locator":"Results/TU-fold outperforms existing methods without using post-processing steps (paragraph 5); Results/TU-fold outperforms existing methods without using post-processing steps (paragraph 7)"}],"limitations":[{"text":"The comparator has its own training and post-processing conventions; TU-Fold augmentation must not be assigned to UFold.","source_ids":["tu-fold-2025"],"source_locator":"Introduction (paragraph 7); Introduction (paragraph 6)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["RNA nucleotide sequences","UFold","RNA secondary-structure base pairs"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["tu-fold-2025"],"source_locator":"Materials and methods/Training method (paragraph 1); Materials and methods/Training method (paragraph 3)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","UFold is the comparison-table label; that label does not specify an immutable weight revision.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-e4710b1c3facf2","kind":"model","name":"ESM-2 embedding + paper classifier","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["proteins-complexes"]},"source_ids":["clathrin-plm-2025"],"links":[],"attributes":{"entity_level":"method","version":"not stated in table","reported_name":"ESM-2 embedding + paper classifier","historical_missing_metadata":{"checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"This record tracks the ESM-2-feature configuration in a supervised clathrin-classification study.","summary_source_ids":["clathrin-plm-2025"],"summary_source_locator":"Materials and methods/Feature selection method (paragraph 1); Introduction (paragraph 3)","sections":[{"title":"How the evaluated method works","body":"Pretrained protein embeddings are used as features for a downstream classifier. The paper also studies fused multi-model features and an LSTM; those combined PLM-CLA results must remain distinct from this ESM-2-only feature row.","source_ids":["clathrin-plm-2025"],"source_locator":"Materials and methods/Overall framework of PLM-CLA (paragraph 1); Results and discussions/The effect of feature selection methods on the predictive performance (paragraph 1)"},{"title":"Underlying method and version boundaries","body":"ESM-2 is a transformer protein language-model family. The official repository exposes residue embeddings, sequence-level pooling and models at several sizes; the study configuration determines which of these is evaluated.","source_ids":["evidence-reported-base-esm-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies ESM-2 embedding + paper classifier: clathrin protein classification. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["clathrin-plm-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-b2-clathrin-plm-2025"}],"facts":[{"label":"Model type","value":"Protein sequence transformer; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-esm-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"Pretrained protein embeddings are used as features for a downstream classifier. The paper also studies fused multi-model features and an LSTM; those combined PLM-CLA results must remain distinct from this ESM-2-only feature row.","status":"source_checked","source_ids":["clathrin-plm-2025"],"source_locator":"Materials and methods/Overall framework of PLM-CLA (paragraph 1); Results and discussions/The effect of feature selection methods on the predictive performance (paragraph 1)"},{"label":"Biological inputs","value":"Protein amino-acid sequences","status":"source_checked","source_ids":["clathrin-plm-2025"],"source_locator":"Introduction (paragraph 2); Conclusion (paragraph 1)"},{"label":"Outputs","value":"Clathrin versus non-clathrin classification","status":"source_checked","source_ids":["clathrin-plm-2025"],"source_locator":"Results and discussions/Performance evaluation of individual feature embeddings (paragraph 1); Results and discussions/The effect of feature selection methods on the predictive performance (paragraph 1)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["clathrin-plm-2025","evidence-reported-base-esm-readme-md"],"source_locator":"Materials and methods/Dataset construction; Materials and methods/Protein language model; Materials and methods/Feature selection method; Materials and methods/Overall framework of PLM-CLA; Materials and methods/Performance evaluation; Results and discussions/The effect of feature selection methods on the predictive performance; Results and discussions/Comparison of PLM-CLA with existing methods; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"ESM-2 embedding + paper classifier is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["clathrin-plm-2025"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"Le2019 and Zhang2020 clathrin datasets; Zhang2020 applies a 0.7 BLAST redundancy threshold and excludes sequences shorter than 220 residues.","status":"source_checked","source_ids":["clathrin-plm-2025"],"source_locator":"Materials and methods/Dataset construction (paragraph 1); Results and discussions/Performance evaluation of PLM-CLA on other benchmark independent test datasets (paragraph 1)"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["clathrin-plm-2025","evidence-reported-base-esm-readme-md"],"source_locator":"Materials and methods/Dataset construction; Materials and methods/Protein language model; Materials and methods/Feature selection method; Materials and methods/Overall framework of PLM-CLA; Materials and methods/Performance evaluation; Results and discussions/The effect of feature selection methods on the predictive performance; Results and discussions/Comparison of PLM-CLA with existing methods; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://github.com/facebookresearch/esm/blob/2b369911bb5b4b0dda914521b9475cad1656b2ac/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-esm-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"MIT (upstream repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-base-esm-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-base-esm-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Allows a controlled comparison of learned protein representations within a fixed classification study.","source_ids":["clathrin-plm-2025"],"source_locator":"Introduction (paragraph 3); Materials and methods/Protein language model (paragraph 1)"}],"limitations":[{"text":"The record identifies a feature-extractor configuration, not an independently validated clathrin capability of the foundation model.","source_ids":["clathrin-plm-2025"],"source_locator":"Results and discussions/Performance evaluation of individual feature embeddings (paragraph 1); Results and discussions/The effect of feature selection methods on the predictive performance (paragraph 1)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Protein amino-acid sequences","ESM-2 embedding + paper classifier","Clathrin versus non-clathrin classification"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["clathrin-plm-2025"],"source_locator":"Materials and methods/Overall framework of PLM-CLA (paragraph 1); Results and discussions/The effect of feature selection methods on the predictive performance (paragraph 1)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","ESM-2 embedding + paper classifier is the comparison-table label; that label does not specify an immutable weight revision.","A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-e6ba198c2ac996","kind":"model","name":"MDL4Microbiome","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["microbes-communities"]},"source_ids":["mdl4microbiome-2022"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"MDL4Microbiome","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"MDL4Microbiome predicts study-specific phenotype labels from several molecular summaries of metagenomic samples.","summary_source_ids":["mdl4microbiome-2022"],"summary_source_locator":"Introduction (paragraph 2); Methods/Generation of feature sets (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"A multimodal neural classifier combines conventional taxonomic profiles, genome-level relative abundance and metabolic-functional features.","source_ids":["mdl4microbiome-2022"],"source_locator":"Methods/Generation of feature sets (paragraph 1); Results/Performance evaluation with various model architectures and parameters (paragraph 2)"},{"title":"What was evaluated","body":"The linked evaluation record identifies MDL4Microbiome: microbiome disease-state classification. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["mdl4microbiome-2022"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b4-020"}],"facts":[{"label":"Model type","value":"Study-specific predictive method; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["mdl4microbiome-2022"],"source_locator":"Methods/Generation of feature sets (paragraph 1); Results/Performance evaluation with various model architectures and parameters (paragraph 2)"},{"label":"Architecture / procedure","value":"A multimodal neural classifier combines conventional taxonomic profiles, genome-level relative abundance and metabolic-functional features.","status":"source_checked","source_ids":["mdl4microbiome-2022"],"source_locator":"Methods/Generation of feature sets (paragraph 1); Results/Performance evaluation with various model architectures and parameters (paragraph 2)"},{"label":"Biological inputs","value":"Taxonomic, genome-abundance and metabolic-functional features derived from metagenomes","status":"source_checked","source_ids":["mdl4microbiome-2022"],"source_locator":"Conclusion (paragraph 1); Results/Performance evaluation with various model architectures and parameters (paragraph 2)"},{"label":"Outputs","value":"Study phenotype classifications","status":"source_checked","source_ids":["mdl4microbiome-2022"],"source_locator":"Supplementary Information (paragraph 1); Conclusion (paragraph 2)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["mdl4microbiome-2022","evidence-reported-mdl4microbiome-2022-readme-md"],"source_locator":"Methods/Data preparation and preprocessing; Methods/Generation of feature sets; Methods/Construction of multimodal deep learning model; Methods/Performance evaluation; Results/Performance evaluation with various model architectures and parameters; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"MDL4Microbiome is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["mdl4microbiome-2022"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"Disease-associated microbiome cohorts described in the paper, evaluated with leave-one-out cross-validation.","status":"source_checked","source_ids":["mdl4microbiome-2022"],"source_locator":"Methods/Generation of feature sets (paragraph 1); Introduction (paragraph 3)"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["mdl4microbiome-2022","evidence-reported-mdl4microbiome-2022-readme-md"],"source_locator":"Methods/Data preparation and preprocessing; Methods/Generation of feature sets; Methods/Construction of multimodal deep learning model; Methods/Performance evaluation; Results/Performance evaluation with various model architectures and parameters; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/DMnBI/MDL4Microbiome/blob/0b2076cbd31af62224bd7a90ef0130e4cac50017/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-mdl4microbiome-2022-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","status":"unreported","source_ids":["evidence-reported-mdl4microbiome-2022-readme-md"],"source_locator":"README.md and repository-root licence-file search"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-mdl4microbiome-2022-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Combines complementary composition and functional measurements rather than relying on a single abundance table.","source_ids":["mdl4microbiome-2022"],"source_locator":"Conclusion (paragraph 1); Results/Performance evaluation with various model architectures and parameters (paragraph 1)"}],"limitations":[{"text":"Retrospective disease-cohort classification is not prospective clinical validation; population, assay and batch effects can affect transfer.","source_ids":["mdl4microbiome-2022"],"source_locator":"Methods/Performance evaluation (paragraph 1); Introduction (paragraph 4)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Taxonomic, genome-abundance and metabolic-functional features derived from metagenomes","MDL4Microbiome","Study phenotype classifications"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["mdl4microbiome-2022"],"source_locator":"Methods/Generation of feature sets (paragraph 1); Results/Performance evaluation with various model architectures and parameters (paragraph 2)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","MDL4Microbiome is the comparison-table label; that label does not specify an immutable weight revision.","A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-e78e3886df0d3a","kind":"model","name":"VirSorter","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["microbes-communities"]},"source_ids":["viral-contig-simulation-2021"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"VirSorter","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"This viral-contig detector is compared on simulated metagenomes with varying taxonomic composition and complexity.","summary_source_ids":["viral-contig-simulation-2021"],"summary_source_locator":"Conclusion (paragraph 1); Methods/Building simulated Illumina metagenomes (paragraph 3)","sections":[{"title":"How the evaluated method works","body":"The benchmark runs existing viral-identification tools on assembled contigs and evaluates both standalone viruses and integrated prophages.","source_ids":["viral-contig-simulation-2021"],"source_locator":"Methods/Tools used in simulation study (paragraph 1); Methods/Tool performance scoring (paragraph 1)"},{"title":"What was evaluated","body":"The linked evaluation record identifies VirSorter: Simulated prophage-contig detection. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["viral-contig-simulation-2021"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b3-026"}],"facts":[{"label":"Model type","value":"Study-specific predictive method; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["viral-contig-simulation-2021"],"source_locator":"Methods/Tools used in simulation study (paragraph 1); Methods/Tool performance scoring (paragraph 1)"},{"label":"Architecture / procedure","value":"The benchmark runs existing viral-identification tools on assembled contigs and evaluates both standalone viruses and integrated prophages.","status":"source_checked","source_ids":["viral-contig-simulation-2021"],"source_locator":"Methods/Tools used in simulation study (paragraph 1); Methods/Tool performance scoring (paragraph 1)"},{"label":"Biological inputs","value":"Metagenomic contigs","status":"source_checked","source_ids":["viral-contig-simulation-2021"],"source_locator":"Methods/Building simulated Illumina metagenomes (paragraph 3); Results/Tool performance by contig length (paragraph 1)"},{"label":"Outputs","value":"Viral-contig or prophage predictions","status":"source_checked","source_ids":["viral-contig-simulation-2021"],"source_locator":"Results/Prophage identification performance (paragraph 1); Discussion (paragraph 1)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["viral-contig-simulation-2021","evidence-reported-viral-contig-simulation-2021-readme-md"],"source_locator":"Methods/Phybrid, a hybrid gene content and nucleotide feature set for viral classification; Methods/Gene content feature set creation; Methods/Model and hyperparameter selection; Methods/Building simulated Illumina metagenomes; Methods/Integrated prophage identification; Methods/Tools used in simulation study; Methods/Tool performance scoring; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"VirSorter is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["viral-contig-simulation-2021"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"The study evaluates released software on simulated assembled contigs, rather than fitting this comparator on the study’s labels. The exact upstream training/reference inventory is not restated.","status":"source_checked","source_ids":["viral-contig-simulation-2021"],"source_locator":"Methods / Tools used in simulation study"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["viral-contig-simulation-2021","evidence-reported-viral-contig-simulation-2021-readme-md"],"source_locator":"Methods/Phybrid, a hybrid gene content and nucleotide feature set for viral classification; Methods/Gene content feature set creation; Methods/Model and hyperparameter selection; Methods/Building simulated Illumina metagenomes; Methods/Integrated prophage identification; Methods/Tools used in simulation study; Methods/Tool performance scoring; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/Strong-Lab/Viral_Classification_in_Metagenomics/blob/f583cbff6b022ce3a7e3870003e22e14769566fa/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-viral-contig-simulation-2021-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"MIT (study repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-viral-contig-simulation-2021-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-viral-contig-simulation-2021-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Uses several simulation conditions and separately examines integrated viral sequences.","source_ids":["viral-contig-simulation-2021"],"source_locator":"Discussion (paragraph 4); Results/Overall tool performance (paragraph 1)"}],"limitations":[{"text":"Sensitivity and false-positive behaviour depend on contig length, community composition and reference coverage.","source_ids":["viral-contig-simulation-2021"],"source_locator":"Methods/Building simulated Illumina metagenomes (paragraph 4); Results/Tool performance by contig length (paragraph 1)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Metagenomic contigs","VirSorter","Viral-contig or prophage predictions"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["viral-contig-simulation-2021"],"source_locator":"Methods/Tools used in simulation study (paragraph 1); Methods/Tool performance scoring (paragraph 1)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","VirSorter is the comparison-table label; that label does not specify an immutable weight revision.","A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-e7d203bd99ca99","kind":"model","name":"NABAS+","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["microbes-communities"]},"source_ids":["nabas-plus-2025"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"NABAS+","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"NABAS+ performs species-level profiling of Illumina shotgun reads through reference alignment and stringent filtering.","summary_source_ids":["nabas-plus-2025"],"summary_source_locator":"Introduction (paragraph 3); Materials and methods/Statistical comparison and visualization (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"BWA aligns reads to a curated RefSeq database with one selected genome per species. Quality filters retain genomes supported by reliable mapped reads.","source_ids":["nabas-plus-2025"],"source_locator":"Materials and methods/Datasets and reference databases/CAMI samples (paragraph 3); Materials and methods/Running the classifiers (paragraph 2)"},{"title":"What was evaluated","body":"The linked evaluation record identifies NABAS+: Metagenomic taxonomic classification. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["nabas-plus-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-039"}],"facts":[{"label":"Model type","value":"Study-specific predictive method; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["nabas-plus-2025"],"source_locator":"Materials and methods/Datasets and reference databases/CAMI samples (paragraph 3); Materials and methods/Running the classifiers (paragraph 2)"},{"label":"Architecture / procedure","value":"BWA aligns reads to a curated RefSeq database with one selected genome per species. Quality filters retain genomes supported by reliable mapped reads.","status":"source_checked","source_ids":["nabas-plus-2025"],"source_locator":"Materials and methods/Datasets and reference databases/CAMI samples (paragraph 3); Materials and methods/Running the classifiers (paragraph 2)"},{"label":"Biological inputs","value":"Illumina shotgun metagenomic reads and a curated species reference database","status":"source_checked","source_ids":["nabas-plus-2025"],"source_locator":"Introduction (paragraph 3); Discussion (paragraph 6)"},{"label":"Outputs","value":"Species identifications and microbiome profiles","status":"source_checked","source_ids":["nabas-plus-2025"],"source_locator":"Materials and methods/Statistical comparison and visualization (paragraph 7); Discussion (paragraph 5)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["nabas-plus-2025","evidence-reported-nabas-plus-2025-readme-md"],"source_locator":"Materials and methods/Selecting and setting up classifiers for the initial gut metagenome analysis; Materials and methods/Selecting and setting up reference classifiers for NABAS+ benchmarking; Materials and methods/Datasets and reference databases/Human gut microbiome samples: acquisition and sequencing; Materials and methods/Datasets and reference databases/Consent for sample collection; Materials and methods/Datasets and reference databases/Real-world clinical dataset; Materials and methods/Datasets and reference databases/CAMI samples; Materials and methods/Datasets and reference databases/Recreating the CAMI sample19; Materials and methods/Datasets and reference databases/Zymo community standards; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"NABAS+ is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["nabas-plus-2025"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"Reference-genome curation rather than learned neural weights.","status":"source_checked","source_ids":["nabas-plus-2025"],"source_locator":"Materials and methods/Datasets and reference databases/Recreating the CAMI sample19 (paragraph 1); Materials and methods/Datasets and reference databases/CAMI samples (paragraph 3)"},{"label":"Context limits","value":"Not applicable to a pretrained sequence-token window; read/contig lengths and the reference-database or comparison configuration determine the analysed input.","status":"inapplicable","source_ids":["nabas-plus-2025"],"source_locator":"Materials and methods/Datasets and reference databases/CAMI samples (paragraph 3); Discussion (paragraph 6)"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/TakacsBertalan/NABAS_paper_scripts/blob/7cab4d317a2c362988e7b96fb33f92a9c79a9fdc/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-nabas-plus-2025-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"CC BY-NC 4.0 (study repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-nabas-plus-2025-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-nabas-plus-2025-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Tests simulated gastrooral communities and deeply sequenced microbial standards.","source_ids":["nabas-plus-2025"],"source_locator":"Abstract (paragraph 1); Materials and methods/Datasets and reference databases/Zymo community standards (paragraph 1)"}],"limitations":[{"text":"Alignment and reference curation trade computational cost against precision; retrospective pathogen examples do not establish clinical diagnostic performance.","source_ids":["nabas-plus-2025"],"source_locator":"Results/Testing classifiers on the CAMI gastrooral in silico data (paragraph 2); Materials and methods/Statistical comparison and visualization (paragraph 2)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Illumina shotgun metagenomic reads and a curated species reference database","NABAS+","Species identifications and microbiome profiles"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["nabas-plus-2025"],"source_locator":"Materials and methods/Datasets and reference databases/CAMI samples (paragraph 3); Materials and methods/Running the classifiers (paragraph 2)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","NABAS+ is the comparison-table label; that label does not specify an immutable weight revision.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-eae60780097101","kind":"model","name":"Chai-1","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["molecular-interactions"]},"source_ids":["lipp-2026"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"Chai-1","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"This configuration predicts lipid–protein poses in the LiPP benchmark.","summary_source_ids":["lipp-2026"],"summary_source_locator":"Results/Curation of the LiPP Benchmark\nSet for Modeling Lipid–Protein\nInteractions (paragraph 1); Materials\nand Methods/Curation of\nthe LiPP Test Set, Precutoff Set, and Similarity\nAssessments between the Different LiPP Data Sets (paragraph 2)","sections":[{"title":"How the evaluated method works","body":"Chai-1 jointly predicts the lipid–protein complex from sequence and molecular identity under the paper’s default co-folding settings.","source_ids":["lipp-2026"],"source_locator":"Materials\nand Methods/Molecular Docking and Structure Prediction\nProtocols (paragraph 1); Materials\nand Methods/Protein Structure Errors (paragraph 1)"},{"title":"Underlying method and version boundaries","body":"Chai-1 predicts biomolecular structures through the released inference implementation. Its input options, MSA/template use and sampling budget are part of each evaluated configuration.","source_ids":["evidence-reported-base-chai-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies Chai-1: Lipid–protein binding pose. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["lipp-2026"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-041"}],"facts":[{"label":"Model type","value":"Biomolecular structure predictor; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-chai-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"Chai-1 jointly predicts the lipid–protein complex from sequence and molecular identity under the paper’s default co-folding settings.","status":"source_checked","source_ids":["lipp-2026"],"source_locator":"Materials\nand Methods/Molecular Docking and Structure Prediction\nProtocols (paragraph 1); Materials\nand Methods/Protein Structure Errors (paragraph 1)"},{"label":"Biological inputs","value":"Protein amino-acid sequence and lipid SMILES/CCD identity","status":"source_checked","source_ids":["lipp-2026"],"source_locator":"Materials\nand Methods/Molecular Docking and Structure Prediction\nProtocols (paragraph 1); Materials\nand Methods/Curation of the LiPP Benchmark Set (paragraph 1)"},{"label":"Outputs","value":"Lipid–protein complex or docking-pose coordinates","status":"source_checked","source_ids":["lipp-2026"],"source_locator":"Materials\nand Methods/Protein Structure Errors (paragraph 1); Materials\nand Methods/Molecular Docking and Structure Prediction\nProtocols (paragraph 1)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["lipp-2026","evidence-reported-base-chai-readme-md"],"source_locator":"Materials\nand Methods/Curation of the LiPP Benchmark Set; Materials\nand Methods/Curation of\nthe LiPP Test Set, Precutoff Set, and Similarity\nAssessments between the Different LiPP Data Sets; Materials\nand Methods/Protein Similarity; Materials\nand Methods/Lipid Similarity; Materials\nand Methods/Molecular Docking and Structure Prediction\nProtocols; Materials\nand Methods/AlphaFold 3; Materials\nand Methods/Chai-1; Materials\nand Methods/RoseTTAFold AA; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"chai_lab0.2.0; ESM embeddings enabled;200 denoising steps,3 trunk recycles, seed 42, five diffusion samples ranked by ipTM.","status":"source_checked","source_ids":["lipp-2026"],"source_locator":"Materials and Methods / Chai-1"},{"label":"Training data / fitting","value":"The LiPP authors identify a PDB training cutoff of 21 January 2021 in Table 1. They evaluate the released predictor without retraining; this is the study’s description, not a new audit of every upstream training sample.","status":"source_checked","source_ids":["lipp-2026"],"source_locator":"Table 1, Chai-1 row; Molecular Docking and Structure Prediction Protocols"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["lipp-2026","evidence-reported-base-chai-readme-md"],"source_locator":"Materials\nand Methods/Curation of the LiPP Benchmark Set; Materials\nand Methods/Curation of\nthe LiPP Test Set, Precutoff Set, and Similarity\nAssessments between the Different LiPP Data Sets; Materials\nand Methods/Protein Similarity; Materials\nand Methods/Lipid Similarity; Materials\nand Methods/Molecular Docking and Structure Prediction\nProtocols; Materials\nand Methods/AlphaFold 3; Materials\nand Methods/Chai-1; Materials\nand Methods/RoseTTAFold AA; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://github.com/chaidiscovery/chai-lab/blob/66c38d1fe5c6756a89ff8596b1dea87d305ec06f/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-chai-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"Apache 2.0 (upstream repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-base-chai-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"Apache 2.0 for both Chai-1 code and model weights, explicitly stated in the official README; paper-specific derived artifacts are separate.","status":"source_checked","source_ids":["evidence-reported-base-chai-readme-md"],"source_locator":"README.md; Licence section"}],"strengths":[{"text":"Tests lipid chemistry and physical plausibility alongside RMSD.","source_ids":["lipp-2026"],"source_locator":"Results/Protein Prediction Errors in Structure Prediction Methods (paragraph 1); Materials\nand Methods/Protein Structure Errors (paragraph 1)"}],"limitations":[{"text":"Input information differs between docking and co-folding methods, and the post-cutoff test subset is much smaller than the full benchmark.","source_ids":["lipp-2026"],"source_locator":"Results/Protein Prediction Errors in Structure Prediction Methods (paragraph 1); Results/Curation of the LiPP Benchmark\nSet for Modeling Lipid–Protein\nInteractions (paragraph 3)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Protein amino-acid sequence and lipid SMILES/CCD identity","Chai-1","Lipid–protein complex or docking-pose coordinates"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["lipp-2026"],"source_locator":"Materials\nand Methods/Molecular Docking and Structure Prediction\nProtocols (paragraph 1); Materials\nand Methods/Protein Structure Errors (paragraph 1)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-ed7f0db85facb1","kind":"model","name":"RNAfold","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["rna-transcriptomes"]},"source_ids":["debfold-2024"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"RNAfold","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"RNAfold is the thermodynamic baseline and an upstream component considered in the DEBFold study.","summary_source_ids":["debfold-2024"],"summary_source_locator":"Methods and Data Sets/DEBFold Workflow (paragraph 1); Methods and Data Sets/DEBFold Workflow/Stage 1: Structure Location Folding Probability Estimation (paragraph 3)","sections":[{"title":"How the evaluated method works","body":"ViennaRNA predicts a minimum-free-energy secondary structure using its thermodynamic energy model; the paper runs default settings.","source_ids":["debfold-2024"],"source_locator":"Methods and Data Sets/DEBFold Workflow/Stage 2: Score-Constrained Optimization Folding (paragraph 1); Methods and Data Sets/DEBFold Workflow (paragraph 1)"},{"title":"Underlying method and version boundaries","body":"The ViennaRNA package computes minimum-free-energy structures, partition functions and associated structure probabilities. RNAfold is a procedure with energy parameters, not a neural language-model checkpoint.","source_ids":["evidence-reported-base-rnafold-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies RNAfold: RNA secondary structure. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["debfold-2024"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-016"}],"facts":[{"label":"Model type","value":"Thermodynamic RNA folding algorithm; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-rnafold-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"ViennaRNA predicts a minimum-free-energy secondary structure using its thermodynamic energy model; the paper runs default settings.","status":"source_checked","source_ids":["debfold-2024"],"source_locator":"Methods and Data Sets/DEBFold Workflow/Stage 2: Score-Constrained Optimization Folding (paragraph 1); Methods and Data Sets/DEBFold Workflow (paragraph 1)"},{"label":"Biological inputs","value":"RNA nucleotide sequence","status":"source_checked","source_ids":["debfold-2024"],"source_locator":"Methods and Data Sets/DEBFold Workflow/Stage 1: Structure Location Folding Probability Estimation (paragraph 3); Methods and Data Sets/Family-Wise Processed RNA Structure Ground-Truth Data Set/Contamination-Free Family-Wise Independent Test Set for Evaluating\nExisting Deep-Learning-Based Structure Prediction Tools (paragraph 1)"},{"label":"Outputs","value":"RNA secondary structure","status":"source_checked","source_ids":["debfold-2024"],"source_locator":"Conclusions (paragraph 1); Methods and Data Sets/Structure Prediction Evaluation Metrics (paragraph 1)"},{"label":"Parameters","value":"Not applicable as a neural parameter count.","status":"inapplicable","source_ids":["debfold-2024"],"source_locator":"Methods and Data Sets/DEBFold Workflow/Stage 1: Structure Location Folding Probability Estimation (paragraph 3); Methods and Data Sets/DEBFold Workflow/Model Training Hyperparameters (paragraph 1)"},{"label":"Known versions / configuration","value":"RNAfold as reported in the DEBFold study; do not assume the 2.6.4 version used in the later BPfold paper.","status":"unreported","source_ids":["debfold-2024"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"Experimentally informed thermodynamic parameters rather than neural pretraining.","status":"source_checked","source_ids":["debfold-2024"],"source_locator":"Methods and Data Sets/DEBFold Workflow/Model Training Hyperparameters (paragraph 1); Methods and Data Sets/DEBFold Workflow (paragraph 1)"},{"label":"Context limits","value":"RNA length is constrained by the RNAfold implementation, algorithm and available memory; there is no learned fixed-token context window.","status":"inapplicable","source_ids":["debfold-2024"],"source_locator":"Table tbl3 (paragraph 1); Results and Discussion/DEBFold Is Robust against Different Thermodynamics-Constrained\nOptimization Algorithms (paragraph 1)"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://github.com/ViennaRNA/ViennaRNA/blob/1ffec79f5e258896160f7362ced8263450f371dc/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-rnafold-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"ViennaRNA licence/disclaimer; see the pinned full text for scope and conditions. (upstream repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-base-rnafold-license-txt"],"source_locator":"license.txt; complete licence text"},{"label":"Weights licence","value":"Not applicable: RNAfold uses thermodynamic energy parameters rather than pretrained neural weights.","status":"inapplicable","source_ids":["debfold-2024"],"source_locator":"Methods and Data Sets/DEBFold Workflow (paragraph 1); Methods and Data Sets/DEBFold Workflow/Stage 1: Structure Location Folding Probability Estimation (paragraph 3)"}],"strengths":[{"text":"Provides a non-neural physical prior against which learned folding models can be assessed.","source_ids":["debfold-2024"],"source_locator":"Results and Discussion/DEBFold Is Robust against Different Thermodynamics-Constrained\nOptimization Algorithms (paragraph 1); Results and Discussion/DEBFold Outperforms Previous Thermodynamics-Based RNA Structure\nPrediction Tools (paragraph 1)"}],"limitations":[{"text":"The conventional nested-pair structure formulation does not model arbitrary pseudoknots or the complete in-vivo folding environment.","source_ids":["debfold-2024"],"source_locator":"Results and Discussion/Limitations of DEBFold (paragraph 1); Introduction (paragraph 5)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["RNA nucleotide sequence","RNAfold","RNA secondary structure"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["debfold-2024"],"source_locator":"Methods and Data Sets/DEBFold Workflow/Stage 2: Score-Constrained Optimization Folding (paragraph 1); Methods and Data Sets/DEBFold Workflow (paragraph 1)"},"coverage":"limited","gaps":["RNAfold as reported in the DEBFold study; do not assume the 2.6.4 version used in the later BPfold paper.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-ee1ae8162c7d67","kind":"model","name":"ARSENAL+ChromBPNet","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["dna-genomes"]},"source_ids":["arsenal-regulatory-dna-2026"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"ARSENAL+ChromBPNet","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"This evaluated pipeline combines ARSENAL regulatory-DNA embeddings with ChromBPNet to predict chromatin accessibility.","summary_source_ids":["arsenal-regulatory-dna-2026"],"summary_source_locator":"Results/ARSENAL embeddings improve supervised sequence models of chromatin accessibility (paragraph 2); Results/ARSENAL embeddings improve supervised sequence models of chromatin accessibility (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"ARSENAL is a short-context masked DNA language model with motif-discovery regularisation. Its embeddings provide additional inputs to a supervised ChromBPNet predictor; the linked result therefore evaluates the combined system.","source_ids":["arsenal-regulatory-dna-2026"],"source_locator":"Abstract (paragraph 1); Results/The ARSENAL model: targeted pretraining with motif discovery regularization (paragraph 2)"},{"title":"What was evaluated","body":"The linked evaluation record identifies ARSENAL+ChromBPNet: regulatory-variant scoring. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["arsenal-regulatory-dna-2026"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b4-001"}],"facts":[{"label":"Model type","value":"Study-specific predictive method; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["arsenal-regulatory-dna-2026"],"source_locator":"Abstract (paragraph 1); Results/The ARSENAL model: targeted pretraining with motif discovery regularization (paragraph 2)"},{"label":"Architecture / procedure","value":"ARSENAL is a short-context masked DNA language model with motif-discovery regularisation. Its embeddings provide additional inputs to a supervised ChromBPNet predictor; the linked result therefore evaluates the combined system.","status":"source_checked","source_ids":["arsenal-regulatory-dna-2026"],"source_locator":"Abstract (paragraph 1); Results/The ARSENAL model: targeted pretraining with motif discovery regularization (paragraph 2)"},{"label":"Biological inputs","value":"Regulatory DNA sequence and task-specific supervised accessibility data","status":"source_checked","source_ids":["arsenal-regulatory-dna-2026"],"source_locator":"Results/ARSENAL improves zero-shot motif discovery (paragraph 6); Methods/Supervised Variant Effect Prediction (paragraph 2)"},{"label":"Outputs","value":"Chromatin-accessibility predictions and derived regulatory-variant scores","status":"source_checked","source_ids":["arsenal-regulatory-dna-2026"],"source_locator":"Methods/Zero-Shot Variant Effect Prediction (paragraph 1); Results/ARSENAL improves zero-shot variant effect prediction in regulatory QTL benchmarks (paragraph 3)"},{"label":"Parameters","value":"Eight transformer blocks, hidden width 768, eight attention heads and feed-forward width 3,072; a combined ARSENAL–ChromBPNet parameter total is not reported in the inspected architecture section.","status":"unreported","source_ids":["arsenal-regulatory-dna-2026"],"source_locator":"Methods/Architecture and Pretraining (paragraph 3); Methods/Supervised Model Training (paragraph 2)"},{"label":"Known versions / configuration","value":"ARSENAL+ChromBPNet is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["arsenal-regulatory-dna-2026"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"ARSENAL pretraining uses approximately 2.3 million ENCODE candidate cis-regulatory elements; chromosomes 6 and 21 are validation and 5, 10, 14, 18, 20 and 22 are test.","status":"source_checked","source_ids":["arsenal-regulatory-dna-2026"],"source_locator":"Results/The ARSENAL model: targeted pretraining with motif discovery regularization (paragraph 1); Methods/Pretraining Data (paragraph 1)"},{"label":"Context limits","value":"ARSENAL encodes 350-bp cCRE windows; ChromBPNet integration supplies a separate supervised prediction procedure.","status":"source_checked","source_ids":["arsenal-regulatory-dna-2026"],"source_locator":"Table T1 (paragraph 1); Methods/Sequence Generation (paragraph 7)"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/kundajelab/regulatory_lm/blob/2c4594f5b7649df3f1372238916f6d95bd31ba1d/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-arsenal-regulatory-dna-2026-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","status":"unreported","source_ids":["evidence-reported-arsenal-regulatory-dna-2026-readme-md"],"source_locator":"README.md and repository-root licence-file search"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-arsenal-regulatory-dna-2026-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Regulatory-enriched pretraining and motif regularisation target features that may be diluted in whole-genome pretraining.","source_ids":["arsenal-regulatory-dna-2026"],"source_locator":"Results/The ARSENAL model: targeted pretraining with motif discovery regularization (paragraph 1); Discussion (paragraph 3)"}],"limitations":[{"text":"This source is a preprint, and improvements from supervised ChromBPNet integration must not be attributed to the pretrained encoder alone.","source_ids":["arsenal-regulatory-dna-2026"],"source_locator":"Methods/Supervised Variant Effect Prediction (paragraph 1); Methods/Supervised Model Training (paragraph 2)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Regulatory DNA sequence and task-specific supervised accessibility data","ARSENAL+ChromBPNet","Chromatin-accessibility predictions and derived regulatory-variant scores"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["arsenal-regulatory-dna-2026"],"source_locator":"Abstract (paragraph 1); Results/The ARSENAL model: targeted pretraining with motif discovery regularization (paragraph 2)"},"coverage":"limited","gaps":["Eight transformer blocks, hidden width 768, eight attention heads and feed-forward width 3,072; a combined ARSENAL–ChromBPNet parameter total is not reported in the inspected architecture section.","ARSENAL+ChromBPNet is the comparison-table label; that label does not specify an immutable weight revision.","No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-f0c630d0565e64","kind":"model","name":"MINGLE","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["cells-tissues"]},"source_ids":["scatac-llmda-2026"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"MINGLE","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"MINGLE is a mutual-information-based scATAC-seq annotation comparator in the scLLMDA study.","summary_source_ids":["scatac-llmda-2026"],"summary_source_locator":"Materials and methods/Cell type annotation via graph domain adaptation/Capture the global consistency relationship of each graph. (paragraph 5); Discussion (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"The method combines cell similarity and topological structure to transfer cell labels; the comparison uses common preprocessing.","source_ids":["scatac-llmda-2026"],"source_locator":"Materials and methods/Benchmark methods (paragraph 1); Results/UMAP visualization comparison across methods (paragraph 1)"},{"title":"What was evaluated","body":"The linked evaluation record identifies MINGLE: Cross-platform scATAC cell-type annotation. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["scatac-llmda-2026"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b3-006"}],"facts":[{"label":"Model type","value":"Study-specific predictive method; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["scatac-llmda-2026"],"source_locator":"Materials and methods/Benchmark methods (paragraph 1); Results/UMAP visualization comparison across methods (paragraph 1)"},{"label":"Architecture / procedure","value":"The method combines cell similarity and topological structure to transfer cell labels; the comparison uses common preprocessing.","status":"source_checked","source_ids":["scatac-llmda-2026"],"source_locator":"Materials and methods/Benchmark methods (paragraph 1); Results/UMAP visualization comparison across methods (paragraph 1)"},{"label":"Biological inputs","value":"scATAC-seq accessibility data and reference cell annotations","status":"source_checked","source_ids":["scatac-llmda-2026"],"source_locator":"Materials and methods/Feature extraction from genomic sequences (paragraph 4); Introduction (paragraph 1)"},{"label":"Outputs","value":"Cell-type annotations","status":"source_checked","source_ids":["scatac-llmda-2026"],"source_locator":"Results/UMAP visualization comparison across methods (paragraph 5); Results/Intra-platform cell type annotation (paragraph 2)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["scatac-llmda-2026","evidence-reported-scatac-llmda-2026-readme-md"],"source_locator":"Materials and methods/Benchmark datasets; Materials and methods/Benchmark methods; Materials and methods/Problem definition; Materials and methods/Feature extraction from genomic sequences; Materials and methods/Cell type annotation via graph domain adaptation/Graph construction.; Materials and methods/Cell type annotation via graph domain adaptation/Capture the local consistency relationship of each graph.; Materials and methods/Cell type annotation via graph domain adaptation/Capture the global consistency relationship of each graph.; Materials and methods/Cell type annotation via graph domain adaptation/Feature fusion via attention.; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"MINGLE is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["scatac-llmda-2026"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"Train on each stated reference scATAC-seq dataset and evaluate transfer to its query dataset. The benchmark states that all methods use SANGO’s preprocessing pipeline; MINGLE-specific optimisation details are not restated.","status":"source_checked","source_ids":["scatac-llmda-2026"],"source_locator":"Materials and methods / Benchmark methods; Tables 1–3"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["scatac-llmda-2026","evidence-reported-scatac-llmda-2026-readme-md"],"source_locator":"Materials and methods/Benchmark datasets; Materials and methods/Benchmark methods; Materials and methods/Problem definition; Materials and methods/Feature extraction from genomic sequences; Materials and methods/Cell type annotation via graph domain adaptation/Graph construction.; Materials and methods/Cell type annotation via graph domain adaptation/Capture the local consistency relationship of each graph.; Materials and methods/Cell type annotation via graph domain adaptation/Capture the global consistency relationship of each graph.; Materials and methods/Cell type annotation via graph domain adaptation/Feature fusion via attention.; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/sheng-guan-2001/scLLMDA/blob/5e24025710bb068312d50a5749ef6bb838ef5a32/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-scatac-llmda-2026-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","status":"unreported","source_ids":["evidence-reported-scatac-llmda-2026-readme-md"],"source_locator":"README.md and repository-root licence-file search"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-scatac-llmda-2026-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Provides a neighbourhood/topology-aware annotation reference.","source_ids":["scatac-llmda-2026"],"source_locator":"Results/Cross-platform cell type annotation (paragraph 1); Abstract (paragraph 2)"}],"limitations":[{"text":"Performance under cross-platform transfer remains task-dependent; the comparator row does not identify every software or reference revision.","source_ids":["scatac-llmda-2026"],"source_locator":"Results/Cross-platform cell type annotation (paragraph 1); Results/UMAP visualization comparison across methods (paragraph 1)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["scATAC-seq accessibility data and reference cell annotations","MINGLE","Cell-type annotations"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["scatac-llmda-2026"],"source_locator":"Materials and methods/Benchmark methods (paragraph 1); Results/UMAP visualization comparison across methods (paragraph 1)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","MINGLE is the comparison-table label; that label does not specify an immutable weight revision.","A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-f23306b94dc7b6","kind":"model","name":"scVI","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["cells-tissues"]},"source_ids":["scxdr-2026"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"scVI","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"This scVI-based comparator is evaluated in the scXDR single-cell drug-response study.","summary_source_ids":["scxdr-2026"],"summary_source_locator":"Abstract (paragraph 2); Methods/Experiment setting (paragraph 2)","sections":[{"title":"How the evaluated method works","body":"A probabilistic variational-autoencoder representation of single-cell expression supports the paper’s downstream response-prediction comparison.","source_ids":["scxdr-2026"],"source_locator":"Methods/Comparison experiments at the individual cell level (paragraph 2); Methods/Comparison experiments at the individual cell level (paragraph 1)"},{"title":"Underlying method and version boundaries","body":"scvi-tools contains probabilistic models for single-cell analyses. scVI, scANVI and study-specific downstream heads are distinct procedures even when distributed through the same software package.","source_ids":["evidence-reported-base-scvi-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies scVI: Cross-dataset single-cell drug response transfer. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["scxdr-2026"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b3-018"}],"facts":[{"label":"Model type","value":"Probabilistic single-cell model; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-scvi-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"A probabilistic variational-autoencoder representation of single-cell expression supports the paper’s downstream response-prediction comparison.","status":"source_checked","source_ids":["scxdr-2026"],"source_locator":"Methods/Comparison experiments at the individual cell level (paragraph 2); Methods/Comparison experiments at the individual cell level (paragraph 1)"},{"label":"Biological inputs","value":"Single-cell expression measurements","status":"source_checked","source_ids":["scxdr-2026"],"source_locator":"Methods/Comparison experiments at the individual cell level (paragraph 2); Methods/Comparison experiments at the cell group level (paragraph 3)"},{"label":"Outputs","value":"Latent representations and the associated response-prediction output","status":"source_checked","source_ids":["scxdr-2026"],"source_locator":"Results/Drug screening and drug response markers (paragraph 3); Results/Pan-cancer level drug clusters and tumor clusters (paragraph 2)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["scxdr-2026","evidence-reported-base-scvi-readme-md"],"source_locator":"Results/Superior performance at the individual cell level compared to various methods; Results/Superior performance at the cell group level compared to various methods; Results/Contribution of model components and architecture to performance; Methods/Data collection and processing; Methods/Model construction; Methods/Experiment setting; Methods/Comparison experiments at the individual cell level; Methods/Comparison experiments at the cell group level; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"scVI is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["scxdr-2026"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"scVI corrects batch effects across single-cell datasets; an additional MLP is then trained on the corrected representation to predict drug response. This evaluates the scVI-plus-MLP pipeline.","status":"source_checked","source_ids":["scxdr-2026"],"source_locator":"Methods / Comparison experiments at the individual cell level; scVI comparator paragraph"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["scxdr-2026","evidence-reported-base-scvi-readme-md"],"source_locator":"Results/Superior performance at the individual cell level compared to various methods; Results/Superior performance at the cell group level compared to various methods; Results/Contribution of model components and architecture to performance; Methods/Data collection and processing; Methods/Model construction; Methods/Experiment setting; Methods/Comparison experiments at the individual cell level; Methods/Comparison experiments at the cell group level; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://github.com/scverse/scvi-tools/blob/73b28e44223621470e582a81a102c107bb22678b/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-scvi-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"BSD 3-Clause (upstream repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-base-scvi-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-base-scvi-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Provides a conventional learned latent representation for comparison with heterogeneous graph transfer.","source_ids":["scxdr-2026"],"source_locator":"Methods/Comparison experiments at the individual cell level (paragraph 2); Methods/Model construction (paragraph 4)"}],"limitations":[{"text":"The response prediction uses a study-specific downstream procedure; scVI alone is not a drug-response assay or validated therapeutic predictor.","source_ids":["scxdr-2026"],"source_locator":"Methods/Case study (paragraph 3); Methods/Comparison experiments at the individual cell level (paragraph 3)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Single-cell expression measurements","scVI","Latent representations and the associated response-prediction output"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["scxdr-2026"],"source_locator":"Methods/Comparison experiments at the individual cell level (paragraph 2); Methods/Comparison experiments at the individual cell level (paragraph 1)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","scVI is the comparison-table label; that label does not specify an immutable weight revision.","A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-f83c0b833411a7","kind":"model","name":"ESM-C","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["proteins-complexes"]},"source_ids":["prime-2026"],"links":[],"attributes":{"entity_level":"method","version":"300M","reported_name":"ESM-C","historical_missing_metadata":{"checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"This protein-language-model configuration is evaluated for viral RBD mutation properties under PRIME.","summary_source_ids":["prime-2026"],"summary_source_locator":"Methods/Model architecture and scale (paragraph 1); Discussion (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"Protein embeddings feed downstream prediction heads, with frozen and fine-tuned encoder conditions explicitly separated.","source_ids":["prime-2026"],"source_locator":"Results/Optimization of neural architectures for real-time deployment (paragraph 1); Methods/Model architecture and scale (paragraph 1)"},{"title":"What was evaluated","body":"The linked evaluation record identifies ESM-C: Mutated RBD binding prediction. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["prime-2026"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-022"}],"facts":[{"label":"Model type","value":"Learned representation pipeline; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["prime-2026"],"source_locator":"Results/Optimization of neural architectures for real-time deployment (paragraph 1); Methods/Model architecture and scale (paragraph 1)"},{"label":"Architecture / procedure","value":"Protein embeddings feed downstream prediction heads, with frozen and fine-tuned encoder conditions explicitly separated.","status":"source_checked","source_ids":["prime-2026"],"source_locator":"Results/Optimization of neural architectures for real-time deployment (paragraph 1); Methods/Model architecture and scale (paragraph 1)"},{"label":"Biological inputs","value":"SARS-CoV-2 receptor-binding-domain sequences","status":"source_checked","source_ids":["prime-2026"],"source_locator":"Results/Benchmark data collection for viral phenotype prediction (paragraph 3); Discussion (paragraph 7)"},{"label":"Outputs","value":"Binding-affinity and expression predictions","status":"source_checked","source_ids":["prime-2026"],"source_locator":"Methods/Computational resources and reproducibility (paragraph 2); Results/Optimization of neural architectures for real-time deployment (paragraph 2)"},{"label":"Parameters","value":"300 million parameters, as identified for this row","status":"source_checked","source_ids":["prime-2026"],"source_locator":"Abstract (paragraph 1); Methods/Clustering and phylogenetic analysis/Clustering (paragraph 1)"},{"label":"Known versions / configuration","value":"300M","status":"source_checked","source_ids":["prime-2026"],"source_locator":"Table Tab1 (paragraph 1); Declarations/Consent for publication (paragraph 1)"},{"label":"Training data / fitting","value":"347,432 RBD sequences in the study; position-stratified partitions test unseen mutational sites.","status":"source_checked","source_ids":["prime-2026"],"source_locator":"Methods/Position-stratified validation protocol/Leakage control (paragraph 1); Background (paragraph 5)"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["prime-2026","evidence-reported-prime-2026-readme-md"],"source_locator":"Results/Optimization of neural architectures for real-time deployment; Methods/Data collection and curation; Methods/Data collection and curation/Outbreak dataset; Methods/Data collection and curation/DMS dataset; Methods/Data collection and curation/BetaCov dataset; Methods/Model architecture and scale; Methods/Model architecture and scale/ESM-2 family; Methods/Model architecture and scale/ESM-C family; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/lanl/prime/blob/d940c51aa0f475b8945789e89761fab0687d5b74/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-prime-2026-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"MIT-style permission text with the repository’s US-government/Triad notice; see the pinned licence. (study repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-prime-2026-license-md"],"source_locator":"LICENSE.md; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-prime-2026-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Contrasts random splitting with a position-stratified protocol that exposes leakage from closely related variants.","source_ids":["prime-2026"],"source_locator":"Methods/Position-stratified validation protocol (paragraph 1); Methods/Position-stratified validation protocol/Leakage control (paragraph 1)"}],"limitations":[{"text":"Random-split performance is not evidence of generalisation to unseen mutation sites; the row’s model size and adaptation must stay attached to its score.","source_ids":["prime-2026"],"source_locator":"Background (paragraph 5); Discussion (paragraph 2)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["SARS-CoV-2 receptor-binding-domain sequences","ESM-C","Binding-affinity and expression predictions"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["prime-2026"],"source_locator":"Results/Optimization of neural architectures for real-time deployment (paragraph 1); Methods/Model architecture and scale (paragraph 1)"},"coverage":"limited","gaps":["A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-f8f0257b98749a","kind":"model","name":"SPIN + ESM2-35M","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["proteins-complexes"]},"source_ids":["spin-protein-function-2026"],"links":[],"attributes":{"entity_level":"method","version":"ESM2-35M frozen","reported_name":"SPIN + ESM2-35M","historical_missing_metadata":{"checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"SPIN uses frozen ESM-2 35M representations to classify proteins into family-specific functional classes.","summary_source_ids":["spin-protein-function-2026"],"summary_source_locator":"3 Results/3.4 Analysis of the SH3 protein family and representational augmentation across clades (paragraph 4); 3 Results/3.3 SPIN on thioredoxin, a widespread protein family (paragraph 1)","sections":[{"title":"How the evaluated method works","body":"The method adapts a question-answering-style architecture to optionally predict a domain span and classify the selected region; it can also operate without domain annotations.","source_ids":["spin-protein-function-2026"],"source_locator":"2 Materials and methods/2.1 SPIN architecture (paragraph 1); 2 Materials and methods/2.1 SPIN architecture (paragraph 7)"},{"title":"Underlying method and version boundaries","body":"ESM-2 is a transformer protein language-model family. The official repository exposes residue embeddings, sequence-level pooling and models at several sizes; the study configuration determines which of these is evaluated.","source_ids":["evidence-reported-base-esm-readme-md"],"source_locator":"README.md; introduction, model description, pretrained-model and usage sections at pinned revision"},{"title":"What was evaluated","body":"The linked evaluation record identifies SPIN + ESM2-35M: protein function annotation. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["spin-protein-function-2026"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-b2-spin-protein-function-2026"}],"facts":[{"label":"Model type","value":"Protein sequence transformer; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["evidence-reported-base-esm-readme-md"],"source_locator":"README.md model description"},{"label":"Architecture / procedure","value":"The method adapts a question-answering-style architecture to optionally predict a domain span and classify the selected region; it can also operate without domain annotations.","status":"source_checked","source_ids":["spin-protein-function-2026"],"source_locator":"2 Materials and methods/2.1 SPIN architecture (paragraph 1); 2 Materials and methods/2.1 SPIN architecture (paragraph 7)"},{"label":"Biological inputs","value":"Protein sequences, optionally with domain annotations","status":"source_checked","source_ids":["spin-protein-function-2026"],"source_locator":"2 Materials and methods/2.1 SPIN architecture (paragraph 1); 2 Materials and methods/2.9 Datasets (paragraph 1)"},{"label":"Outputs","value":"Family-specific functional classes and optional domain spans","status":"source_checked","source_ids":["spin-protein-function-2026"],"source_locator":"3 Results/3.3 SPIN on thioredoxin, a widespread protein family (paragraph 1); 3 Results/3.4 Analysis of the SH3 protein family and representational augmentation across clades (paragraph 4)"},{"label":"Parameters","value":"35-million-parameter frozen ESM-2 feature extractor; total classifier size is separate.","status":"source_checked","source_ids":["spin-protein-function-2026"],"source_locator":"Table vbag064-T1 (paragraph 1); 2 Materials and methods/2.5 An experimental setup for comparative analysis (paragraph 4)"},{"label":"Known versions / configuration","value":"ESM2-35M frozen","status":"source_checked","source_ids":["spin-protein-function-2026"],"source_locator":"Table vbag064-T1 (paragraph 1); 2 Materials and methods/2.5 An experimental setup for comparative analysis (paragraph 4)"},{"label":"Training data / fitting","value":"An independent predictor is fitted per protein family with three random seeds. Functional labels may come from computational annotation; domain boundaries contribute to the domain-span loss when supplied.","status":"source_checked","source_ids":["spin-protein-function-2026"],"source_locator":"2 Materials and methods/2.3 Family-specific model tuning (paragraph 1); 2 Materials and methods/2.4 Bootstrap analysis of domain span predictions (paragraph 1)"},{"label":"Context limits","value":"Sequences longer than 1,024 amino acids are excluded in this study.","status":"source_checked","source_ids":["spin-protein-function-2026"],"source_locator":"2 Materials and methods/2.1 SPIN architecture (paragraph 5); 3 Results/3.4 Analysis of the SH3 protein family and representational augmentation across clades (paragraph 2)"},{"label":"Access","value":"Official upstream implementation and usage documentation: https://github.com/facebookresearch/esm/blob/2b369911bb5b4b0dda914521b9475cad1656b2ac/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-base-esm-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"MIT (upstream repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).","status":"source_checked","source_ids":["evidence-reported-base-esm-license"],"source_locator":"LICENSE; complete licence text"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-base-esm-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Restricting classification to relevant domains provides an explicit route to family-specific functional distinctions.","source_ids":["spin-protein-function-2026"],"source_locator":"3 Results/3.3 SPIN on thioredoxin, a widespread protein family (paragraph 1); 3 Results (paragraph 1)"}],"limitations":[{"text":"The defined functional classes and domain assumptions limit the claim; this is not an unrestricted protein-function annotator.","source_ids":["spin-protein-function-2026"],"source_locator":"2 Materials and methods/2.1 SPIN architecture (paragraph 2); 5 Discussion and conclusions (paragraph 4)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Protein sequences, optionally with domain annotations","SPIN + ESM2-35M","Family-specific functional classes and optional domain spans"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["spin-protein-function-2026"],"source_locator":"2 Materials and methods/2.1 SPIN architecture (paragraph 1); 2 Materials and methods/2.1 SPIN architecture (paragraph 7)"},"coverage":"limited","gaps":["The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-fa2da404b4d08e","kind":"model","name":"DEELIG","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["molecular-interactions"]},"source_ids":["deelig-2021"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"DEELIG","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"DEELIG predicts protein–ligand binding affinity from separately supplied protein and ligand information.","summary_source_ids":["deelig-2021"],"summary_source_locator":"Materials and Methods/Novel data set: raw data (paragraph 3); Materials and Methods/Novel data set: raw data (paragraph 2)","sections":[{"title":"How the evaluated method works","body":"Convolutional networks learn interactions between featurised protein pockets and ligands. The study explores atomic and residue-level representations, including grid-based atomic descriptors.","source_ids":["deelig-2021"],"source_locator":"Materials and Methods/Feature extraction/Protein-pocket features (paragraph 1); Discussion (paragraph 2)"},{"title":"What was evaluated","body":"The linked evaluation record identifies DEELIG: Protein–ligand binding affinity prediction. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["deelig-2021"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b3-042"}],"facts":[{"label":"Model type","value":"Convolutional neural network; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["deelig-2021"],"source_locator":"Materials and Methods/Feature extraction/Protein-pocket features (paragraph 1); Discussion (paragraph 2)"},{"label":"Architecture / procedure","value":"Convolutional networks learn interactions between featurised protein pockets and ligands. The study explores atomic and residue-level representations, including grid-based atomic descriptors.","status":"source_checked","source_ids":["deelig-2021"],"source_locator":"Materials and Methods/Feature extraction/Protein-pocket features (paragraph 1); Discussion (paragraph 2)"},{"label":"Biological inputs","value":"High-resolution protein structures and non-peptide ligands, without requiring a docked complex as the user input","status":"source_checked","source_ids":["deelig-2021"],"source_locator":"Abstract (paragraph 1); Introduction (paragraph 3)"},{"label":"Outputs","value":"Predicted protein–ligand binding affinity","status":"source_checked","source_ids":["deelig-2021"],"source_locator":"Materials and Methods/Novel data set: raw data (paragraph 3); Materials and Methods/Novel data set: raw data (paragraph 2)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["deelig-2021","evidence-reported-deelig-2021-readme-md"],"source_locator":"Materials and Methods/Novel data set: raw data; Materials and Methods/Data set refinement; Materials and Methods/Feature extraction; Materials and Methods/Feature extraction/Protein-pocket features; Materials and Methods/Feature extraction/Ligand features; Materials and Methods/Feature extraction/Grid formation; Materials and Methods/Strategies; Materials and Methods/Strategies/Atomic model/Preprocessing; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"DEELIG is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["deelig-2021"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"An in-house prepared protein–ligand dataset described in the study, with explicitly separated training, validation and test records.","status":"source_checked","source_ids":["deelig-2021"],"source_locator":"Materials and Methods/Strategies/Composite model/Training (paragraph 2); Materials and Methods/Data set refinement (paragraph 2)"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["deelig-2021","evidence-reported-deelig-2021-readme-md"],"source_locator":"Materials and Methods/Novel data set: raw data; Materials and Methods/Data set refinement; Materials and Methods/Feature extraction; Materials and Methods/Feature extraction/Protein-pocket features; Materials and Methods/Feature extraction/Ligand features; Materials and Methods/Feature extraction/Grid formation; Materials and Methods/Strategies; Materials and Methods/Strategies/Atomic model/Preprocessing; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/asadahmedtech/DEELIG/blob/3a3993fc903c40f1ce904111c8e085c79fb45df6/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-deelig-2021-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","status":"unreported","source_ids":["evidence-reported-deelig-2021-readme-md"],"source_locator":"README.md and repository-root licence-file search"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-deelig-2021-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"The user-facing procedure does not require a precomputed docked ligand pose.","source_ids":["deelig-2021"],"source_locator":"Introduction (paragraph 3); Introduction (paragraph 5)"}],"limitations":[{"text":"The paper’s applicability conditions include high-resolution protein structures and non-peptide ligands; this does not establish unrestricted chemical or receptor coverage.","source_ids":["deelig-2021"],"source_locator":"Materials and Methods/Novel data set: raw data (paragraph 3); Materials and Methods/Novel data set: raw data (paragraph 2)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["High-resolution protein structures and non-peptide ligands, without requiring a docked complex as the user input","DEELIG","Predicted protein–ligand binding affinity"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["deelig-2021"],"source_locator":"Materials and Methods/Feature extraction/Protein-pocket features (paragraph 1); Discussion (paragraph 2)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","DEELIG is the comparison-table label; that label does not specify an immutable weight revision.","A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-fcf2cd29a81aae","kind":"model","name":"scGen","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["cells-tissues"]},"source_ids":["insilico-perturbation-auprc-2025"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"scGen","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"This scGen configuration is evaluated for predicted expression responses and recovery of differentially expressed genes.","summary_source_ids":["insilico-perturbation-auprc-2025"],"summary_source_locator":"Results/DEG prediction on cell-level responses under single stimulus across multiple cell types (paragraph 3); Results/DEG prediction on cell-level responses under single stimulus across multiple cell types/High \\documentclass[12pt]{minimal}\n\\usepackage{amsmath}\n\\usepackage{wasysym}\n\\usepackage{amsfonts}\n\\usepackage{amssymb}\n\\usepackage{amsbsy}\n\\usepackage{upgreek}\n\\usepackage{mathrsfs}\n\\setlength{\\oddsidemargin}{-69pt}\n\\begin{document}\n$R^{2}$\\end{document} does not imply ability to identify differentially expressed genes (paragraph 6)","sections":[{"title":"How the evaluated method works","body":"A learned single-cell perturbation predictor is assessed with differential-expression precision/recall in addition to overall expression-fit metrics.","source_ids":["insilico-perturbation-auprc-2025"],"source_locator":"Results/DEG prediction on cell-level responses under single stimulus across multiple cell types/Comparative analysis of two-factor and scGen models for DEG prediction (paragraph 13); Materials and methods/Assessing model performance via differential expression classification (paragraph 12)"},{"title":"What was evaluated","body":"The linked evaluation record identifies scGen: differentially expressed gene identification. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["insilico-perturbation-auprc-2025"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b4-016"}],"facts":[{"label":"Model type","value":"Study-specific predictive method; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["insilico-perturbation-auprc-2025"],"source_locator":"Results/DEG prediction on cell-level responses under single stimulus across multiple cell types/Comparative analysis of two-factor and scGen models for DEG prediction (paragraph 13); Materials and methods/Assessing model performance via differential expression classification (paragraph 12)"},{"label":"Architecture / procedure","value":"A learned single-cell perturbation predictor is assessed with differential-expression precision/recall in addition to overall expression-fit metrics.","status":"source_checked","source_ids":["insilico-perturbation-auprc-2025"],"source_locator":"Results/DEG prediction on cell-level responses under single stimulus across multiple cell types/Comparative analysis of two-factor and scGen models for DEG prediction (paragraph 13); Materials and methods/Assessing model performance via differential expression classification (paragraph 12)"},{"label":"Biological inputs","value":"Single-cell expression profiles and perturbation conditions","status":"source_checked","source_ids":["insilico-perturbation-auprc-2025"],"source_locator":"Results/DEG prediction on population-level responses under multiple perturbations across multiple cell types (paragraph 2); Results/DEG prediction on population-level responses under multiple perturbations across multiple cell types (paragraph 3)"},{"label":"Outputs","value":"Predicted perturbed expression profiles","status":"source_checked","source_ids":["insilico-perturbation-auprc-2025"],"source_locator":"Materials and methods/Limitations of traditional performance metrics in perturbation response prediction (paragraph 4); Results/DEG prediction on cell-level responses under single stimulus across multiple cell types/Comparative analysis of two-factor and scGen models for DEG prediction (paragraph 3)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["insilico-perturbation-auprc-2025","evidence-reported-insilico-perturbation-auprc-2025-readme-md"],"source_locator":"Materials and methods/Mathematical formulation of cellular perturbation experiments; Materials and methods/Limitations of traditional performance metrics in perturbation response prediction; Materials and methods/Assessing model performance via differential expression classification; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"scGen is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["insilico-perturbation-auprc-2025"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"The study uses a processed PBMC dataset of 18,868 cells, seven cell types and 6,998 highly variable genes. Seven training iterations each withhold the stimulated cells of one cell type, evaluating an out-of-sample perturbation response.","status":"source_checked","source_ids":["insilico-perturbation-auprc-2025"],"source_locator":"Results / DEG prediction on cell-level responses under single stimulus across multiple cell types"},{"label":"Context limits","value":"The evaluated input has 6,998 highly variable genes per cell; this is a feature set, not a sequence-token limit.","status":"source_checked","source_ids":["insilico-perturbation-auprc-2025"],"source_locator":"Results / DEG prediction on cell-level responses under single stimulus across multiple cell types"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/hxzhu491/Cell-Perturbation-evaluation-Metric/blob/3b5f8a2ed001c074936287ece478747376c8f5bf/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-insilico-perturbation-auprc-2025-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","status":"unreported","source_ids":["evidence-reported-insilico-perturbation-auprc-2025-readme-md"],"source_locator":"README.md and repository-root licence-file search"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-insilico-perturbation-auprc-2025-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"The evaluation tests biologically relevant differential-expression recovery rather than relying only on global R².","source_ids":["insilico-perturbation-auprc-2025"],"source_locator":"Results/DEG prediction on cell-level responses under single stimulus across multiple cell types/High \\documentclass[12pt]{minimal}\n\\usepackage{amsmath}\n\\usepackage{wasysym}\n\\usepackage{amsfonts}\n\\usepackage{amssymb}\n\\usepackage{amsbsy}\n\\usepackage{upgreek}\n\\usepackage{mathrsfs}\n\\setlength{\\oddsidemargin}{-69pt}\n\\begin{document}\n$R^{2}$\\end{document} does not imply ability to identify differentially expressed genes (paragraph 8); Materials and methods/Limitations of traditional performance metrics in perturbation response prediction (paragraph 6)"}],"limitations":[{"text":"High overall expression agreement can coexist with poor recovery of differentially expressed genes; both evaluation targets must remain visible.","source_ids":["insilico-perturbation-auprc-2025"],"source_locator":"Abstract (paragraph 1); Results/DEG prediction on cell-level responses under single stimulus across multiple cell types (paragraph 3)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Single-cell expression profiles and perturbation conditions","scGen","Predicted perturbed expression profiles"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["insilico-perturbation-auprc-2025"],"source_locator":"Results/DEG prediction on cell-level responses under single stimulus across multiple cell types/Comparative analysis of two-factor and scGen models for DEG prediction (paragraph 13); Materials and methods/Assessing model performance via differential expression classification (paragraph 12)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","scGen is the comparison-table label; that label does not specify an immutable weight revision.","No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-model-fdac4c1ec8a433","kind":"model","name":"ProtT5 embeddings + ensemble classifier","description":"Identity as reported in this paper. Unspecified versions are not assumed equivalent to other papers.","status":"needs_review","facets":{"areas":["proteins-complexes"]},"source_ids":["protein-binding-sites-2023"],"links":[],"attributes":{"entity_level":"method","version":null,"reported_name":"ProtT5 embeddings + ensemble classifier","historical_missing_metadata":{"version":"not_reported_in_legacy_extract","checkpoint_revision":"not_reported_in_legacy_extract","training_data":"not_reported_in_legacy_extract","licence":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"EDLMPPI uses ProtT5 embeddings and an ensemble predictor to identify protein–protein interaction sites.","summary_source_ids":["protein-binding-sites-2023"],"summary_source_locator":"Conclusions (paragraph 1); Abstract (paragraph 2)","sections":[{"title":"How the evaluated method works","body":"ProtT5 converts the sequence into distributed residue representations, which feed the study’s ensemble deep-learning binding-site classifier.","source_ids":["protein-binding-sites-2023"],"source_locator":"Abstract (paragraph 2); Abstract (paragraph 1)"},{"title":"What was evaluated","body":"The linked evaluation record identifies ProtT5 embeddings + ensemble classifier: protein-protein binding-site prediction. Its dataset, split, adaptation and evidence origin remain attached to the reported results.","source_ids":["protein-binding-sites-2023"],"source_locator":"The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b4-010"}],"facts":[{"label":"Model type","value":"Study-specific predictive method; this record is the paper-specific evaluated configuration.","status":"source_checked","source_ids":["protein-binding-sites-2023"],"source_locator":"Abstract (paragraph 2); Abstract (paragraph 1)"},{"label":"Architecture / procedure","value":"ProtT5 converts the sequence into distributed residue representations, which feed the study’s ensemble deep-learning binding-site classifier.","status":"source_checked","source_ids":["protein-binding-sites-2023"],"source_locator":"Abstract (paragraph 2); Abstract (paragraph 1)"},{"label":"Biological inputs","value":"Protein amino-acid sequences","status":"source_checked","source_ids":["protein-binding-sites-2023"],"source_locator":"Methods/Feature descriptors/Dynamic global contextual information (paragraph 2); Methods/Feature descriptors/Multi-source biological features (paragraph 4)"},{"label":"Outputs","value":"Per-residue protein–protein binding-site predictions","status":"source_checked","source_ids":["protein-binding-sites-2023"],"source_locator":"Methods/Datasets (paragraph 2); Abstract (paragraph 2)"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["protein-binding-sites-2023","evidence-reported-protein-binding-sites-2023-readme-md"],"source_locator":"Results and discussion/Comparing EDLMPPI with different deep learning architectures; Results and discussion/Comparing EDLMPPI with other PPIs prediction methods; Methods/Datasets; Methods/Feature descriptors; Methods/Feature descriptors/Dynamic global contextual information; Methods/Feature descriptors/Multi-source biological features; Methods/Ensemble deep memory capsule network; Methods/Ensemble deep memory capsule network/Deep memory network; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision"},{"label":"Known versions / configuration","value":"ProtT5 embeddings + ensemble classifier is the comparison-table label; that label does not specify an immutable weight revision.","status":"unreported","source_ids":["protein-binding-sites-2023"],"source_locator":"Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label."},{"label":"Training data / fitting","value":"Dset_448, Dset_72 and Dset_164 are the study’s benchmark collections; their training/test roles are retained in the associated evaluation records.","status":"source_checked","source_ids":["protein-binding-sites-2023"],"source_locator":"Methods/Datasets (paragraph 1); Methods/Datasets (paragraph 2)"},{"label":"Context limits","value":"A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["protein-binding-sites-2023","evidence-reported-protein-binding-sites-2023-readme-md"],"source_locator":"Results and discussion/Comparing EDLMPPI with different deep learning architectures; Results and discussion/Comparing EDLMPPI with other PPIs prediction methods; Methods/Datasets; Methods/Feature descriptors; Methods/Feature descriptors/Dynamic global contextual information; Methods/Feature descriptors/Multi-source biological features; Methods/Ensemble deep memory capsule network; Methods/Ensemble deep memory capsule network/Deep memory network; inspected for explicit maximum input length (dataset lengths and family-wide limits are not substituted); README.md at pinned repository revision"},{"label":"Access","value":"Official study implementation and usage documentation: https://github.com/houzl3416/EDLMPPI/blob/78e4a7b36bb83ccf4274786b859125178804f434/README.md. This pinned documentation revision is not automatically the evaluated weight revision.","status":"source_checked","source_ids":["evidence-reported-protein-binding-sites-2023-readme-md"],"source_locator":"README.md; installation, model download and usage instructions"},{"label":"Code licence","value":"No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","status":"unreported","source_ids":["evidence-reported-protein-binding-sites-2023-readme-md"],"source_locator":"README.md and repository-root licence-file search"},{"label":"Weights licence","value":"The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","status":"unreported","source_ids":["evidence-reported-protein-binding-sites-2023-readme-md"],"source_locator":"README.md; checkpoint/access documentation and licence scope"}],"strengths":[{"text":"Uses pretrained sequence features for a label-scarce residue-level prediction task.","source_ids":["protein-binding-sites-2023"],"source_locator":"Conclusions (paragraph 1); Methods/Feature descriptors/Multi-source biological features (paragraph 2)"}],"limitations":[{"text":"Residue-level binding-site prediction differs from deciding whether an arbitrary pair of proteins interacts.","source_ids":["protein-binding-sites-2023"],"source_locator":"Methods/Datasets (paragraph 2); Results and discussion/Protein binding domains analysis (paragraph 2)"}],"diagram":{"title":"Evaluated procedure (conceptual)","steps":["Protein amino-acid sequences","ProtT5 embeddings + ensemble classifier","Per-residue protein–protein binding-site predictions"],"caption":"Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.","source_ids":["protein-binding-sites-2023"],"source_locator":"Abstract (paragraph 2); Abstract (paragraph 1)"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","ProtT5 embeddings + ensemble classifier is the comparison-table label; that label does not specify an immutable weight revision.","A maximum input/context length for this exact evaluated configuration is not established by the inspected sources.","No explicit code licence was established from the paper’s availability statement and inspected repository-root documentation.","The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately.","An immutable checkpoint or fitted-artifact digest is not supplied by the comparison-table identity."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction."}}}}
{"id":"reported-task-003d746a129c9b","kind":"benchmark","name":"differentially expressed gene identification","description":"","status":"needs_review","facets":{"areas":["cells-tissues"],"tasks":["differentially expressed gene identification"]},"source_ids":["insilico-perturbation-auprc-2025"],"links":[{"relation":"dataset","target_id":"reported-dataset-7bf2cf7d2b2d01"}],"attributes":{"entity_level":"task","version":null,"task":"differentially expressed gene identification","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Differential-expression evaluation compares predicted gene responses against observed differential-expression labels.","summary_source_ids":["insilico-perturbation-auprc-2025"],"summary_source_locator":"Methods: In silico models and evaluation of performance; cached text lines 100–103","sections":[{"title":"Evaluation methodology","body":"Cell-type–perturbation combinations with responses available across donors. A subset of complete combinations is held out for testing; models are fitted separately for each donor. Gene-expression R-squared is contrasted with precision–recall evaluation of differentially expressed genes. Cell-type-only, perturbation-only and two-factor linear models, plus SI-A. Complete cell-type–perturbation combinations are held out. Models are fitted separately for each donor, so this is not an unseen-donor generalization test.","source_ids":["insilico-perturbation-auprc-2025"],"source_locator":"Methods: In silico models and evaluation of performance; cached text lines 100–103"}],"facts":[{"label":"Datasets","value":"Cell-type–perturbation combinations with responses available across donors.","status":"source_checked","source_ids":["insilico-perturbation-auprc-2025"],"source_locator":"Methods: In silico models and evaluation of performance; cached text lines 100–103"},{"label":"Splits","value":"A subset of complete combinations is held out for testing; models are fitted separately for each donor.","status":"source_checked","source_ids":["insilico-perturbation-auprc-2025"],"source_locator":"Methods: In silico models and evaluation of performance; cached text lines 100–103"},{"label":"Metrics","value":"Gene-expression R-squared is contrasted with precision–recall evaluation of differentially expressed genes.","status":"source_checked","source_ids":["insilico-perturbation-auprc-2025"],"source_locator":"Methods: In silico models and evaluation of performance; cached text lines 100–103"},{"label":"Baselines","value":"Cell-type-only, perturbation-only and two-factor linear models, plus SI-A.","status":"source_checked","source_ids":["insilico-perturbation-auprc-2025"],"source_locator":"Methods: In silico models and evaluation of performance; cached text lines 100–103"},{"label":"Leakage controls","value":"Complete cell-type–perturbation combinations are held out. Models are fitted separately for each donor, so this is not an unseen-donor generalization test.","status":"source_checked","source_ids":["insilico-perturbation-auprc-2025"],"source_locator":"Methods: In silico models and evaluation of performance; cached text lines 100–103"},{"label":"Uncertainty","value":"Figure 6 shows the distribution of AUPRC across 100 held-out cell-type–perturbation pairs and marks their averages. These are between-pair distributions, not confidence intervals for an aggregate score; no such interval is specified in this evaluation passage.","status":"source_checked","source_ids":["insilico-perturbation-auprc-2025"],"source_locator":"Results: In silico models and evaluation of performance; Figure 6 caption"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["insilico-perturbation-auprc-2025"],"source_locator":"Methods: In silico models and evaluation of performance; cached text lines 100–103"},{"label":"Organisms","value":"Human peripheral blood mononuclear cells from three healthy donors, measured 24 hours after compound treatment.","status":"source_checked","source_ids":["insilico-perturbation-auprc-2025"],"source_locator":"Results: DEG prediction on population-level responses under multiple perturbations across multiple cell types"},{"label":"Assays","value":"Measured expression responses and differential-expression targets.","status":"source_checked","source_ids":["insilico-perturbation-auprc-2025"],"source_locator":"Methods: In silico models and evaluation of performance; cached text lines 100–103"},{"label":"Allowed inputs","value":"Cell type, perturbation identity and expression information permitted by the chosen model.","status":"source_checked","source_ids":["insilico-perturbation-auprc-2025"],"source_locator":"Methods: In silico models and evaluation of performance; cached text lines 100–103"},{"label":"Adaptation","value":"Models fit separately per donor with complete cell-type–perturbation combinations held out.","status":"source_checked","source_ids":["insilico-perturbation-auprc-2025"],"source_locator":"Methods: In silico models and evaluation of performance; cached text lines 100–103"}],"strengths":[{"text":"Holding out complete cell-type–perturbation combinations tests compositional response prediction.","source_ids":["insilico-perturbation-auprc-2025"],"source_locator":"Methods: In silico models and evaluation of performance; cached text lines 100–103"}],"limitations":[{"text":"Holding out cell-type–perturbation combinations is not donor-held-out evaluation. Distributions across testing pairs quantify heterogeneity across prediction problems, not confidence intervals for generalization to new donors.","source_ids":["insilico-perturbation-auprc-2025"],"source_locator":"Results: In silico models and evaluation of performance; Figure 6 caption; Results: DEG prediction on population-level responses under multiple perturbations across multiple cell types"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Cell type, perturbation identity and expression information permitted by the chosen model.","Evaluation: A subset of complete combinations is held out for testing; models are fitted separately for each donor.","Readout: Gene-expression R-squared is contrasted with precision–recall evaluation of differentially expressed genes."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["insilico-perturbation-auprc-2025"],"source_locator":"Methods: In silico models and evaluation of performance; cached text lines 100–103"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Targeted full-paper and supplement review of the outstanding task fields, with original dataset metadata checked where accessible. Source-scoped omissions are explicit; no independent benchmark reproduction or numerical-result change."}}}}
{"id":"reported-task-00e594df6a182d","kind":"benchmark","name":"Mutated RBD binding prediction","description":"","status":"needs_review","facets":{"areas":["proteins-complexes"],"tasks":["Mutated RBD binding prediction"]},"source_ids":["prime-2026"],"links":[{"relation":"dataset","target_id":"reported-dataset-becc215358afd0"}],"attributes":{"entity_level":"task","version":null,"task":"Mutated RBD binding prediction","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"PRIME examines how protein representations generalise across a published viral-variant dataset. This record covers its binding-prediction endpoint.","summary_source_ids":["prime-2026"],"summary_source_locator":"Abstract; Results: Benchmark data collection for viral phenotype prediction; Table 1 task and column headings","sections":[{"title":"What the evaluation establishes","body":"The paper contrasts ordinary random assessment with assessment across withheld mutation positions. These settings ask different generalisation questions and their scores should not be pooled. Binding and expression are separate measured labels; neither is a direct measure of infection, transmission or clinical risk.","source_ids":["prime-2026"],"source_locator":"Abstract; Results: Benchmark data collection for viral phenotype prediction; Table 1 task and column headings"}],"facts":[{"label":"Entity type","value":"Paper-specific evaluation task; this profile is a descriptive evidence summary.","status":"source_checked","source_ids":["prime-2026"],"source_locator":"Abstract; Results: Benchmark data collection for viral phenotype prediction; Table 1 task and column headings"},{"label":"Datasets","value":"The paper’s PRIME receptor-binding-domain variant collection; this is a paper-specific task, not a universal protein benchmark.","status":"source_checked","source_ids":["prime-2026"],"source_locator":"Abstract; Results: Benchmark data collection for viral phenotype prediction; Table 1 task and column headings"},{"label":"Organisms","value":"SARS-CoV-2 in the reported variant dataset.","status":"source_checked","source_ids":["prime-2026"],"source_locator":"Abstract; Results: Benchmark data collection for viral phenotype prediction; Table 1 task and column headings"},{"label":"Assays","value":"Previously published binding and expression measurements provide phenotype labels; this record concerns binding prediction.","status":"source_checked","source_ids":["prime-2026"],"source_locator":"Abstract; Results: Benchmark data collection for viral phenotype prediction; Table 1 task and column headings"},{"label":"Splits","value":"Random and position-stratified settings are distinct. The linked results retain the reported setting; this profile does not supply an executable split manifest.","status":"source_checked","source_ids":["prime-2026"],"source_locator":"Abstract; Results: Benchmark data collection for viral phenotype prediction; Table 1 task and column headings"},{"label":"Allowed inputs","value":"Protein-sequence representations associated with measured phenotype labels.","status":"source_checked","source_ids":["prime-2026"],"source_locator":"Abstract; Results: Benchmark data collection for viral phenotype prediction; Table 1 task and column headings"},{"label":"Adaptation","value":"The paper distinguishes frozen representations from adapted models. Individual evaluation records retain that distinction.","status":"source_checked","source_ids":["prime-2026"],"source_locator":"Abstract; Results: Benchmark data collection for viral phenotype prediction; Table 1 task and column headings"},{"label":"Metrics","value":"R² for the linked binding-regression results; preserve the endpoint and split beside each value.","status":"source_checked","source_ids":["prime-2026"],"source_locator":"Abstract; Results: Benchmark data collection for viral phenotype prediction; Table 1 task and column headings"},{"label":"Baselines","value":"The paper compares pretrained encoder configurations. The exact method named in each retained result remains its identity; no model recommendation follows from this summary.","status":"source_checked","source_ids":["prime-2026"],"source_locator":"Abstract; Results: Benchmark data collection for viral phenotype prediction; Table 1 task and column headings"}],"strengths":[{"text":"The contrasting assessment settings expose how a reported score depends on the generalisation question.","source_ids":["prime-2026"],"source_locator":"Abstract; Results: Benchmark data collection for viral phenotype prediction; Table 1 task and column headings"}],"limitations":[{"text":"Prediction of a measured binding label does not establish infectivity or clinical risk. A source-checked score alone does not verify the full dataset lineage or split manifest.","source_ids":["prime-2026"],"source_locator":"Abstract; Results: Benchmark data collection for viral phenotype prediction; Table 1 task and column headings"}],"diagram":{"title":"Conceptual assessment outline","steps":["Published phenotype dataset","Distinguish assessment settings","Evaluate the specified endpoint","Read the metric with its evidence"],"caption":"Conceptual overview of the published statistical assessment.","source_ids":["prime-2026"],"source_locator":"Abstract; Results: Benchmark data collection for viral phenotype prediction; Table 1 task and column headings"},"coverage":"limited","gaps":["This profile reviews the evaluation question, data type and statistical readout. It does not reconstruct an executable protocol, verify exact split membership or independently reproduce the experiment."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected the cited primary abstract and descriptive computational-evaluation sections. Review covers the descriptive claims shown; no executable protocol was reconstructed. Original numerical records retain their prior transcription review."}}}}
{"id":"reported-task-016f70615f2cfc","kind":"benchmark","name":"RNA secondary structure","description":"","status":"needs_review","facets":{"areas":["rna-transcriptomes"],"tasks":["RNA secondary structure"]},"source_ids":["debfold-2024"],"links":[{"relation":"dataset","target_id":"reported-dataset-18ebde58579c2b"}],"attributes":{"entity_level":"task","version":null,"task":"RNA secondary structure","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"RNA secondary-structure assessment emphasizes held-out structural families rather than only novel sequences from familiar families.","summary_source_ids":["debfold-2024"],"summary_source_locator":"Methods: Family-Wise Processed RNA Structure Ground-Truth Data Set; Contamination-Free Family-Wise Independent Test Set; Structure Prediction Evaluation Metrics; cached text lines 26–31, 36–37","sections":[{"title":"Evaluation methodology","body":"bpRNA-1m-derived structural families and a separate bpRNA-new evaluation collection. Family-wise training/validation/test partitioning with cross-validation inside the training process. Base-pair precision, recall and structure F1. Existing RNA structure predictors, including SPOT-RNA and SPOT-RNA2, are considered in the independent-set design. A separate bpRNA-new test set is constructed because some comparators were trained on bpRNA-1m.","source_ids":["debfold-2024"],"source_locator":"Methods: Family-Wise Processed RNA Structure Ground-Truth Data Set; Contamination-Free Family-Wise Independent Test Set; Structure Prediction Evaluation Metrics; cached text lines 26–31, 36–37"}],"facts":[{"label":"Datasets","value":"bpRNA-1m-derived structural families and a separate bpRNA-new evaluation collection.","status":"source_checked","source_ids":["debfold-2024"],"source_locator":"Methods: Family-Wise Processed RNA Structure Ground-Truth Data Set; Contamination-Free Family-Wise Independent Test Set; Structure Prediction Evaluation Metrics; cached text lines 26–31, 36–37"},{"label":"Splits","value":"Family-wise training/validation/test partitioning with cross-validation inside the training process.","status":"source_checked","source_ids":["debfold-2024"],"source_locator":"Methods: Family-Wise Processed RNA Structure Ground-Truth Data Set; Contamination-Free Family-Wise Independent Test Set; Structure Prediction Evaluation Metrics; cached text lines 26–31, 36–37"},{"label":"Metrics","value":"Base-pair precision, recall and structure F1.","status":"source_checked","source_ids":["debfold-2024"],"source_locator":"Methods: Family-Wise Processed RNA Structure Ground-Truth Data Set; Contamination-Free Family-Wise Independent Test Set; Structure Prediction Evaluation Metrics; cached text lines 26–31, 36–37"},{"label":"Baselines","value":"Existing RNA structure predictors, including SPOT-RNA and SPOT-RNA2, are considered in the independent-set design.","status":"source_checked","source_ids":["debfold-2024"],"source_locator":"Methods: Family-Wise Processed RNA Structure Ground-Truth Data Set; Contamination-Free Family-Wise Independent Test Set; Structure Prediction Evaluation Metrics; cached text lines 26–31, 36–37"},{"label":"Leakage controls","value":"A separate bpRNA-new test set is constructed because some comparators were trained on bpRNA-1m.","status":"source_checked","source_ids":["debfold-2024"],"source_locator":"Methods: Family-Wise Processed RNA Structure Ground-Truth Data Set; Contamination-Free Family-Wise Independent Test Set; Structure Prediction Evaluation Metrics; cached text lines 26–31, 36–37"},{"label":"Uncertainty","value":"The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","status":"unreported","source_ids":["debfold-2024"],"source_locator":"Methods: Family-Wise Processed RNA Structure Ground-Truth Data Set; Contamination-Free Family-Wise Independent Test Set; Structure Prediction Evaluation Metrics; cached text lines 26–31, 36–37"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["debfold-2024"],"source_locator":"Methods: Family-Wise Processed RNA Structure Ground-Truth Data Set; Contamination-Free Family-Wise Independent Test Set; Structure Prediction Evaluation Metrics; cached text lines 26–31, 36–37"},{"label":"Organisms","value":"The evaluated RNA sets are selected by Rfam family, length and source rather than by organism. The dataset-construction sections identify bpRNA/Rfam and PDB origins but do not enumerate the organism composition of TestSet α, β and γ.","status":"unreported","source_ids":["debfold-2024"],"source_locator":"Methods and Data Sets: Family-Wise Processed RNA Structure Ground-Truth Data Set, including all three test-set construction subsections"},{"label":"Assays","value":"Reference secondary structures.","status":"source_checked","source_ids":["debfold-2024"],"source_locator":"Methods: Family-Wise Processed RNA Structure Ground-Truth Data Set; Contamination-Free Family-Wise Independent Test Set; Structure Prediction Evaluation Metrics; cached text lines 26–31, 36–37"},{"label":"Allowed inputs","value":"RNA sequences.","status":"source_checked","source_ids":["debfold-2024"],"source_locator":"Methods: Family-Wise Processed RNA Structure Ground-Truth Data Set; Contamination-Free Family-Wise Independent Test Set; Structure Prediction Evaluation Metrics; cached text lines 26–31, 36–37"},{"label":"Adaptation","value":"Supervised secondary-structure prediction with family-wise partitions and training cross-validation.","status":"source_checked","source_ids":["debfold-2024"],"source_locator":"Methods: Family-Wise Processed RNA Structure Ground-Truth Data Set; Contamination-Free Family-Wise Independent Test Set; Structure Prediction Evaluation Metrics; cached text lines 26–31, 36–37"}],"strengths":[{"text":"Family-wise splitting tests transfer beyond within-family sequence similarity.","source_ids":["debfold-2024"],"source_locator":"Methods: Family-Wise Processed RNA Structure Ground-Truth Data Set; Contamination-Free Family-Wise Independent Test Set; Structure Prediction Evaluation Metrics; cached text lines 26–31, 36–37"}],"limitations":[{"text":"Family composition, sequence length and training overlap differ between the independent test sets. The paper explicitly flags potentially optimistic results for some pretrained comparators with overlapping Rfam families.","source_ids":["debfold-2024"],"source_locator":"Methods and Data Sets: Family-Wise Processed RNA Structure Ground-Truth Data Set, including all three test-set construction subsections"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: RNA sequences.","Evaluation: Family-wise training/validation/test partitioning with cross-validation inside the training process.","Readout: Base-pair precision, recall and structure F1."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["debfold-2024"],"source_locator":"Methods: Family-Wise Processed RNA Structure Ground-Truth Data Set; Contamination-Free Family-Wise Independent Test Set; Structure Prediction Evaluation Metrics; cached text lines 26–31, 36–37"},"coverage":"limited","gaps":["Uncertainty: The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","Organisms: The evaluated RNA sets are selected by Rfam family, length and source rather than by organism. The dataset-construction sections identify bpRNA/Rfam and PDB origins but do not enumerate the organism composition of TestSet α, β and γ."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Targeted full-paper and supplement review of the outstanding task fields, with original dataset metadata checked where accessible. Source-scoped omissions are explicit; no independent benchmark reproduction or numerical-result change."}}}}
{"id":"reported-task-031186b57c62de","kind":"benchmark","name":"Human thymus cell-type classification","description":"","status":"needs_review","facets":{"areas":["cells-tissues"],"tasks":["Human thymus cell-type classification"]},"source_ids":["mouse-geneformer-2025"],"links":[{"relation":"dataset","target_id":"reported-dataset-477a9082515406"}],"attributes":{"entity_level":"task","version":null,"task":"Human thymus cell-type classification","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Human thymus annotation evaluates transfer from a mouse-trained gene representation after homolog mapping.","summary_source_ids":["mouse-geneformer-2025"],"summary_source_locator":"Results: cross-species human cell classification; cached text line 41","sections":[{"title":"Evaluation methodology","body":"Human thymus scRNA-seq data from GSE144870, alongside separate breast and cortex evaluations. Human thymus cells come from GSE144870. After mapping human genes to mouse homologues, the fine-tuning experiment randomly uses 80% of cells for training and 20% for testing; the zero-shot column is a separate evaluation setting. Cell-type classification accuracy. Mouse-Geneformer and human Geneformer. The source states these datasets are outside Genecorpus-30M; this review does not treat its feature-distance check as proof of complete leakage absence.","source_ids":["mouse-geneformer-2025"],"source_locator":"Results: cross-species human cell classification; cached text line 41; Materials and methods: Human cell type classification using mouse-Geneformer; Table 4"}],"facts":[{"label":"Datasets","value":"Human thymus scRNA-seq data from GSE144870, alongside separate breast and cortex evaluations.","status":"source_checked","source_ids":["mouse-geneformer-2025"],"source_locator":"Results: cross-species human cell classification; cached text line 41"},{"label":"Splits","value":"Human thymus cells come from GSE144870. After mapping human genes to mouse homologues, the fine-tuning experiment randomly uses 80% of cells for training and 20% for testing; the zero-shot column is a separate evaluation setting.","status":"source_checked","source_ids":["mouse-geneformer-2025"],"source_locator":"Materials and methods: Human cell type classification using mouse-Geneformer; Table 4"},{"label":"Metrics","value":"Cell-type classification accuracy.","status":"source_checked","source_ids":["mouse-geneformer-2025"],"source_locator":"Results: cross-species human cell classification; cached text line 41"},{"label":"Baselines","value":"Mouse-Geneformer and human Geneformer.","status":"source_checked","source_ids":["mouse-geneformer-2025"],"source_locator":"Results: cross-species human cell classification; cached text line 41"},{"label":"Leakage controls","value":"The source states these datasets are outside Genecorpus-30M; this review does not treat its feature-distance check as proof of complete leakage absence.","status":"source_checked","source_ids":["mouse-geneformer-2025"],"source_locator":"Results: cross-species human cell classification; cached text line 41"},{"label":"Uncertainty","value":"Table 4 prints one accuracy and F1 value per model/setting. Its caption and the human classification method do not report confidence intervals, standard deviations or repeated-split uncertainty for the thymus result.","status":"unreported","source_ids":["mouse-geneformer-2025"],"source_locator":"Materials and methods: Human cell type classification using mouse-Geneformer; Table 4 caption"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["mouse-geneformer-2025"],"source_locator":"Results: cross-species human cell classification; cached text line 41"},{"label":"Organisms","value":"Human thymus for the linked test; mouse pretraining is a separate data source.","status":"source_checked","source_ids":["mouse-geneformer-2025"],"source_locator":"Results: cross-species human cell classification; cached text line 41"},{"label":"Assays","value":"Human thymus scRNA-seq cell-type annotations.","status":"source_checked","source_ids":["mouse-geneformer-2025"],"source_locator":"Results: cross-species human cell classification; cached text line 41"},{"label":"Allowed inputs","value":"Gene-expression profiles represented by Geneformer models.","status":"source_checked","source_ids":["mouse-geneformer-2025"],"source_locator":"Results: cross-species human cell classification; cached text line 41"},{"label":"Adaptation","value":"Cross-species representation transfer compared with the human Geneformer baseline.","status":"source_checked","source_ids":["mouse-geneformer-2025"],"source_locator":"Results: cross-species human cell classification; cached text line 41"}],"strengths":[],"limitations":[{"text":"The fine-tuning experiment randomly partitions cells; it does not establish donor-held-out transfer. Mapping human genes to mouse homologues changes the available feature set.","source_ids":["mouse-geneformer-2025"],"source_locator":"Materials and methods: Human cell type classification using mouse-Geneformer; Table 4; Materials and methods: Human cell type classification using mouse-Geneformer; Table 4 caption"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Gene-expression profiles represented by Geneformer models.","Evaluation: Human thymus cells come from GSE144870. After mapping human genes to mouse homologues, the fine-tuning experiment randomly uses 80% of cells for training and 20% for testing; the zero-shot column is a separate evaluation setting.","Readout: Cell-type classification accuracy."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["mouse-geneformer-2025"],"source_locator":"Results: cross-species human cell classification; cached text line 41; Materials and methods: Human cell type classification using mouse-Geneformer; Table 4"},"coverage":"limited","gaps":["Uncertainty: Table 4 prints one accuracy and F1 value per model/setting. Its caption and the human classification method do not report confidence intervals, standard deviations or repeated-split uncertainty for the thymus result."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Targeted full-paper and supplement review of the outstanding task fields, with original dataset metadata checked where accessible. Source-scoped omissions are explicit; no independent benchmark reproduction or numerical-result change."}}}}
{"id":"reported-task-0647b0364def8f","kind":"benchmark","name":"antibody deamidation-site prediction","description":"","status":"needs_review","facets":{"areas":["proteins-complexes"],"tasks":["antibody deamidation-site prediction"]},"source_ids":["antibody-deamidation-plm-2024"],"links":[{"relation":"dataset","target_id":"reported-dataset-0edd8f724db696"}],"attributes":{"entity_level":"task","version":null,"task":"antibody deamidation-site prediction","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Antibody deamidation-site classification is assessed with class-sensitive metrics and an independent antibody dataset.","summary_source_ids":["antibody-deamidation-plm-2024"],"summary_source_locator":"Results §§3.4–3.5; cached text lines 42–50","sections":[{"title":"Evaluation methodology","body":"Labeled active and inactive sites, including a separate six-antibody evaluation collection. Five-fold stratified cross-validation on training data, then evaluation on the independent collection. Accuracy, precision, recall, specificity, F1, MCC and ROC-AUC; the source explicitly cautions that accuracy alone hides class imbalance. Global-embedding and local-sequence ablations; published decision-tree/random-forest models and NGOME. Independent testing withholds complete antibodies from the training collection. The source does not establish whether within-training cross-validation also groups every site by antibody.","source_ids":["antibody-deamidation-plm-2024"],"source_locator":"Results §§3.4–3.5; cached text lines 42–50"}],"facts":[{"label":"Datasets","value":"Labeled active and inactive sites, including a separate six-antibody evaluation collection.","status":"source_checked","source_ids":["antibody-deamidation-plm-2024"],"source_locator":"Results §§3.4–3.5; cached text lines 42–50"},{"label":"Splits","value":"Five-fold stratified cross-validation on training data, then evaluation on the independent collection.","status":"source_checked","source_ids":["antibody-deamidation-plm-2024"],"source_locator":"Results §§3.4–3.5; cached text lines 42–50"},{"label":"Metrics","value":"Accuracy, precision, recall, specificity, F1, MCC and ROC-AUC; the source explicitly cautions that accuracy alone hides class imbalance.","status":"source_checked","source_ids":["antibody-deamidation-plm-2024"],"source_locator":"Results §§3.4–3.5; cached text lines 42–50"},{"label":"Baselines","value":"Global-embedding and local-sequence ablations; published decision-tree/random-forest models and NGOME.","status":"source_checked","source_ids":["antibody-deamidation-plm-2024"],"source_locator":"Results §§3.4–3.5; cached text lines 42–50"},{"label":"Leakage controls","value":"Independent testing withholds complete antibodies from the training collection. The source does not establish whether within-training cross-validation also groups every site by antibody.","status":"source_checked","source_ids":["antibody-deamidation-plm-2024"],"source_locator":"Results §§3.4–3.5; cached text lines 42–50"},{"label":"Uncertainty","value":"Table 1 gives values with ± terms for five-fold stratified cross-validation, whereas Table 2 gives point values for the independent test. The table caption does not define the ± terms as a standard deviation, standard error or confidence interval; that interpretation remains unreported.","status":"source_checked","source_ids":["antibody-deamidation-plm-2024"],"source_locator":"Model-performance discussion and Tables 1–2"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["antibody-deamidation-plm-2024"],"source_locator":"Results §§3.4–3.5; cached text lines 42–50"},{"label":"Organisms","value":"NISTmAb is a humanized IgG1 antibody. The other antibodies were produced in Chinese hamster ovary cells; this expression host should not be mistaken for their sequence species. The complete sequence-origin composition of the proprietary antibody panel is not reported in Methods 2.1–2.2.","status":"source_checked","source_ids":["antibody-deamidation-plm-2024"],"source_locator":"Methods 2.1 Chemicals and Reagents; 2.2 Accelerated Thermal Stress"},{"label":"Assays","value":"Labeled antibody deamidation sites.","status":"source_checked","source_ids":["antibody-deamidation-plm-2024"],"source_locator":"Results §§3.4–3.5; cached text lines 42–50"},{"label":"Allowed inputs","value":"Antibody sequences and candidate residue positions.","status":"source_checked","source_ids":["antibody-deamidation-plm-2024"],"source_locator":"Results §§3.4–3.5; cached text lines 42–50"},{"label":"Adaptation","value":"Supervised classifier with stratified cross-validation and an independent antibody evaluation.","status":"source_checked","source_ids":["antibody-deamidation-plm-2024"],"source_locator":"Results §§3.4–3.5; cached text lines 42–50"}],"strengths":[],"limitations":[{"text":"Sites within one antibody may be correlated. Five-fold stratification alone does not establish antibody-disjoint evaluation, and expression host does not identify an antibody’s sequence species.","source_ids":["antibody-deamidation-plm-2024"],"source_locator":"Model-performance discussion and Tables 1–2; Methods 2.1 Chemicals and Reagents; 2.2 Accelerated Thermal Stress"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Antibody sequences and candidate residue positions.","Evaluation: Five-fold stratified cross-validation on training data, then evaluation on the independent collection.","Readout: Accuracy, precision, recall, specificity, F1, MCC and ROC-AUC; the source explicitly cautions that accuracy alone hides class imbalance."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["antibody-deamidation-plm-2024"],"source_locator":"Results §§3.4–3.5; cached text lines 42–50"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Targeted full-paper and supplement review of the outstanding task fields, with original dataset metadata checked where accessible. Source-scoped omissions are explicit; no independent benchmark reproduction or numerical-result change."}}}}
{"id":"reported-task-09c3100b77dcc5","kind":"benchmark","name":"protein-protein interaction prediction","description":"","status":"needs_review","facets":{"areas":["molecular-interactions"],"tasks":["protein-protein interaction prediction"]},"source_ids":["esm2-amp-2025"],"links":[{"relation":"dataset","target_id":"reported-dataset-9135087a16af1c"}],"attributes":{"entity_level":"task","version":null,"task":"protein-protein interaction prediction","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Protein-interaction classification contrasts ordinary pair-level validation with a more stringent protein-partitioned benchmark.","summary_source_ids":["esm2-amp-2025"],"summary_source_locator":"Methods: Protein–protein interaction datasets; Results: strictly constructed dataset; cached text lines 10–12, 58–59, 67–68","sections":[{"title":"Evaluation methodology","body":"Pan_dataset and an independent real_test collection; Bernett data provide an additional evaluation. Five-fold cross-validation on Pan_dataset; separate Bernett training/validation/test partitions. AUC and accuracy are among the reported metrics. The Bernett comparison includes performance values supplied by previous studies. The real_test set filters high-similarity proteins; Bernett uses graph partitioning and within-partition redundancy filtering.","source_ids":["esm2-amp-2025"],"source_locator":"Methods: Protein–protein interaction datasets; Results: strictly constructed dataset; cached text lines 10–12, 58–59, 67–68"}],"facts":[{"label":"Datasets","value":"Pan_dataset and an independent real_test collection; Bernett data provide an additional evaluation.","status":"source_checked","source_ids":["esm2-amp-2025"],"source_locator":"Methods: Protein–protein interaction datasets; Results: strictly constructed dataset; cached text lines 10–12, 58–59, 67–68"},{"label":"Splits","value":"Five-fold cross-validation on Pan_dataset; separate Bernett training/validation/test partitions.","status":"source_checked","source_ids":["esm2-amp-2025"],"source_locator":"Methods: Protein–protein interaction datasets; Results: strictly constructed dataset; cached text lines 10–12, 58–59, 67–68"},{"label":"Metrics","value":"AUC and accuracy are among the reported metrics.","status":"source_checked","source_ids":["esm2-amp-2025"],"source_locator":"Methods: Protein–protein interaction datasets; Results: strictly constructed dataset; cached text lines 10–12, 58–59, 67–68"},{"label":"Baselines","value":"The Bernett comparison includes performance values supplied by previous studies.","status":"source_checked","source_ids":["esm2-amp-2025"],"source_locator":"Methods: Protein–protein interaction datasets; Results: strictly constructed dataset; cached text lines 10–12, 58–59, 67–68"},{"label":"Leakage controls","value":"The real_test set filters high-similarity proteins; Bernett uses graph partitioning and within-partition redundancy filtering.","status":"source_checked","source_ids":["esm2-amp-2025"],"source_locator":"Methods: Protein–protein interaction datasets; Results: strictly constructed dataset; cached text lines 10–12, 58–59, 67–68"},{"label":"Uncertainty","value":"The Pan_dataset experiment averages five cross-validation folds. Tables 2 and 4 present point metrics without score confidence intervals; quoted Bernett comparator values are not additional independent repetitions.","status":"source_checked","source_ids":["esm2-amp-2025"],"source_locator":"Methods: Protein–protein interaction datasets; Results and Tables 2, 4"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["esm2-amp-2025"],"source_locator":"Methods: Protein–protein interaction datasets; Results: strictly constructed dataset; cached text lines 10–12, 58–59, 67–68"},{"label":"Organisms","value":"Pan_dataset, real_test and Bernett are human interaction datasets. The separate Multi_species collection contains C. elegans, D. melanogaster and E. coli; performance on this collection must retain its multi-species context.","status":"source_checked","source_ids":["esm2-amp-2025"],"source_locator":"Methods: Protein–protein interaction datasets; Table 1"},{"label":"Assays","value":"Protein-pair interaction labels.","status":"source_checked","source_ids":["esm2-amp-2025"],"source_locator":"Methods: Protein–protein interaction datasets; Results: strictly constructed dataset; cached text lines 10–12, 58–59, 67–68"},{"label":"Allowed inputs","value":"Pairs of protein sequences represented using ESM2.","status":"source_checked","source_ids":["esm2-amp-2025"],"source_locator":"Methods: Protein–protein interaction datasets; Results: strictly constructed dataset; cached text lines 10–12, 58–59, 67–68"},{"label":"Adaptation","value":"Supervised pair classification with cross-validation and separate independent tests.","status":"source_checked","source_ids":["esm2-amp-2025"],"source_locator":"Methods: Protein–protein interaction datasets; Results: strictly constructed dataset; cached text lines 10–12, 58–59, 67–68"}],"strengths":[],"limitations":[{"text":"Pair-level cross-validation and protein-disjoint partitioning answer different generalization questions. Literature comparator scores retain their original execution provenance.","source_ids":["esm2-amp-2025"],"source_locator":"Methods: Protein–protein interaction datasets; Results: strictly constructed dataset; cached text lines 10–12, 58–59, 67–68"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Pairs of protein sequences represented using ESM2.","Evaluation: Five-fold cross-validation on Pan_dataset; separate Bernett training/validation/test partitions.","Readout: AUC and accuracy are among the reported metrics."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["esm2-amp-2025"],"source_locator":"Methods: Protein–protein interaction datasets; Results: strictly constructed dataset; cached text lines 10–12, 58–59, 67–68"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Targeted full-paper and supplement review of the outstanding task fields, with original dataset metadata checked where accessible. Source-scoped omissions are explicit; no independent benchmark reproduction or numerical-result change."}}}}
{"id":"reported-task-0c92cda11228c4","kind":"benchmark","name":"Antibody–antigen interaction prediction using folded complexes","description":"","status":"needs_review","facets":{"areas":["molecular-interactions"],"tasks":["Antibody–antigen interaction prediction using folded complexes"]},"source_ids":["antibody-flexibility-2025"],"links":[{"relation":"dataset","target_id":"reported-dataset-ee26acbd6e8cf7"}],"attributes":{"entity_level":"task","version":null,"task":"Antibody–antigen interaction prediction using folded complexes","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Antibody–antigen interaction evaluation uses a curated SAbDab collection alongside a separate protein-interaction dataset.","summary_source_ids":["antibody-flexibility-2025"],"summary_source_locator":"Methods: Dataset; cached text lines 72–76; comparative evaluation and ablation passages","sections":[{"title":"Evaluation methodology","body":"SAbDab antibody–antigen complexes and the dMaSIF protein–protein interaction dataset. The antibody–antigen collection is randomly partitioned with a validation subset and separate test complexes. ROC-AUC for the interaction task, comparing original and folded complex inputs; paratope/epitope metrics are separate evaluations. The interaction predictor is evaluated on original structures and complexes folded by AlphaFold3, Chai-1 and Boltz-1; these are alternative input-structure conditions. The source describes sequence deduplication and structural-similarity filtering before the random split.","source_ids":["antibody-flexibility-2025"],"source_locator":"Methods: Dataset; cached text lines 72–76; comparative evaluation and ablation passages"}],"facts":[{"label":"Datasets","value":"SAbDab antibody–antigen complexes and the dMaSIF protein–protein interaction dataset.","status":"source_checked","source_ids":["antibody-flexibility-2025"],"source_locator":"Methods: Dataset; cached text lines 72–76; comparative evaluation and ablation passages"},{"label":"Splits","value":"The antibody–antigen collection is randomly partitioned with a validation subset and separate test complexes.","status":"source_checked","source_ids":["antibody-flexibility-2025"],"source_locator":"Methods: Dataset; cached text lines 72–76; comparative evaluation and ablation passages"},{"label":"Metrics","value":"ROC-AUC for the interaction task, comparing original and folded complex inputs; paratope/epitope metrics are separate evaluations.","status":"source_checked","source_ids":["antibody-flexibility-2025"],"source_locator":"Methods: Dataset; cached text lines 72–76; comparative evaluation and ablation passages"},{"label":"Baselines","value":"The interaction predictor is evaluated on original structures and complexes folded by AlphaFold3, Chai-1 and Boltz-1; these are alternative input-structure conditions.","status":"source_checked","source_ids":["antibody-flexibility-2025"],"source_locator":"Methods: Dataset; cached text lines 72–76; comparative evaluation and ablation passages"},{"label":"Leakage controls","value":"The source describes sequence deduplication and structural-similarity filtering before the random split.","status":"source_checked","source_ids":["antibody-flexibility-2025"],"source_locator":"Methods: Dataset; cached text lines 72–76; comparative evaluation and ablation passages"},{"label":"Uncertainty","value":"Table 5 reports folded-complex interaction AUC-ROC with ± terms, without defining their estimator in its caption. Tables 3–4 summarize folding RMSD and model confidence outputs separately; pLDDT, PAE and pTM are not confidence intervals on comparative interaction performance.","status":"source_checked","source_ids":["antibody-flexibility-2025"],"source_locator":"Tables 3–5 captions; Methods: Dataset and model evaluation"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["antibody-flexibility-2025"],"source_locator":"Methods: Dataset; cached text lines 72–76; comparative evaluation and ablation passages"},{"label":"Organisms","value":"The folding comparison uses antibody–antigen complexes selected from SAbDab. The dataset-construction text identifies structural and redundancy filters but does not provide a species census for the evaluated complex subset. Antibody and antigen origins must be distinguished.","status":"unreported","source_ids":["antibody-flexibility-2025"],"source_locator":"Methods: Dataset; Table 6"},{"label":"Assays","value":"Experimentally determined complex structures used for interaction assessment.","status":"source_checked","source_ids":["antibody-flexibility-2025"],"source_locator":"Methods: Dataset; cached text lines 72–76; comparative evaluation and ablation passages"},{"label":"Allowed inputs","value":"Folded antibody–antigen complexes.","status":"source_checked","source_ids":["antibody-flexibility-2025"],"source_locator":"Methods: Dataset; cached text lines 72–76; comparative evaluation and ablation passages"},{"label":"Adaptation","value":"Task training uses the curated complex training partition.","status":"source_checked","source_ids":["antibody-flexibility-2025"],"source_locator":"Methods: Dataset; cached text lines 72–76; comparative evaluation and ablation passages"}],"strengths":[],"limitations":[{"text":"The paratope/epitope classification and folded-complex interaction evaluations have different outcomes. Their thresholds and splits cannot be substituted for one another; predicted structure confidence is not benchmark uncertainty.","source_ids":["antibody-flexibility-2025"],"source_locator":"Tables 3–5 captions; Methods: Dataset and model evaluation; Methods: Dataset; Table 6"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Folded antibody–antigen complexes.","Evaluation: The antibody–antigen collection is randomly partitioned with a validation subset and separate test complexes.","Readout: ROC-AUC for the interaction task, comparing original and folded complex inputs; paratope/epitope metrics are separate evaluations."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["antibody-flexibility-2025"],"source_locator":"Methods: Dataset; cached text lines 72–76; comparative evaluation and ablation passages"},"coverage":"limited","gaps":["Organisms: The folding comparison uses antibody–antigen complexes selected from SAbDab. The dataset-construction text identifies structural and redundancy filters but does not provide a species census for the evaluated complex subset. Antibody and antigen origins must be distinguished."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Targeted full-paper and supplement review of the outstanding task fields, with original dataset metadata checked where accessible. Source-scoped omissions are explicit; no independent benchmark reproduction or numerical-result change."}}}}
{"id":"reported-task-132da895d4c381","kind":"benchmark","name":"Enhancer classification","description":"","status":"needs_review","facets":{"areas":["dna-genomes"],"tasks":["Enhancer classification"]},"source_ids":["enbed-2024"],"links":[{"relation":"dataset","target_id":"reported-dataset-f0bf60a62ad7c3"}],"attributes":{"entity_level":"task","version":null,"task":"Enhancer classification","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Enhancer classification is evaluated within two published genomic sequence benchmark collections.","summary_source_ids":["enbed-2024"],"summary_source_locator":"Results §1.2.1; Methods §§2.5.1–2.5.2; cached text lines 18–19, 70–74; matching task comparison table/ablation captions","sections":[{"title":"Evaluation methodology","body":"Genomic Benchmarks includes human enhancer datasets; Nucleotide Transformer tasks include enhancer/non-enhancer and enhancer-strength labels. For the Nucleotide Transformer enhancer tasks, Supplementary Table 1 lists 14,968 training sequences and 400 test sequences of length 200. This set combines the original strong/weak/non-enhancer collection with 6,000 synthetic enhancers and 6,000 synthetic non-enhancers. Those counts do not describe the separate Genomic Benchmarks enhancer datasets. Classification accuracy is described for the benchmark comparison. Enformer, DNABERT-2, Nucleotide Transformer and HyenaDNA in the enhancer-containing NT task comparison. The enhancer-specific methods and supplement do not specify a chromosome holdout, homology threshold or synthetic-parent separation across these splits. The clade split and 95% similarity threshold described later in the supplement apply to mutation generation, not this enhancer evaluation.","source_ids":["enbed-2024","evidence-task-final-a-enbed-2024-vbae117-supplementary-data-pdf"],"source_locator":"Results §1.2.1; Methods §§2.5.1–2.5.2; cached text lines 18–19, 70–74; matching task comparison table/ablation captions; Supplement: Nucleotide Transformer benchmark dataset description and Table 1; Supplement: Nucleotide Transformer benchmark description and Table 1; separate Mutation generation section"}],"facts":[{"label":"Datasets","value":"Genomic Benchmarks includes human enhancer datasets; Nucleotide Transformer tasks include enhancer/non-enhancer and enhancer-strength labels.","status":"source_checked","source_ids":["enbed-2024"],"source_locator":"Results §1.2.1; Methods §§2.5.1–2.5.2; cached text lines 18–19, 70–74; matching task comparison table/ablation captions"},{"label":"Splits","value":"For the Nucleotide Transformer enhancer tasks, Supplementary Table 1 lists 14,968 training sequences and 400 test sequences of length 200. This set combines the original strong/weak/non-enhancer collection with 6,000 synthetic enhancers and 6,000 synthetic non-enhancers. Those counts do not describe the separate Genomic Benchmarks enhancer datasets.","status":"source_checked","source_ids":["enbed-2024","evidence-task-final-a-enbed-2024-vbae117-supplementary-data-pdf"],"source_locator":"Supplement: Nucleotide Transformer benchmark dataset description and Table 1"},{"label":"Metrics","value":"Classification accuracy is described for the benchmark comparison.","status":"source_checked","source_ids":["enbed-2024"],"source_locator":"Results §1.2.1; Methods §§2.5.1–2.5.2; cached text lines 18–19, 70–74; matching task comparison table/ablation captions"},{"label":"Baselines","value":"Enformer, DNABERT-2, Nucleotide Transformer and HyenaDNA in the enhancer-containing NT task comparison.","status":"source_checked","source_ids":["enbed-2024"],"source_locator":"Results §1.2.1; Methods §§2.5.1–2.5.2; cached text lines 18–19, 70–74; matching task comparison table/ablation captions"},{"label":"Leakage controls","value":"The enhancer-specific methods and supplement do not specify a chromosome holdout, homology threshold or synthetic-parent separation across these splits. The clade split and 95% similarity threshold described later in the supplement apply to mutation generation, not this enhancer evaluation.","status":"unreported","source_ids":["enbed-2024","evidence-task-final-a-enbed-2024-vbae117-supplementary-data-pdf"],"source_locator":"Supplement: Nucleotide Transformer benchmark description and Table 1; separate Mutation generation section"},{"label":"Uncertainty","value":"The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","status":"unreported","source_ids":["enbed-2024"],"source_locator":"Results §1.2.1; Methods §§2.5.1–2.5.2; cached text lines 18–19, 70–74; matching task comparison table/ablation captions"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["enbed-2024"],"source_locator":"Results §1.2.1; Methods §§2.5.1–2.5.2; cached text lines 18–19, 70–74; matching task comparison table/ablation captions"},{"label":"Organisms","value":"Human enhancer datasets within Genomic Benchmarks and Nucleotide Transformer tasks.","status":"source_checked","source_ids":["enbed-2024"],"source_locator":"Results §1.2.1; Methods §§2.5.1–2.5.2; cached text lines 18–19, 70–74; matching task comparison table/ablation captions"},{"label":"Assays","value":"Enhancer identity/strength annotations.","status":"source_checked","source_ids":["enbed-2024"],"source_locator":"Results §1.2.1; Methods §§2.5.1–2.5.2; cached text lines 18–19, 70–74; matching task comparison table/ablation captions"},{"label":"Allowed inputs","value":"Genomic sequence.","status":"source_checked","source_ids":["enbed-2024"],"source_locator":"Results §1.2.1; Methods §§2.5.1–2.5.2; cached text lines 18–19, 70–74; matching task comparison table/ablation captions"},{"label":"Adaptation","value":"Task-level enhancer classification; exact task configuration must be selected from the evaluated benchmark collection.","status":"source_checked","source_ids":["enbed-2024"],"source_locator":"Results §1.2.1; Methods §§2.5.1–2.5.2; cached text lines 18–19, 70–74; matching task comparison table/ablation captions"}],"strengths":[],"limitations":[{"text":"ENBED reports separate enhancer identification and enhancer-type tasks. Their splits and synthetic augmentation must stay distinct from the mutation-generation experiment’s phylogenetic holdouts.","source_ids":["enbed-2024","evidence-task-final-a-enbed-2024-vbae117-supplementary-data-pdf"],"source_locator":"Supplement: Nucleotide Transformer benchmark dataset description and Table 1; Supplement: Nucleotide Transformer benchmark description and Table 1; separate Mutation generation section"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Genomic sequence.","Evaluation: For the Nucleotide Transformer enhancer tasks, Supplementary Table 1 lists 14,968 training sequences and 400 test sequences of length 200. This set combines the original strong/weak/non-enhancer collection with 6,000 synthetic enhancers and 6,000 synthetic non-enhancers. Those counts do not describe the separate Genomic Benchmarks enhancer datasets.","Readout: Classification accuracy is described for the benchmark comparison."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["enbed-2024","evidence-task-final-a-enbed-2024-vbae117-supplementary-data-pdf"],"source_locator":"Results §1.2.1; Methods §§2.5.1–2.5.2; cached text lines 18–19, 70–74; matching task comparison table/ablation captions; Supplement: Nucleotide Transformer benchmark dataset description and Table 1"},"coverage":"limited","gaps":["Leakage controls: The enhancer-specific methods and supplement do not specify a chromosome holdout, homology threshold or synthetic-parent separation across these splits. The clade split and 95% similarity threshold described later in the supplement apply to mutation generation, not this enhancer evaluation.","Uncertainty: The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Targeted full-paper and supplement review of the outstanding task fields, with original dataset metadata checked where accessible. Source-scoped omissions are explicit; no independent benchmark reproduction or numerical-result change."}}}}
{"id":"reported-task-13dfe6b33e71ed","kind":"benchmark","name":"polyadenylation site detection","description":"","status":"needs_review","facets":{"areas":["dna-genomes"],"tasks":["polyadenylation site detection"]},"source_ids":["polya-glm-2025"],"links":[{"relation":"dataset","target_id":"reported-dataset-55f200c9481409"}],"attributes":{"entity_level":"task","version":null,"task":"polyadenylation site detection","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Polyadenylation-site classification compares two background definitions and two adaptation regimes.","summary_source_ids":["polya-glm-2025"],"summary_source_locator":"Methods: dataset construction and evaluation; Results Table 1; cached text lines 12–15, 37, 49–51","sections":[{"title":"Evaluation methodology","body":"GENCODE v38 human annotations with balanced Gene-Gene and Gene-Intergene datasets; a separate mouse transfer evaluation. Five-fold evaluation with 60:20:20 training/validation/test proportions per fold. Accuracy, precision, recall, F1 and AUC, averaged across folds. DNABERT-2, Nucleotide Transformer and HyenaDNA under few-shot and fine-tuned conditions. Negative examples exclude genic regions overlapping GENCODE v38 polyadenylation annotations. This is a label-construction control; the evaluation description does not establish chromosome or homologous-sequence separation between the five-fold training and test partitions.","source_ids":["polya-glm-2025"],"source_locator":"Methods: dataset construction and evaluation; Results Table 1; cached text lines 12–15, 37, 49–51; Methods: dataset construction and evaluation framework"}],"facts":[{"label":"Datasets","value":"GENCODE v38 human annotations with balanced Gene-Gene and Gene-Intergene datasets; a separate mouse transfer evaluation.","status":"source_checked","source_ids":["polya-glm-2025"],"source_locator":"Methods: dataset construction and evaluation; Results Table 1; cached text lines 12–15, 37, 49–51"},{"label":"Splits","value":"Five-fold evaluation with 60:20:20 training/validation/test proportions per fold.","status":"source_checked","source_ids":["polya-glm-2025"],"source_locator":"Methods: dataset construction and evaluation; Results Table 1; cached text lines 12–15, 37, 49–51"},{"label":"Metrics","value":"Accuracy, precision, recall, F1 and AUC, averaged across folds.","status":"source_checked","source_ids":["polya-glm-2025"],"source_locator":"Methods: dataset construction and evaluation; Results Table 1; cached text lines 12–15, 37, 49–51"},{"label":"Baselines","value":"DNABERT-2, Nucleotide Transformer and HyenaDNA under few-shot and fine-tuned conditions.","status":"source_checked","source_ids":["polya-glm-2025"],"source_locator":"Methods: dataset construction and evaluation; Results Table 1; cached text lines 12–15, 37, 49–51"},{"label":"Leakage controls","value":"Negative examples exclude genic regions overlapping GENCODE v38 polyadenylation annotations. This is a label-construction control; the evaluation description does not establish chromosome or homologous-sequence separation between the five-fold training and test partitions.","status":"source_checked","source_ids":["polya-glm-2025"],"source_locator":"Methods: dataset construction and evaluation framework"},{"label":"Uncertainty","value":"Table 1 reports averages over five folds. For few-shot predictions, ten random prototype pairs per fold are averaged. No confidence interval or standard-error estimator for the table’s aggregate metrics is specified there.","status":"source_checked","source_ids":["polya-glm-2025"],"source_locator":"Methods: few-shot evaluation; Table 1 caption"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["polya-glm-2025"],"source_locator":"Methods: dataset construction and evaluation; Results Table 1; cached text lines 12–15, 37, 49–51"},{"label":"Organisms","value":"Human; separate transfer evaluation in mouse.","status":"source_checked","source_ids":["polya-glm-2025"],"source_locator":"Methods: dataset construction and evaluation; Results Table 1; cached text lines 12–15, 37, 49–51"},{"label":"Assays","value":"GENCODE polyadenylation annotations.","status":"source_checked","source_ids":["polya-glm-2025"],"source_locator":"Methods: dataset construction and evaluation; Results Table 1; cached text lines 12–15, 37, 49–51"},{"label":"Allowed inputs","value":"DNA/RNA sequence context for candidate polyadenylation sites.","status":"source_checked","source_ids":["polya-glm-2025"],"source_locator":"Methods: dataset construction and evaluation; Results Table 1; cached text lines 12–15, 37, 49–51"},{"label":"Adaptation","value":"Few-shot and task fine-tuning settings are compared separately.","status":"source_checked","source_ids":["polya-glm-2025"],"source_locator":"Methods: dataset construction and evaluation; Results Table 1; cached text lines 12–15, 37, 49–51"}],"strengths":[],"limitations":[{"text":"Changing the negative-region definition changes the classification problem. Negative-label overlap exclusion alone does not establish chromosome- or homology-disjoint evaluation.","source_ids":["polya-glm-2025"],"source_locator":"Methods: dataset construction and evaluation framework; Methods: few-shot evaluation; Table 1 caption"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: DNA/RNA sequence context for candidate polyadenylation sites.","Evaluation: Five-fold evaluation with 60:20:20 training/validation/test proportions per fold.","Readout: Accuracy, precision, recall, F1 and AUC, averaged across folds."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["polya-glm-2025"],"source_locator":"Methods: dataset construction and evaluation; Results Table 1; cached text lines 12–15, 37, 49–51"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Targeted full-paper and supplement review of the outstanding task fields, with original dataset metadata checked where accessible. Source-scoped omissions are explicit; no independent benchmark reproduction or numerical-result change."}}}}
{"id":"reported-task-167f08013c270e","kind":"benchmark","name":"Cross-dataset single-cell drug response transfer","description":"","status":"needs_review","facets":{"areas":["cells-tissues"],"tasks":["Cross-dataset single-cell drug response transfer"]},"source_ids":["scxdr-2026"],"links":[{"relation":"dataset","target_id":"reported-dataset-f7210686a78474"}],"attributes":{"entity_level":"task","version":null,"task":"Cross-dataset single-cell drug response transfer","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Cross-dataset response prediction evaluates transfer across batches, drugs, tumors and patients.","summary_source_ids":["scxdr-2026"],"summary_source_locator":"Methods: Data collection and processing; group-level evaluation; cached text lines 55–57, 84–86; uncertainty/repeat-run/statistical-comparison passages; matching task comparison table/ablation captions","sections":[{"title":"Evaluation methodology","body":"Twelve scRNA-seq datasets organized into four transfer scenarios and multiple source-to-target tasks. The paper defines 20 cross-dataset transfers across batch, drug, tumor and patient scenarios. Comparator protocols differ: scVI/ComBat combine corrected datasets before a train/test partition, whereas transfer methods retain source/target roles. The exact per-task split must accompany any comparison. Cell-group accuracy averages within-cluster prediction accuracy; a separate malignant-cell-group accuracy is also described. Bulk-transfer scDEAL, SCAD, CaDRReS-Sc and DREEP plus conventional MLP/SVM/correlation references; single-cell transfer comparators form another comparison group. Training and evaluation transfer between datasets, drugs, tumors or patients as listed in Supplementary Table S2. Table S11 excludes drug–cell relations from the representation graph’s listed relation types. These controls do not by themselves establish an audit of target-label use during every adaptation or hyperparameter-selection step.","source_ids":["scxdr-2026","evidence-task-final-a-scxdr-2026-42003-2025-9418-moesm1-esm-pdf"],"source_locator":"Methods: Data collection and processing; group-level evaluation; cached text lines 55–57, 84–86; uncertainty/repeat-run/statistical-comparison passages; matching task comparison table/ablation captions; Methods: Data collection and processing; Comparison experiments at the individual cell level; Supplementary Table S2; Supplementary Tables S2 and S11"}],"facts":[{"label":"Datasets","value":"Twelve scRNA-seq datasets organized into four transfer scenarios and multiple source-to-target tasks.","status":"source_checked","source_ids":["scxdr-2026"],"source_locator":"Methods: Data collection and processing; group-level evaluation; cached text lines 55–57, 84–86; uncertainty/repeat-run/statistical-comparison passages; matching task comparison table/ablation captions"},{"label":"Splits","value":"The paper defines 20 cross-dataset transfers across batch, drug, tumor and patient scenarios. Comparator protocols differ: scVI/ComBat combine corrected datasets before a train/test partition, whereas transfer methods retain source/target roles. The exact per-task split must accompany any comparison.","status":"source_checked","source_ids":["scxdr-2026"],"source_locator":"Methods: Data collection and processing; Comparison experiments at the individual cell level; Supplementary Table S2"},{"label":"Metrics","value":"Cell-group accuracy averages within-cluster prediction accuracy; a separate malignant-cell-group accuracy is also described.","status":"source_checked","source_ids":["scxdr-2026"],"source_locator":"Methods: Data collection and processing; group-level evaluation; cached text lines 55–57, 84–86; uncertainty/repeat-run/statistical-comparison passages; matching task comparison table/ablation captions"},{"label":"Baselines","value":"Bulk-transfer scDEAL, SCAD, CaDRReS-Sc and DREEP plus conventional MLP/SVM/correlation references; single-cell transfer comparators form another comparison group.","status":"source_checked","source_ids":["scxdr-2026"],"source_locator":"Methods: Data collection and processing; group-level evaluation; cached text lines 55–57, 84–86; uncertainty/repeat-run/statistical-comparison passages; matching task comparison table/ablation captions"},{"label":"Leakage controls","value":"Training and evaluation transfer between datasets, drugs, tumors or patients as listed in Supplementary Table S2. Table S11 excludes drug–cell relations from the representation graph’s listed relation types. These controls do not by themselves establish an audit of target-label use during every adaptation or hyperparameter-selection step.","status":"source_checked","source_ids":["scxdr-2026","evidence-task-final-a-scxdr-2026-42003-2025-9418-moesm1-esm-pdf"],"source_locator":"Supplementary Tables S2 and S11"},{"label":"Uncertainty","value":"Transfer-scenario comparisons report mean and standard deviation; variation across source/target tasks must not be interpreted as an individual-cell confidence interval.","status":"source_checked","source_ids":["scxdr-2026"],"source_locator":"Methods: Data collection and processing; group-level evaluation; cached text lines 55–57, 84–86; uncertainty/repeat-run/statistical-comparison passages; matching task comparison table/ablation captions"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["scxdr-2026"],"source_locator":"Methods: Data collection and processing; group-level evaluation; cached text lines 55–57, 84–86; uncertainty/repeat-run/statistical-comparison passages; matching task comparison table/ablation captions"},{"label":"Organisms","value":"Human (Homo sapiens). The nine GEO accessions listed in Supplementary Table S1, including the repeated patient/drug subsets of GSE147326, all identify their sample organism as Homo sapiens (taxon 9606).","status":"source_checked","source_ids":["scxdr-2026","evidence-task-final-a-gse134839-soft","evidence-task-final-a-gse149214-soft","evidence-task-final-a-gse108394-soft","evidence-task-final-a-gse164614-soft","evidence-task-final-a-gse230538-soft","evidence-task-final-a-gse117872-soft","evidence-task-final-a-gse127298-soft","evidence-task-final-a-gse140440-soft","evidence-task-final-a-gse147326-soft"],"source_locator":"Supplementary Table S1; GEO Series_sample_organism and Series_sample_taxid fields"},{"label":"Assays","value":"Single-cell drug-response measurements.","status":"source_checked","source_ids":["scxdr-2026"],"source_locator":"Methods: Data collection and processing; group-level evaluation; cached text lines 55–57, 84–86; uncertainty/repeat-run/statistical-comparison passages; matching task comparison table/ablation captions"},{"label":"Allowed inputs","value":"Source and target single-cell expression representations.","status":"source_checked","source_ids":["scxdr-2026"],"source_locator":"Methods: Data collection and processing; group-level evaluation; cached text lines 55–57, 84–86; uncertainty/repeat-run/statistical-comparison passages; matching task comparison table/ablation captions"},{"label":"Adaptation","value":"Cross-dataset transfer; exact target-label access must be distinguished by transfer scenario.","status":"source_checked","source_ids":["scxdr-2026"],"source_locator":"Methods: Data collection and processing; group-level evaluation; cached text lines 55–57, 84–86; uncertainty/repeat-run/statistical-comparison passages; matching task comparison table/ablation captions"}],"strengths":[],"limitations":[{"text":"Cell-level, cluster-averaged and malignant-group metrics are distinct aggregates. Transfer between datasets does not by itself establish that every comparator used identical target-label access or hyperparameter selection.","source_ids":["scxdr-2026","evidence-task-final-a-scxdr-2026-42003-2025-9418-moesm1-esm-pdf","evidence-task-final-a-gse134839-soft","evidence-task-final-a-gse149214-soft","evidence-task-final-a-gse108394-soft","evidence-task-final-a-gse164614-soft","evidence-task-final-a-gse230538-soft","evidence-task-final-a-gse117872-soft","evidence-task-final-a-gse127298-soft","evidence-task-final-a-gse140440-soft","evidence-task-final-a-gse147326-soft"],"source_locator":"Methods: Data collection and processing; Comparison experiments at the individual cell level; Supplementary Table S2; Supplementary Tables S2 and S11; Supplementary Table S1; GEO Series_sample_organism and Series_sample_taxid fields"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Source and target single-cell expression representations.","Evaluation: The paper defines 20 cross-dataset transfers across batch, drug, tumor and patient scenarios. Comparator protocols differ: scVI/ComBat combine corrected datasets before a train/test partition, whereas transfer methods retain source/target roles. The exact per-task split must accompany any comparison.","Readout: Cell-group accuracy averages within-cluster prediction accuracy; a separate malignant-cell-group accuracy is also described."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["scxdr-2026"],"source_locator":"Methods: Data collection and processing; group-level evaluation; cached text lines 55–57, 84–86; uncertainty/repeat-run/statistical-comparison passages; matching task comparison table/ablation captions; Methods: Data collection and processing; Comparison experiments at the individual cell level; Supplementary Table S2"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Targeted full-paper and supplement review of the outstanding task fields, with original dataset metadata checked where accessible. Source-scoped omissions are explicit; no independent benchmark reproduction or numerical-result change."}}}}
{"id":"reported-task-1c74661df2c401","kind":"benchmark","name":"flu-vaccine mRNA property prediction","description":"","status":"needs_review","facets":{"areas":["rna-transcriptomes"],"tasks":["flu-vaccine mRNA property prediction"]},"source_ids":["codonbert-vaccines-2024"],"links":[{"relation":"dataset","target_id":"reported-dataset-54b9bc432928d6"}],"attributes":{"entity_level":"task","version":null,"task":"flu-vaccine mRNA property prediction","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"The flu-vaccine entry refers to a supervised mRNA-property evaluation dataset within the CodonBERT paper.","summary_source_ids":["codonbert-vaccines-2024"],"summary_source_locator":"Results: supervised learning datasets; cached text lines 14–17; task metric definitions and corresponding results table; matching task comparison table/ablation captions","sections":[{"title":"Evaluation methodology","body":"A separately generated flu-vaccine mRNA dataset with measured expression-related labels. Each downstream dataset uses a shared 70:15:15 training/validation/test split across methods, according to Table 1. Spearman rank correlation for the vaccine-expression regression task; classification accuracy belongs to a different E. coli task. Plain TextCNN, RNABERT/TextCNN, RNA-FM/TextCNN, TF-IDF and Codon2vec/TextCNN configurations in the expression-regression table. The shared split controls partition differences between methods. Homology or construct-family exclusion is not specified in the checked downstream-task split statement. The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","source_ids":["codonbert-vaccines-2024"],"source_locator":"Results: supervised learning datasets; cached text lines 14–17; task metric definitions and corresponding results table; matching task comparison table/ablation captions"}],"facts":[{"label":"Datasets","value":"A separately generated flu-vaccine mRNA dataset with measured expression-related labels.","status":"source_checked","source_ids":["codonbert-vaccines-2024"],"source_locator":"Results: supervised learning datasets; cached text lines 14–17; task metric definitions and corresponding results table; matching task comparison table/ablation captions"},{"label":"Splits","value":"Each downstream dataset uses a shared 70:15:15 training/validation/test split across methods, according to Table 1.","status":"source_checked","source_ids":["codonbert-vaccines-2024"],"source_locator":"Results: supervised learning datasets; cached text lines 14–17; task metric definitions and corresponding results table; matching task comparison table/ablation captions"},{"label":"Metrics","value":"Spearman rank correlation for the vaccine-expression regression task; classification accuracy belongs to a different E. coli task.","status":"source_checked","source_ids":["codonbert-vaccines-2024"],"source_locator":"Results: supervised learning datasets; cached text lines 14–17; task metric definitions and corresponding results table; matching task comparison table/ablation captions"},{"label":"Baselines","value":"Plain TextCNN, RNABERT/TextCNN, RNA-FM/TextCNN, TF-IDF and Codon2vec/TextCNN configurations in the expression-regression table.","status":"source_checked","source_ids":["codonbert-vaccines-2024"],"source_locator":"Results: supervised learning datasets; cached text lines 14–17; task metric definitions and corresponding results table; matching task comparison table/ablation captions"},{"label":"Leakage controls","value":"The shared split controls partition differences between methods. Homology or construct-family exclusion is not specified in the checked downstream-task split statement.","status":"source_checked","source_ids":["codonbert-vaccines-2024"],"source_locator":"Results: supervised learning datasets; cached text lines 14–17; task metric definitions and corresponding results table; matching task comparison table/ablation captions"},{"label":"Uncertainty","value":"The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","status":"unreported","source_ids":["codonbert-vaccines-2024"],"source_locator":"Results: supervised learning datasets; cached text lines 14–17; task metric definitions and corresponding results table; matching task comparison table/ablation captions"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["codonbert-vaccines-2024"],"source_locator":"Results: supervised learning datasets; cached text lines 14–17; task metric definitions and corresponding results table; matching task comparison table/ablation captions"},{"label":"Organisms","value":"Influenza-vaccine mRNA constructs.","status":"source_checked","source_ids":["codonbert-vaccines-2024"],"source_locator":"Results: supervised learning datasets; cached text lines 14–17; task metric definitions and corresponding results table; matching task comparison table/ablation captions"},{"label":"Assays","value":"Measured expression-related mRNA labels.","status":"source_checked","source_ids":["codonbert-vaccines-2024"],"source_locator":"Results: supervised learning datasets; cached text lines 14–17; task metric definitions and corresponding results table; matching task comparison table/ablation captions"},{"label":"Allowed inputs","value":"mRNA sequence representations.","status":"source_checked","source_ids":["codonbert-vaccines-2024"],"source_locator":"Results: supervised learning datasets; cached text lines 14–17; task metric definitions and corresponding results table; matching task comparison table/ablation captions"},{"label":"Adaptation","value":"Supervised downstream expression regression; all methods use the same task split.","status":"source_checked","source_ids":["codonbert-vaccines-2024"],"source_locator":"Results: supervised learning datasets; cached text lines 14–17; task metric definitions and corresponding results table; matching task comparison table/ablation captions"}],"strengths":[],"limitations":[{"text":"Exact split membership, regression aggregation, baseline execution remain unresolved for the separate vaccine-expression dataset.","source_ids":["codonbert-vaccines-2024"],"source_locator":"Results: supervised learning datasets; cached text lines 14–17; task metric definitions and corresponding results table; matching task comparison table/ablation captions"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: mRNA sequence representations.","Evaluation: Supervised downstream expression regression; all methods use the same task split.","Readout: Spearman rank correlation for the vaccine-expression regression task; classification accuracy belongs to a different E. coli task."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["codonbert-vaccines-2024"],"source_locator":"Results: supervised learning datasets; cached text lines 14–17; task metric definitions and corresponding results table; matching task comparison table/ablation captions"},"coverage":"limited","gaps":["Exact split membership, regression aggregation, baseline execution remain unresolved for the separate vaccine-expression dataset."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Task-specific computational methodology and field context checked in the cited primary-source artifact. Source-backed fields, inapplicable evaluator dimensions and unresolved details are distinguished. Numerical results were not reproduced."}}}}
{"id":"reported-task-1ebf9b408517f9","kind":"benchmark","name":"Enzyme functional identity prediction","description":"","status":"needs_review","facets":{"areas":["proteins-complexes"],"tasks":["Enzyme functional identity prediction"]},"source_ids":["fujisan-2024"],"links":[{"relation":"dataset","target_id":"reported-dataset-5197cca532f89d"}],"attributes":{"entity_level":"task","version":null,"task":"Enzyme functional identity prediction","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Enzyme functional-identity classification predicts whether a protein pair shares its annotated reaction function.","summary_source_ids":["fujisan-2024"],"summary_source_locator":"Methods: Dataset construction; classification and metrics; Results: model comparison; cached text lines 10–11, 25, 28, 52, 55; uncertainty/repeat-run/statistical-comparison passages","sections":[{"title":"Evaluation methodology","body":"Swiss-Prot reaction annotations and AlphaFold Database structures; balanced same-function and different-function pairs. Random pair partitions provide training, validation and test subsets; model hyperparameters are tuned using cross-validation. Accuracy, false-positive rate, MCC, precision, recall, F1, AUROC and AUPR. LightGBM and multiple conventional classifiers. The low-sequence-similarity evaluation excludes pairs from the original dataset; random splitting of the main pair dataset does not establish protein-identity holdout.","source_ids":["fujisan-2024"],"source_locator":"Methods: Dataset construction; classification and metrics; Results: model comparison; cached text lines 10–11, 25, 28, 52, 55; uncertainty/repeat-run/statistical-comparison passages"}],"facts":[{"label":"Datasets","value":"Swiss-Prot reaction annotations and AlphaFold Database structures; balanced same-function and different-function pairs.","status":"source_checked","source_ids":["fujisan-2024"],"source_locator":"Methods: Dataset construction; classification and metrics; Results: model comparison; cached text lines 10–11, 25, 28, 52, 55; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Splits","value":"Random pair partitions provide training, validation and test subsets; model hyperparameters are tuned using cross-validation.","status":"source_checked","source_ids":["fujisan-2024"],"source_locator":"Methods: Dataset construction; classification and metrics; Results: model comparison; cached text lines 10–11, 25, 28, 52, 55; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Metrics","value":"Accuracy, false-positive rate, MCC, precision, recall, F1, AUROC and AUPR.","status":"source_checked","source_ids":["fujisan-2024"],"source_locator":"Methods: Dataset construction; classification and metrics; Results: model comparison; cached text lines 10–11, 25, 28, 52, 55; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Baselines","value":"LightGBM and multiple conventional classifiers.","status":"source_checked","source_ids":["fujisan-2024"],"source_locator":"Methods: Dataset construction; classification and metrics; Results: model comparison; cached text lines 10–11, 25, 28, 52, 55; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Leakage controls","value":"The low-sequence-similarity evaluation excludes pairs from the original dataset; random splitting of the main pair dataset does not establish protein-identity holdout.","status":"source_checked","source_ids":["fujisan-2024"],"source_locator":"Methods: Dataset construction; classification and metrics; Results: model comparison; cached text lines 10–11, 25, 28, 52, 55; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Uncertainty","value":"The paper assesses stability using repeated bootstrap iterations; its sampling unit must remain attached to the reported interval.","status":"source_checked","source_ids":["fujisan-2024"],"source_locator":"Methods: Dataset construction; classification and metrics; Results: model comparison; cached text lines 10–11, 25, 28, 52, 55; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["fujisan-2024"],"source_locator":"Methods: Dataset construction; classification and metrics; Results: model comparison; cached text lines 10–11, 25, 28, 52, 55; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Organisms","value":"Proteins were selected through Swiss-Prot release 2022_04 entries with Rhea reaction annotations and AlphaFold DB v4 structures. Dataset construction does not enumerate organism frequencies for the sampled 100,000 protein pairs, so a species-restricted population cannot be assigned.","status":"unreported","source_ids":["fujisan-2024"],"source_locator":"Materials and methods: Dataset construction"},{"label":"Assays","value":"Swiss-Prot reaction/function annotations with AlphaFold structures.","status":"source_checked","source_ids":["fujisan-2024"],"source_locator":"Methods: Dataset construction; classification and metrics; Results: model comparison; cached text lines 10–11, 25, 28, 52, 55; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Allowed inputs","value":"Pairs of enzyme representations.","status":"source_checked","source_ids":["fujisan-2024"],"source_locator":"Methods: Dataset construction; classification and metrics; Results: model comparison; cached text lines 10–11, 25, 28, 52, 55; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Adaptation","value":"Supervised same-function classification; hyperparameters use cross-validation, with algorithm choice additionally compared on test data.","status":"source_checked","source_ids":["fujisan-2024"],"source_locator":"Methods: Dataset construction; classification and metrics; Results: model comparison; cached text lines 10–11, 25, 28, 52, 55; uncertainty/repeat-run/statistical-comparison passages"}],"strengths":[],"limitations":[{"text":"The source reports selecting the best algorithm after test-set comparison, which limits interpretation as a final untouched test. Pair-level splitting does not establish protein-disjoint generalization.","source_ids":["fujisan-2024"],"source_locator":"Methods: Dataset construction; classification and metrics; Results: model comparison; cached text lines 10–11, 25, 28, 52, 55; uncertainty/repeat-run/statistical-comparison passages"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Pairs of enzyme representations.","Evaluation: Random pair partitions provide training, validation and test subsets; model hyperparameters are tuned using cross-validation.","Readout: Accuracy, false-positive rate, MCC, precision, recall, F1, AUROC and AUPR."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["fujisan-2024"],"source_locator":"Methods: Dataset construction; classification and metrics; Results: model comparison; cached text lines 10–11, 25, 28, 52, 55; uncertainty/repeat-run/statistical-comparison passages"},"coverage":"limited","gaps":["Organisms: Proteins were selected through Swiss-Prot release 2022_04 entries with Rhea reaction annotations and AlphaFold DB v4 structures. Dataset construction does not enumerate organism frequencies for the sampled 100,000 protein pairs, so a species-restricted population cannot be assigned."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Targeted full-paper and supplement review of the outstanding task fields, with original dataset metadata checked where accessible. Source-scoped omissions are explicit; no independent benchmark reproduction or numerical-result change."}}}}
{"id":"reported-task-22024610c4d658","kind":"benchmark","name":"enhancer prediction","description":"","status":"needs_review","facets":{"areas":["dna-genomes"],"tasks":["enhancer prediction"]},"source_ids":["hi-enhancer-2025"],"links":[{"relation":"dataset","target_id":"reported-dataset-a8610f2b80cdf0"}],"attributes":{"entity_level":"task","version":null,"task":"enhancer prediction","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Enhancer classification evaluates a stacked predictor using functional genomic signal features.","summary_source_ids":["hi-enhancer-2025"],"summary_source_locator":"Methods §§2.1.1–2.1.2; cached text lines 13–14, 20–22; task metric definitions and corresponding results table","sections":[{"title":"Evaluation methodology","body":"HCT116 and A549 annotations assembled from ENCODE accessibility, activity and chromatin-mark data, with class-imbalanced negatives. The first layer uses a training/validation partition; subsequent stacking uses held-out base-model predictions and cross-validation. Accuracy, AUROC and AUPRC; fold-wise standard deviations are reported separately from runtime. AutoGluon base classifiers and a KAN meta-classifier are components of the evaluation. For the separate sequence-based boundary stage, Supplementary Text S6 reports CD-HIT filtering of iEnhancer-2L sequences but gives an unusual “>20% similarity” threshold without a reproducible command. The signal-based region detector instead uses the blending splits in Text S4. Neither passage establishes chromosome separation across that detector’s evaluation folds.","source_ids":["hi-enhancer-2025","evidence-task-final-a-hi-enhancer-2025-btaf441-supplementary-data-docx"],"source_locator":"Methods §§2.1.1–2.1.2; cached text lines 13–14, 20–22; task metric definitions and corresponding results table; Supplementary Text S4 Dataset division and comparison with DECODE; Text S6 Benchmark datasets and performance evaluation metrics"}],"facts":[{"label":"Datasets","value":"HCT116 and A549 annotations assembled from ENCODE accessibility, activity and chromatin-mark data, with class-imbalanced negatives.","status":"source_checked","source_ids":["hi-enhancer-2025"],"source_locator":"Methods §§2.1.1–2.1.2; cached text lines 13–14, 20–22; task metric definitions and corresponding results table"},{"label":"Splits","value":"The first layer uses a training/validation partition; subsequent stacking uses held-out base-model predictions and cross-validation.","status":"source_checked","source_ids":["hi-enhancer-2025"],"source_locator":"Methods §§2.1.1–2.1.2; cached text lines 13–14, 20–22; task metric definitions and corresponding results table"},{"label":"Metrics","value":"Accuracy, AUROC and AUPRC; fold-wise standard deviations are reported separately from runtime.","status":"source_checked","source_ids":["hi-enhancer-2025"],"source_locator":"Methods §§2.1.1–2.1.2; cached text lines 13–14, 20–22; task metric definitions and corresponding results table"},{"label":"Baselines","value":"AutoGluon base classifiers and a KAN meta-classifier are components of the evaluation.","status":"source_checked","source_ids":["hi-enhancer-2025"],"source_locator":"Methods §§2.1.1–2.1.2; cached text lines 13–14, 20–22; task metric definitions and corresponding results table"},{"label":"Leakage controls","value":"For the separate sequence-based boundary stage, Supplementary Text S6 reports CD-HIT filtering of iEnhancer-2L sequences but gives an unusual “>20% similarity” threshold without a reproducible command. The signal-based region detector instead uses the blending splits in Text S4. Neither passage establishes chromosome separation across that detector’s evaluation folds.","status":"source_checked","source_ids":["hi-enhancer-2025","evidence-task-final-a-hi-enhancer-2025-btaf441-supplementary-data-docx"],"source_locator":"Supplementary Text S4 Dataset division and comparison with DECODE; Text S6 Benchmark datasets and performance evaluation metrics"},{"label":"Uncertainty","value":"Standard deviations across five cross-validation folds are reported for the classification metrics.","status":"source_checked","source_ids":["hi-enhancer-2025"],"source_locator":"Methods §§2.1.1–2.1.2; cached text lines 13–14, 20–22; task metric definitions and corresponding results table"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["hi-enhancer-2025"],"source_locator":"Methods §§2.1.1–2.1.2; cached text lines 13–14, 20–22; task metric definitions and corresponding results table"},{"label":"Organisms","value":"Human HCT116 and A549 cells.","status":"source_checked","source_ids":["hi-enhancer-2025"],"source_locator":"Methods §§2.1.1–2.1.2; cached text lines 13–14, 20–22; task metric definitions and corresponding results table"},{"label":"Assays","value":"ENCODE accessibility, activity and chromatin-mark annotations.","status":"source_checked","source_ids":["hi-enhancer-2025"],"source_locator":"Methods §§2.1.1–2.1.2; cached text lines 13–14, 20–22; task metric definitions and corresponding results table"},{"label":"Allowed inputs","value":"Genomic regions with functional/epigenetic signals used by the classifier; this is not a sequence-only evaluation.","status":"source_checked","source_ids":["hi-enhancer-2025"],"source_locator":"Methods §§2.1.1–2.1.2; cached text lines 13–14, 20–22; task metric definitions and corresponding results table"},{"label":"Adaptation","value":"Supervised stacked prediction; second-stage learning uses held-out base-model predictions.","status":"source_checked","source_ids":["hi-enhancer-2025"],"source_locator":"Methods §§2.1.1–2.1.2; cached text lines 13–14, 20–22; task metric definitions and corresponding results table"}],"strengths":[],"limitations":[{"text":"The epigenetic-signal region detector and the sequence-based boundary stage use different data and splits. The supplement’s unusual CD-HIT threshold requires an implementation-level clarification before claiming homology-disjoint evaluation.","source_ids":["hi-enhancer-2025","evidence-task-final-a-hi-enhancer-2025-btaf441-supplementary-data-docx"],"source_locator":"Supplementary Text S4 Dataset division and comparison with DECODE; Text S6 Benchmark datasets and performance evaluation metrics"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Genomic regions with functional/epigenetic signals used by the classifier; this is not a sequence-only evaluation.","Evaluation: The first layer uses a training/validation partition; subsequent stacking uses held-out base-model predictions and cross-validation.","Readout: Accuracy, AUROC and AUPRC; fold-wise standard deviations are reported separately from runtime."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["hi-enhancer-2025"],"source_locator":"Methods §§2.1.1–2.1.2; cached text lines 13–14, 20–22; task metric definitions and corresponding results table"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Targeted full-paper and supplement review of the outstanding task fields, with original dataset metadata checked where accessible. Source-scoped omissions are explicit; no independent benchmark reproduction or numerical-result change."}}}}
{"id":"reported-task-2cbac97dd849f5","kind":"benchmark","name":"Enhancer-target gene prediction","description":"","status":"needs_review","facets":{"areas":["dna-genomes"],"tasks":["Enhancer-target gene prediction"]},"source_ids":["dnalongbench-2025"],"links":[{"relation":"dataset","target_id":"reported-dataset-fcb5752d916d5d"}],"attributes":{"entity_level":"task","version":null,"task":"Enhancer-target gene prediction","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Enhancer-target classification tests whether sequence models distinguish experimentally annotated interaction candidates.","summary_source_ids":["dnalongbench-2025"],"summary_source_locator":"Proposed Dataset: enhancer-promoter interactions; Experiments; cached text lines 27–28, 48","sections":[{"title":"Evaluation methodology","body":"K562 enhancer-promoter candidate pairs with positive and negative annotations from prior studies. The enhancer-target task paragraph identifies the curated K562 pairs and evaluation metric but does not state its partition assignment; nearby split descriptions belong to other tasks. AUROC for enhancer-promoter interaction classification. A compact CNN and DNA foundation models are compared with the activity-by-contact expert model. The enhancer-target paragraph does not establish gene- or overlapping-interval-disjoint partitions; unrelated task split rules cannot be imported. The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","source_ids":["dnalongbench-2025"],"source_locator":"Proposed Dataset: enhancer-promoter interactions; Experiments; cached text lines 27–28, 48"}],"facts":[{"label":"Datasets","value":"K562 enhancer-promoter candidate pairs with positive and negative annotations from prior studies.","status":"source_checked","source_ids":["dnalongbench-2025"],"source_locator":"Proposed Dataset: enhancer-promoter interactions; Experiments; cached text lines 27–28, 48"},{"label":"Splits","value":"The enhancer-target task paragraph identifies the curated K562 pairs and evaluation metric but does not state its partition assignment; nearby split descriptions belong to other tasks.","status":"unreported","source_ids":["dnalongbench-2025"],"source_locator":"Proposed Dataset: enhancer-promoter interactions; Experiments; cached text lines 27–28, 48"},{"label":"Metrics","value":"AUROC for enhancer-promoter interaction classification.","status":"source_checked","source_ids":["dnalongbench-2025"],"source_locator":"Proposed Dataset: enhancer-promoter interactions; Experiments; cached text lines 27–28, 48"},{"label":"Baselines","value":"A compact CNN and DNA foundation models are compared with the activity-by-contact expert model.","status":"source_checked","source_ids":["dnalongbench-2025"],"source_locator":"Proposed Dataset: enhancer-promoter interactions; Experiments; cached text lines 27–28, 48"},{"label":"Leakage controls","value":"The enhancer-target paragraph does not establish gene- or overlapping-interval-disjoint partitions; unrelated task split rules cannot be imported.","status":"unreported","source_ids":["dnalongbench-2025"],"source_locator":"Proposed Dataset: enhancer-promoter interactions; Experiments; cached text lines 27–28, 48"},{"label":"Uncertainty","value":"The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","status":"unreported","source_ids":["dnalongbench-2025"],"source_locator":"Proposed Dataset: enhancer-promoter interactions; Experiments; cached text lines 27–28, 48"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["dnalongbench-2025"],"source_locator":"Proposed Dataset: enhancer-promoter interactions; Experiments; cached text lines 27–28, 48"},{"label":"Organisms","value":"Human K562.","status":"source_checked","source_ids":["dnalongbench-2025"],"source_locator":"Proposed Dataset: enhancer-promoter interactions; Experiments; cached text lines 27–28, 48"},{"label":"Assays","value":"Enhancer–promoter link annotations.","status":"source_checked","source_ids":["dnalongbench-2025"],"source_locator":"Proposed Dataset: enhancer-promoter interactions; Experiments; cached text lines 27–28, 48"},{"label":"Allowed inputs","value":"Long genomic sequence context for enhancer/target-gene pairs.","status":"source_checked","source_ids":["dnalongbench-2025"],"source_locator":"Proposed Dataset: enhancer-promoter interactions; Experiments; cached text lines 27–28, 48"},{"label":"Adaptation","value":"Downstream DNA-model prediction compared with a CNN and the activity-by-contact expert model.","status":"source_checked","source_ids":["dnalongbench-2025"],"source_locator":"Proposed Dataset: enhancer-promoter interactions; Experiments; cached text lines 27–28, 48"}],"strengths":[],"limitations":[{"text":"The activity-by-contact comparator receives functional assays and chromatin-contact data that sequence-only models do not. Splits described for other DNALongBench tasks are not imported into this record.","source_ids":["dnalongbench-2025"],"source_locator":"Proposed Dataset: enhancer-promoter interactions; Experiments; cached text lines 27–28, 48"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Long genomic sequence context for enhancer/target-gene pairs.","Evaluation: Downstream DNA-model prediction compared with a CNN and the activity-by-contact expert model.","Readout: AUROC for enhancer-promoter interaction classification."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["dnalongbench-2025"],"source_locator":"Proposed Dataset: enhancer-promoter interactions; Experiments; cached text lines 27–28, 48"},"coverage":"limited","gaps":["The activity-by-contact comparator receives functional assays and chromatin-contact data that sequence-only models do not. Splits described for other DNALongBench tasks are not imported into this record."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Task-specific computational methodology and field context checked in the cited primary-source artifact. Source-backed fields, inapplicable evaluator dimensions and unresolved details are distinguished. Numerical results were not reproduced."}}}}
{"id":"reported-task-3063ed4da76b4b","kind":"benchmark","name":"Gene-regulatory link prediction","description":"","status":"needs_review","facets":{"areas":["cells-tissues"],"tasks":["Gene-regulatory link prediction"]},"source_ids":["scregnet-2025"],"links":[{"relation":"dataset","target_id":"reported-dataset-2ad2fad5e1cd0a"}],"attributes":{"entity_level":"task","version":null,"task":"Gene-regulatory link prediction","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Gene-regulatory link prediction uses cell-specific reference networks to evaluate inferred regulatory edges.","summary_source_ids":["scregnet-2025"],"summary_source_locator":"Methods: Datasets and data pre-processing; Baseline models and evaluation metrics; Results: ablations; cached text lines 46–48, 55–56, 65; benchmark datasets and per-TF partition paragraph","sections":[{"title":"Evaluation methodology","body":"Seven BEELINE human/mouse scRNA-seq datasets with cell-type-specific ChIP-seq reference networks. For each TF, target/non-target edges are divided 67:33 into training/test, with part of training reserved for validation; all TFs contribute to both arms. AUROC and AUPRC; selected comparisons average results from two gene-panel sizes. GNNLink, GENELink, GNE, CNNC, DeepDRIM, GRN-transformer, PCC, GRNBoost2 and GENIE3. Target genes are separated between training and testing for the same TF. This is not a leave-TF-out design and does not establish globally disjoint gene identities across all TFs. The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","source_ids":["scregnet-2025"],"source_locator":"Methods: Datasets and data pre-processing; Baseline models and evaluation metrics; Results: ablations; cached text lines 46–48, 55–56, 65; benchmark datasets and per-TF partition paragraph"}],"facts":[{"label":"Datasets","value":"Seven BEELINE human/mouse scRNA-seq datasets with cell-type-specific ChIP-seq reference networks.","status":"source_checked","source_ids":["scregnet-2025"],"source_locator":"Methods: Datasets and data pre-processing; Baseline models and evaluation metrics; Results: ablations; cached text lines 46–48, 55–56, 65; benchmark datasets and per-TF partition paragraph"},{"label":"Splits","value":"For each TF, target/non-target edges are divided 67:33 into training/test, with part of training reserved for validation; all TFs contribute to both arms.","status":"source_checked","source_ids":["scregnet-2025"],"source_locator":"Methods: Datasets and data pre-processing; Baseline models and evaluation metrics; Results: ablations; cached text lines 46–48, 55–56, 65; benchmark datasets and per-TF partition paragraph"},{"label":"Metrics","value":"AUROC and AUPRC; selected comparisons average results from two gene-panel sizes.","status":"source_checked","source_ids":["scregnet-2025"],"source_locator":"Methods: Datasets and data pre-processing; Baseline models and evaluation metrics; Results: ablations; cached text lines 46–48, 55–56, 65; benchmark datasets and per-TF partition paragraph"},{"label":"Baselines","value":"GNNLink, GENELink, GNE, CNNC, DeepDRIM, GRN-transformer, PCC, GRNBoost2 and GENIE3.","status":"source_checked","source_ids":["scregnet-2025"],"source_locator":"Methods: Datasets and data pre-processing; Baseline models and evaluation metrics; Results: ablations; cached text lines 46–48, 55–56, 65; benchmark datasets and per-TF partition paragraph"},{"label":"Leakage controls","value":"Target genes are separated between training and testing for the same TF. This is not a leave-TF-out design and does not establish globally disjoint gene identities across all TFs.","status":"source_checked","source_ids":["scregnet-2025"],"source_locator":"Methods: Datasets and data pre-processing; Baseline models and evaluation metrics; Results: ablations; cached text lines 46–48, 55–56, 65; benchmark datasets and per-TF partition paragraph"},{"label":"Uncertainty","value":"The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","status":"unreported","source_ids":["scregnet-2025"],"source_locator":"Methods: Datasets and data pre-processing; Baseline models and evaluation metrics; Results: ablations; cached text lines 46–48, 55–56, 65; benchmark datasets and per-TF partition paragraph"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["scregnet-2025"],"source_locator":"Methods: Datasets and data pre-processing; Baseline models and evaluation metrics; Results: ablations; cached text lines 46–48, 55–56, 65; benchmark datasets and per-TF partition paragraph"},{"label":"Organisms","value":"Human and mouse.","status":"source_checked","source_ids":["scregnet-2025"],"source_locator":"Methods: Datasets and data pre-processing; Baseline models and evaluation metrics; Results: ablations; cached text lines 46–48, 55–56, 65; benchmark datasets and per-TF partition paragraph"},{"label":"Assays","value":"Single-cell RNA-seq with cell-type-specific ChIP-seq regulatory references.","status":"source_checked","source_ids":["scregnet-2025"],"source_locator":"Methods: Datasets and data pre-processing; Baseline models and evaluation metrics; Results: ablations; cached text lines 46–48, 55–56, 65; benchmark datasets and per-TF partition paragraph"},{"label":"Allowed inputs","value":"Single-cell expression and candidate regulatory gene pairs.","status":"source_checked","source_ids":["scregnet-2025"],"source_locator":"Methods: Datasets and data pre-processing; Baseline models and evaluation metrics; Results: ablations; cached text lines 46–48, 55–56, 65; benchmark datasets and per-TF partition paragraph"},{"label":"Adaptation","value":"Supervised regulatory-link prediction compared with neural and conventional network-inference methods.","status":"source_checked","source_ids":["scregnet-2025"],"source_locator":"Methods: Datasets and data pre-processing; Baseline models and evaluation metrics; Results: ablations; cached text lines 46–48, 55–56, 65; benchmark datasets and per-TF partition paragraph"}],"strengths":[],"limitations":[{"text":"Gene-panel definition and incomplete reference networks affect the candidate edge population. Negative-pair construction and cross-TF gene sharing must be distinguished from the stated per-TF target split.","source_ids":["scregnet-2025"],"source_locator":"Methods: Datasets and data pre-processing; Baseline models and evaluation metrics; Results: ablations; cached text lines 46–48, 55–56, 65; benchmark datasets and per-TF partition paragraph"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Single-cell expression and candidate regulatory gene pairs.","Evaluation: Supervised regulatory-link prediction compared with neural and conventional network-inference methods.","Readout: AUROC and AUPRC; selected comparisons average results from two gene-panel sizes."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["scregnet-2025"],"source_locator":"Methods: Datasets and data pre-processing; Baseline models and evaluation metrics; Results: ablations; cached text lines 46–48, 55–56, 65; benchmark datasets and per-TF partition paragraph"},"coverage":"limited","gaps":["Gene-panel definition and incomplete reference networks affect the candidate edge population. Negative-pair construction and cross-TF gene sharing must be distinguished from the stated per-TF target split."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Task-specific computational methodology and field context checked in the cited primary-source artifact. Source-backed fields, inapplicable evaluator dimensions and unresolved details are distinguished. Numerical results were not reproduced."}}}}
{"id":"reported-task-3109f8d0f2b7b5","kind":"benchmark","name":"cross-species conservation prediction","description":"","status":"needs_review","facets":{"areas":["dna-genomes"],"tasks":["cross-species conservation prediction"]},"source_ids":["plantcad2-2025"],"links":[{"relation":"dataset","target_id":"reported-dataset-b6d0ebaca196a6"}],"attributes":{"entity_level":"task","version":null,"task":"cross-species conservation prediction","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Conservation prediction compares zero-shot sequence scores across genomic models with different context and prediction objectives.","summary_source_ids":["plantcad2-2025"],"summary_source_locator":"Methods: Zero-shot evaluation of PlantCAD2, PlantCAD and GPN; Evo2 evaluation; cached text lines 110–115; task metric definitions and corresponding results table","sections":[{"title":"Evaluation methodology","body":"Three conservation tasks use Sorghum bicolor with Andropogoneae alignments, a Poaceae coding-sequence alignment with Pharus latifolius as outgroup, and potato (Solanum tuberosum) with 95 Solanaceae genomes. Each uses its own alignment-derived conservation labels. This is zero-shot scoring rather than fitting on conservation labels. For the Andropogoneae and Poaceae tasks, the evaluated site is centered at position 4,096 in an 8,192-base window and its reference base is masked. Conservation thresholds and selected positions differ by task. AUROC for conserved/non-conserved site classification in the cross-species conservation evaluations. PlantCAD2, PlantCAD, GPN and Evo2. The paper explicitly states that Solanaceae species were absent from PlantCAD2 pretraining, while Evo 2 included multiple Solanum genomes. This supports a scoped distinction for the potato experiment, not a universal claim that every evaluated position or homolog was unseen.","source_ids":["plantcad2-2025"],"source_locator":"Methods: Evolutionary constraint prediction using the zero-shot strategy; Methods: Zero-shot evaluation of PlantCAD2, PlantCAD and GPN; Evo2 evaluation; cached text lines 110–115; task metric definitions and corresponding results table; Results: PlantCAD2 accurately predicts evolutionary conservation with a zero-shot strategy; Methods: Evolutionary constraint prediction using the zero-shot strategy"}],"facts":[{"label":"Datasets","value":"Three conservation tasks use Sorghum bicolor with Andropogoneae alignments, a Poaceae coding-sequence alignment with Pharus latifolius as outgroup, and potato (Solanum tuberosum) with 95 Solanaceae genomes. Each uses its own alignment-derived conservation labels.","status":"source_checked","source_ids":["plantcad2-2025"],"source_locator":"Methods: Evolutionary constraint prediction using the zero-shot strategy"},{"label":"Splits","value":"This is zero-shot scoring rather than fitting on conservation labels. For the Andropogoneae and Poaceae tasks, the evaluated site is centered at position 4,096 in an 8,192-base window and its reference base is masked. Conservation thresholds and selected positions differ by task.","status":"source_checked","source_ids":["plantcad2-2025"],"source_locator":"Methods: Evolutionary constraint prediction using the zero-shot strategy"},{"label":"Metrics","value":"AUROC for conserved/non-conserved site classification in the cross-species conservation evaluations.","status":"source_checked","source_ids":["plantcad2-2025"],"source_locator":"Methods: Zero-shot evaluation of PlantCAD2, PlantCAD and GPN; Evo2 evaluation; cached text lines 110–115; task metric definitions and corresponding results table"},{"label":"Baselines","value":"PlantCAD2, PlantCAD, GPN and Evo2.","status":"source_checked","source_ids":["plantcad2-2025"],"source_locator":"Methods: Zero-shot evaluation of PlantCAD2, PlantCAD and GPN; Evo2 evaluation; cached text lines 110–115; task metric definitions and corresponding results table"},{"label":"Leakage controls","value":"The paper explicitly states that Solanaceae species were absent from PlantCAD2 pretraining, while Evo 2 included multiple Solanum genomes. This supports a scoped distinction for the potato experiment, not a universal claim that every evaluated position or homolog was unseen.","status":"source_checked","source_ids":["plantcad2-2025"],"source_locator":"Results: PlantCAD2 accurately predicts evolutionary conservation with a zero-shot strategy; Methods: Evolutionary constraint prediction using the zero-shot strategy"},{"label":"Uncertainty","value":"Supplemental Table 2 supplies task-, model- and context-specific AUROC point estimates, without uncertainty columns. Figure S2 varies context length; that variation is not a confidence interval or repeated-seed error estimate.","status":"unreported","source_ids":["plantcad2-2025","evidence-task-final-a-plantcad2-2025-media-1-xlsx"],"source_locator":"Supplemental Table 2 worksheet: Task, Model, Context and AUROC columns; Figure S2"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["plantcad2-2025"],"source_locator":"Methods: Zero-shot evaluation of PlantCAD2, PlantCAD and GPN; Evo2 evaluation; cached text lines 110–115; task metric definitions and corresponding results table"},{"label":"Organisms","value":"Plant genomic sequences.","status":"source_checked","source_ids":["plantcad2-2025"],"source_locator":"Methods: Zero-shot evaluation of PlantCAD2, PlantCAD and GPN; Evo2 evaluation; cached text lines 110–115; task metric definitions and corresponding results table"},{"label":"Assays","value":"Conservation/evolutionary-constraint annotations.","status":"source_checked","source_ids":["plantcad2-2025"],"source_locator":"Methods: Zero-shot evaluation of PlantCAD2, PlantCAD and GPN; Evo2 evaluation; cached text lines 110–115; task metric definitions and corresponding results table"},{"label":"Allowed inputs","value":"Plant DNA sequence.","status":"source_checked","source_ids":["plantcad2-2025"],"source_locator":"Methods: Zero-shot evaluation of PlantCAD2, PlantCAD and GPN; Evo2 evaluation; cached text lines 110–115; task metric definitions and corresponding results table"},{"label":"Adaptation","value":"Zero-shot scoring compared across PlantCAD2, PlantCAD, GPN and Evo2.","status":"source_checked","source_ids":["plantcad2-2025"],"source_locator":"Methods: Zero-shot evaluation of PlantCAD2, PlantCAD and GPN; Evo2 evaluation; cached text lines 110–115; task metric definitions and corresponding results table"}],"strengths":[],"limitations":[{"text":"Short-context comparators receive cropped inputs. Evo2 uses a reference-allele likelihood because its autoregressive objective does not support identical masked-token scoring; those are material comparison differences.","source_ids":["plantcad2-2025"],"source_locator":"Methods: Zero-shot evaluation of PlantCAD2, PlantCAD and GPN; Evo2 evaluation; cached text lines 110–115; task metric definitions and corresponding results table"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Plant DNA sequence.","Evaluation: This is zero-shot scoring rather than fitting on conservation labels. For the Andropogoneae and Poaceae tasks, the evaluated site is centered at position 4,096 in an 8,192-base window and its reference base is masked. Conservation thresholds and selected positions differ by task.","Readout: AUROC for conserved/non-conserved site classification in the cross-species conservation evaluations."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["plantcad2-2025"],"source_locator":"Methods: Zero-shot evaluation of PlantCAD2, PlantCAD and GPN; Evo2 evaluation; cached text lines 110–115; task metric definitions and corresponding results table; Methods: Evolutionary constraint prediction using the zero-shot strategy"},"coverage":"limited","gaps":["Uncertainty: Supplemental Table 2 supplies task-, model- and context-specific AUROC point estimates, without uncertainty columns. Figure S2 varies context length; that variation is not a confidence interval or repeated-seed error estimate."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Targeted full-paper and supplement review of the outstanding task fields, with original dataset metadata checked where accessible. Source-scoped omissions are explicit; no independent benchmark reproduction or numerical-result change."}}}}
{"id":"reported-task-369dcfef14c4a9","kind":"benchmark","name":"Simulated metagenome virus-taxon retrieval","description":"","status":"needs_review","facets":{"areas":["microbes-communities"],"tasks":["Simulated metagenome virus-taxon retrieval"]},"source_ids":["lazypipe-2020"],"links":[{"relation":"dataset","target_id":"reported-dataset-450c1af18cc623"}],"attributes":{"entity_level":"task","version":null,"task":"Simulated metagenome virus-taxon retrieval","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Simulated metagenome evaluation measures recovery of known viral taxa at multiple taxonomic levels.","summary_source_ids":["lazypipe-2020"],"summary_source_locator":"Methods §2.3; Results §3.1; cached text lines 27–29, 35–37","sections":[{"title":"Evaluation methodology","body":"MetaShot simulated reads with origin labels mapped through NCBI taxonomy into a CAMI reference profile. An index/query evaluation against a fixed known mixture; the paper separately evaluates mock-community data. OPAL-based precision, recall and F1 for viral taxa and all predicted taxa, with genus/species levels distinguished. Lazypipe variants, Centrifuge, MetaPhlAn2 and Kraken2 are discussed in the comparison. The MetaShot simulation is scored against accession-derived viral and bacterial truth using reference-based classifiers. Methods 2.3 does not report removing the simulated source genomes from the classifiers’ reference databases; this is not established as an unseen-genome test.","source_ids":["lazypipe-2020"],"source_locator":"Methods §2.3; Results §3.1; cached text lines 27–29, 35–37; Methods 2.3 Benchmarking performance"}],"facts":[{"label":"Datasets","value":"MetaShot simulated reads with origin labels mapped through NCBI taxonomy into a CAMI reference profile.","status":"source_checked","source_ids":["lazypipe-2020"],"source_locator":"Methods §2.3; Results §3.1; cached text lines 27–29, 35–37"},{"label":"Splits","value":"An index/query evaluation against a fixed known mixture; the paper separately evaluates mock-community data.","status":"source_checked","source_ids":["lazypipe-2020"],"source_locator":"Methods §2.3; Results §3.1; cached text lines 27–29, 35–37"},{"label":"Metrics","value":"OPAL-based precision, recall and F1 for viral taxa and all predicted taxa, with genus/species levels distinguished.","status":"source_checked","source_ids":["lazypipe-2020"],"source_locator":"Methods §2.3; Results §3.1; cached text lines 27–29, 35–37"},{"label":"Baselines","value":"Lazypipe variants, Centrifuge, MetaPhlAn2 and Kraken2 are discussed in the comparison.","status":"source_checked","source_ids":["lazypipe-2020"],"source_locator":"Methods §2.3; Results §3.1; cached text lines 27–29, 35–37"},{"label":"Leakage controls","value":"The MetaShot simulation is scored against accession-derived viral and bacterial truth using reference-based classifiers. Methods 2.3 does not report removing the simulated source genomes from the classifiers’ reference databases; this is not established as an unseen-genome test.","status":"unreported","source_ids":["lazypipe-2020"],"source_locator":"Methods 2.3 Benchmarking performance"},{"label":"Uncertainty","value":"The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","status":"unreported","source_ids":["lazypipe-2020"],"source_locator":"Methods §2.3; Results §3.1; cached text lines 27–29, 35–37"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["lazypipe-2020"],"source_locator":"Methods §2.3; Results §3.1; cached text lines 27–29, 35–37"},{"label":"Organisms","value":"Simulated viral taxa and background community members.","status":"source_checked","source_ids":["lazypipe-2020"],"source_locator":"Methods §2.3; Results §3.1; cached text lines 27–29, 35–37"},{"label":"Assays","value":"MetaShot reads with known origin taxonomy.","status":"source_checked","source_ids":["lazypipe-2020"],"source_locator":"Methods §2.3; Results §3.1; cached text lines 27–29, 35–37"},{"label":"Allowed inputs","value":"Metagenomic reads and reference resources.","status":"source_checked","source_ids":["lazypipe-2020"],"source_locator":"Methods §2.3; Results §3.1; cached text lines 27–29, 35–37"},{"label":"Adaptation","value":"Index/query analysis of a known mixture; mock-community data are a separate test.","status":"source_checked","source_ids":["lazypipe-2020"],"source_locator":"Methods §2.3; Results §3.1; cached text lines 27–29, 35–37"}],"strengths":[],"limitations":[{"text":"Taxonomic rank changes the assessed outcome. A simulation drawn from known reference organisms does not establish performance on viruses absent from the reference databases.","source_ids":["lazypipe-2020"],"source_locator":"Methods 2.3 Benchmarking performance"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Metagenomic reads and reference resources.","Evaluation: An index/query evaluation against a fixed known mixture; the paper separately evaluates mock-community data.","Readout: OPAL-based precision, recall and F1 for viral taxa and all predicted taxa, with genus/species levels distinguished."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["lazypipe-2020"],"source_locator":"Methods §2.3; Results §3.1; cached text lines 27–29, 35–37"},"coverage":"limited","gaps":["Leakage controls: The MetaShot simulation is scored against accession-derived viral and bacterial truth using reference-based classifiers. Methods 2.3 does not report removing the simulated source genomes from the classifiers’ reference databases; this is not established as an unseen-genome test.","Uncertainty: The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Targeted full-paper and supplement review of the outstanding task fields, with original dataset metadata checked where accessible. Source-scoped omissions are explicit; no independent benchmark reproduction or numerical-result change."}}}}
{"id":"reported-task-3891811dcce8b3","kind":"benchmark","name":"E. coli sigma70 promoter prediction","description":"","status":"needs_review","facets":{"areas":["microbes-communities"],"tasks":["E. coli sigma70 promoter prediction"]},"source_ids":["prokbert-2024"],"links":[{"relation":"dataset","target_id":"reported-dataset-48def1da574597"}],"attributes":{"entity_level":"task","version":null,"task":"E. coli sigma70 promoter prediction","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"The E. coli promoter task is an independent test setting within ProkBERT’s broader prokaryotic evaluation.","summary_source_ids":["prokbert-2024"],"summary_source_locator":"Methods §2.3.1; Results: promoter comparison; cached text lines 66–67, 138–141","sections":[{"title":"Evaluation methodology","body":"PPD-derived promoter data and a separately identified E. coli sigma70 test collection. Independent E. coli testing is described; exact training exclusions remain unextracted. Sensitivity, specificity and accuracy are discussed together to expose false-positive tradeoffs. CNNProm, Sigma70Pred, iPromoter-BnCNN, iPromoter-2L and Promotech. The source warns that some comparator training data may overlap or be closely related to the E. coli evaluation data. The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","source_ids":["prokbert-2024"],"source_locator":"Methods §2.3.1; Results: promoter comparison; cached text lines 66–67, 138–141"}],"facts":[{"label":"Datasets","value":"PPD-derived promoter data and a separately identified E. coli sigma70 test collection.","status":"source_checked","source_ids":["prokbert-2024"],"source_locator":"Methods §2.3.1; Results: promoter comparison; cached text lines 66–67, 138–141"},{"label":"Splits","value":"Independent E. coli testing is described; exact training exclusions remain unextracted.","status":"source_checked","source_ids":["prokbert-2024"],"source_locator":"Methods §2.3.1; Results: promoter comparison; cached text lines 66–67, 138–141"},{"label":"Metrics","value":"Sensitivity, specificity and accuracy are discussed together to expose false-positive tradeoffs.","status":"source_checked","source_ids":["prokbert-2024"],"source_locator":"Methods §2.3.1; Results: promoter comparison; cached text lines 66–67, 138–141"},{"label":"Baselines","value":"CNNProm, Sigma70Pred, iPromoter-BnCNN, iPromoter-2L and Promotech.","status":"source_checked","source_ids":["prokbert-2024"],"source_locator":"Methods §2.3.1; Results: promoter comparison; cached text lines 66–67, 138–141"},{"label":"Leakage controls","value":"The source warns that some comparator training data may overlap or be closely related to the E. coli evaluation data.","status":"source_checked","source_ids":["prokbert-2024"],"source_locator":"Methods §2.3.1; Results: promoter comparison; cached text lines 66–67, 138–141"},{"label":"Uncertainty","value":"The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","status":"unreported","source_ids":["prokbert-2024"],"source_locator":"Methods §2.3.1; Results: promoter comparison; cached text lines 66–67, 138–141"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["prokbert-2024"],"source_locator":"Methods §2.3.1; Results: promoter comparison; cached text lines 66–67, 138–141"},{"label":"Organisms","value":"Escherichia coli for the sigma70 test.","status":"source_checked","source_ids":["prokbert-2024"],"source_locator":"Methods §2.3.1; Results: promoter comparison; cached text lines 66–67, 138–141"},{"label":"Assays","value":"PPD and independent sigma70 promoter annotations.","status":"source_checked","source_ids":["prokbert-2024"],"source_locator":"Methods §2.3.1; Results: promoter comparison; cached text lines 66–67, 138–141"},{"label":"Allowed inputs","value":"DNA promoter-window sequence.","status":"source_checked","source_ids":["prokbert-2024"],"source_locator":"Methods §2.3.1; Results: promoter comparison; cached text lines 66–67, 138–141"},{"label":"Adaptation","value":"Supervised promoter prediction compared with established task-specific methods.","status":"source_checked","source_ids":["prokbert-2024"],"source_locator":"Methods §2.3.1; Results: promoter comparison; cached text lines 66–67, 138–141"}],"strengths":[],"limitations":[{"text":"Comparator source versions remain unextracted.","source_ids":["prokbert-2024"],"source_locator":"Methods §2.3.1; Results: promoter comparison; cached text lines 66–67, 138–141"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: DNA promoter-window sequence.","Evaluation: Supervised promoter prediction compared with established task-specific methods.","Readout: Sensitivity, specificity and accuracy are discussed together to expose false-positive tradeoffs."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["prokbert-2024"],"source_locator":"Methods §2.3.1; Results: promoter comparison; cached text lines 66–67, 138–141"},"coverage":"limited","gaps":["Comparator source versions remain unextracted."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Task-specific computational methodology and field context checked in the cited primary-source artifact. Source-backed fields, inapplicable evaluator dimensions and unresolved details are distinguished. Numerical results were not reproduced."}}}}
{"id":"reported-task-3a3bff34cce634","kind":"benchmark","name":"RNA compound-binding site prediction","description":"","status":"needs_review","facets":{"areas":["rna-transcriptomes"],"tasks":["RNA compound-binding site prediction"]},"source_ids":["cobra-rna-binding-2026"],"links":[{"relation":"dataset","target_id":"reported-dataset-b1af840b76b351"}],"attributes":{"entity_level":"task","version":null,"task":"RNA compound-binding site prediction","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"RNA compound-binding prediction is evaluated at nucleotide level with independent test collections and a structural-generalization check.","summary_source_ids":["cobra-rna-binding-2026"],"summary_source_locator":"Methods: Dataset preparation; Evaluation metrics; Results: Structure-based split dataset evaluation; cached text lines 9–13, 35–36, 61–65","sections":[{"title":"Evaluation methodology","body":"HARIBOSS and TR60 training resources, with RB9, JL10, TL12 and TE18 reserved for testing. Combined HARIBOSS/TR60 data use an 80:10:10 internal partition; RB9, JL10, TL12 and TE18 remain external tests. RNABind structural splits define an additional retraining setting. Precision, recall, F1, MCC, AUROC and AUPRC. Comparisons include RNABind and other structure-aware methods in the structural-split evaluation. Sequences overlapping training and the four external tests are removed; structural-split retraining separately tests structure-level generalization.","source_ids":["cobra-rna-binding-2026"],"source_locator":"Methods: Dataset preparation; Evaluation metrics; Results: Structure-based split dataset evaluation; cached text lines 9–13, 35–36, 61–65"}],"facts":[{"label":"Datasets","value":"HARIBOSS and TR60 training resources, with RB9, JL10, TL12 and TE18 reserved for testing.","status":"source_checked","source_ids":["cobra-rna-binding-2026"],"source_locator":"Methods: Dataset preparation; Evaluation metrics; Results: Structure-based split dataset evaluation; cached text lines 9–13, 35–36, 61–65"},{"label":"Splits","value":"Combined HARIBOSS/TR60 data use an 80:10:10 internal partition; RB9, JL10, TL12 and TE18 remain external tests. RNABind structural splits define an additional retraining setting.","status":"source_checked","source_ids":["cobra-rna-binding-2026"],"source_locator":"Methods: Dataset preparation; Evaluation metrics; Results: Structure-based split dataset evaluation; cached text lines 9–13, 35–36, 61–65"},{"label":"Metrics","value":"Precision, recall, F1, MCC, AUROC and AUPRC.","status":"source_checked","source_ids":["cobra-rna-binding-2026"],"source_locator":"Methods: Dataset preparation; Evaluation metrics; Results: Structure-based split dataset evaluation; cached text lines 9–13, 35–36, 61–65"},{"label":"Baselines","value":"Comparisons include RNABind and other structure-aware methods in the structural-split evaluation.","status":"source_checked","source_ids":["cobra-rna-binding-2026"],"source_locator":"Methods: Dataset preparation; Evaluation metrics; Results: Structure-based split dataset evaluation; cached text lines 9–13, 35–36, 61–65"},{"label":"Leakage controls","value":"Sequences overlapping training and the four external tests are removed; structural-split retraining separately tests structure-level generalization.","status":"source_checked","source_ids":["cobra-rna-binding-2026"],"source_locator":"Methods: Dataset preparation; Evaluation metrics; Results: Structure-based split dataset evaluation; cached text lines 9–13, 35–36, 61–65"},{"label":"Uncertainty","value":"The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","status":"unreported","source_ids":["cobra-rna-binding-2026"],"source_locator":"Methods: Dataset preparation; Evaluation metrics; Results: Structure-based split dataset evaluation; cached text lines 9–13, 35–36, 61–65"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["cobra-rna-binding-2026"],"source_locator":"Methods: Dataset preparation; Evaluation metrics; Results: Structure-based split dataset evaluation; cached text lines 9–13, 35–36, 61–65"},{"label":"Organisms","value":"The evaluation pools PDB-derived RNA–ligand complexes from HARIBOSS, TR60, RB9, TL12, JL10 and TE18. Dataset preparation filters ligands, chain length and overlapping sequences, but does not enumerate the organism composition of each resulting subset.","status":"unreported","source_ids":["cobra-rna-binding-2026"],"source_locator":"Material and methods: Dataset preparation"},{"label":"Assays","value":"RNA compound-binding-site annotations.","status":"source_checked","source_ids":["cobra-rna-binding-2026"],"source_locator":"Methods: Dataset preparation; Evaluation metrics; Results: Structure-based split dataset evaluation; cached text lines 9–13, 35–36, 61–65"},{"label":"Allowed inputs","value":"RNA sequence only; explicit RNA structures are not required for CoBRA inference.","status":"source_checked","source_ids":["cobra-rna-binding-2026"],"source_locator":"Methods: Dataset preparation; Evaluation metrics; Results: Structure-based split dataset evaluation; cached text lines 9–13, 35–36, 61–65"},{"label":"Adaptation","value":"Frozen RNA-model residue embeddings feed a supervised MLP; language-model parameters are not updated.","status":"source_checked","source_ids":["cobra-rna-binding-2026"],"source_locator":"Methods: Dataset preparation; Evaluation metrics; Results: Structure-based split dataset evaluation; cached text lines 9–13, 35–36, 61–65"}],"strengths":[{"text":"Sequence-overlap removal protects the external tests; a separate structural-split experiment probes a different generalization boundary.","source_ids":["cobra-rna-binding-2026"],"source_locator":"Methods: Dataset preparation; Evaluation metrics; Results: Structure-based split dataset evaluation; cached text lines 9–13, 35–36, 61–65"}],"limitations":[{"text":"Metal and nonmetal compounds are both present. Sequence-only and structure-aware comparators use different input information.","source_ids":["cobra-rna-binding-2026"],"source_locator":"Methods: Dataset preparation; Evaluation metrics; Results: Structure-based split dataset evaluation; cached text lines 9–13, 35–36, 61–65"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: RNA sequence only; explicit RNA structures are not required for CoBRA inference.","Evaluation: Combined HARIBOSS/TR60 data use an 80:10:10 internal partition; RB9, JL10, TL12 and TE18 remain external tests. RNABind structural splits define an additional retraining setting.","Readout: Precision, recall, F1, MCC, AUROC and AUPRC."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["cobra-rna-binding-2026"],"source_locator":"Methods: Dataset preparation; Evaluation metrics; Results: Structure-based split dataset evaluation; cached text lines 9–13, 35–36, 61–65"},"coverage":"limited","gaps":["Uncertainty: The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","Organisms: The evaluation pools PDB-derived RNA–ligand complexes from HARIBOSS, TR60, RB9, TL12, JL10 and TE18. Dataset preparation filters ligands, chain length and overlapping sequences, but does not enumerate the organism composition of each resulting subset."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Targeted full-paper and supplement review of the outstanding task fields, with original dataset metadata checked where accessible. Source-scoped omissions are explicit; no independent benchmark reproduction or numerical-result change."}}}}
{"id":"reported-task-3d4dec23120fef","kind":"benchmark","name":"viral sequence detection","description":"","status":"needs_review","facets":{"areas":["microbes-communities"],"tasks":["viral sequence detection"]},"source_ids":["detire-viral-metagenomes-2023"],"links":[{"relation":"dataset","target_id":"reported-dataset-0b54f42a987b1d"}],"attributes":{"entity_level":"task","version":null,"task":"viral sequence detection","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Viral sequence detection evaluates a classifier on temporally separated reference-genome collections.","summary_source_ids":["detire-viral-metagenomes-2023"],"summary_source_locator":"Methods §§2.1, 2.3; cached text lines 11–12, 19–20; matching task comparison table/ablation captions","sections":[{"title":"Evaluation methodology","body":"NCBI viral RefSeq and prokaryotic-host reference sequences. The downstream classifier uses earlier records for training, an intermediate period for validation and later records for testing. Confusion-matrix measures include recall, accuracy, precision and F1. DeepVirFinder, PPR-Meta and CHEER on the temporal and CAMI marine test settings. The embedding stage uses the broader reference collection spanning the later evaluation period; downstream temporal separation alone is not an end-to-end temporal exclusion guarantee. The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","source_ids":["detire-viral-metagenomes-2023"],"source_locator":"Methods §§2.1, 2.3; cached text lines 11–12, 19–20; matching task comparison table/ablation captions"}],"facts":[{"label":"Datasets","value":"NCBI viral RefSeq and prokaryotic-host reference sequences.","status":"source_checked","source_ids":["detire-viral-metagenomes-2023"],"source_locator":"Methods §§2.1, 2.3; cached text lines 11–12, 19–20; matching task comparison table/ablation captions"},{"label":"Splits","value":"The downstream classifier uses earlier records for training, an intermediate period for validation and later records for testing.","status":"source_checked","source_ids":["detire-viral-metagenomes-2023"],"source_locator":"Methods §§2.1, 2.3; cached text lines 11–12, 19–20; matching task comparison table/ablation captions"},{"label":"Metrics","value":"Confusion-matrix measures include recall, accuracy, precision and F1.","status":"source_checked","source_ids":["detire-viral-metagenomes-2023"],"source_locator":"Methods §§2.1, 2.3; cached text lines 11–12, 19–20; matching task comparison table/ablation captions"},{"label":"Baselines","value":"DeepVirFinder, PPR-Meta and CHEER on the temporal and CAMI marine test settings.","status":"source_checked","source_ids":["detire-viral-metagenomes-2023"],"source_locator":"Methods §§2.1, 2.3; cached text lines 11–12, 19–20; matching task comparison table/ablation captions"},{"label":"Leakage controls","value":"The embedding stage uses the broader reference collection spanning the later evaluation period; downstream temporal separation alone is not an end-to-end temporal exclusion guarantee.","status":"source_checked","source_ids":["detire-viral-metagenomes-2023"],"source_locator":"Methods §§2.1, 2.3; cached text lines 11–12, 19–20; matching task comparison table/ablation captions"},{"label":"Uncertainty","value":"The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","status":"unreported","source_ids":["detire-viral-metagenomes-2023"],"source_locator":"Methods §§2.1, 2.3; cached text lines 11–12, 19–20; matching task comparison table/ablation captions"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["detire-viral-metagenomes-2023"],"source_locator":"Methods §§2.1, 2.3; cached text lines 11–12, 19–20; matching task comparison table/ablation captions"},{"label":"Organisms","value":"Viruses and prokaryotic hosts from NCBI reference sequences.","status":"source_checked","source_ids":["detire-viral-metagenomes-2023"],"source_locator":"Methods §§2.1, 2.3; cached text lines 11–12, 19–20; matching task comparison table/ablation captions"},{"label":"Assays","value":"Reference-derived viral/nonviral labels.","status":"source_checked","source_ids":["detire-viral-metagenomes-2023"],"source_locator":"Methods §§2.1, 2.3; cached text lines 11–12, 19–20; matching task comparison table/ablation captions"},{"label":"Allowed inputs","value":"Nucleotide sequence fragments.","status":"source_checked","source_ids":["detire-viral-metagenomes-2023"],"source_locator":"Methods §§2.1, 2.3; cached text lines 11–12, 19–20; matching task comparison table/ablation captions"},{"label":"Adaptation","value":"A downstream classifier uses temporal partitions; sequence embedding was trained on a broader corpus.","status":"source_checked","source_ids":["detire-viral-metagenomes-2023"],"source_locator":"Methods §§2.1, 2.3; cached text lines 11–12, 19–20; matching task comparison table/ablation captions"}],"strengths":[{"text":"Temporal classifier partitions expose distribution shift, while the source permits an explicit check of broader embedding-training overlap.","source_ids":["detire-viral-metagenomes-2023"],"source_locator":"Methods §§2.1, 2.3; cached text lines 11–12, 19–20; matching task comparison table/ablation captions"}],"limitations":[{"text":"This profile covers detection evaluation only. Exact reference overlap and comparator database versions remain unextracted.","source_ids":["detire-viral-metagenomes-2023"],"source_locator":"Methods §§2.1, 2.3; cached text lines 11–12, 19–20; matching task comparison table/ablation captions"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Nucleotide sequence fragments.","Evaluation: A downstream classifier uses temporal partitions; sequence embedding was trained on a broader corpus.","Readout: Confusion-matrix measures include recall, accuracy, precision and F1."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["detire-viral-metagenomes-2023"],"source_locator":"Methods §§2.1, 2.3; cached text lines 11–12, 19–20; matching task comparison table/ablation captions"},"coverage":"limited","gaps":["This profile covers detection evaluation only. Exact reference overlap and comparator database versions remain unextracted."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Task-specific computational methodology and field context checked in the cited primary-source artifact. Source-backed fields, inapplicable evaluator dimensions and unresolved details are distinguished. Numerical results were not reproduced."}}}}
{"id":"reported-task-4420dcdfe8338d","kind":"benchmark","name":"metagenomic genus classification","description":"","status":"needs_review","facets":{"areas":["microbes-communities"],"tasks":["metagenomic genus classification"]},"source_ids":["pc-mer-2024"],"links":[{"relation":"dataset","target_id":"reported-dataset-d28955d5872903"}],"attributes":{"entity_level":"task","version":null,"task":"metagenomic genus classification","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Genus classification compares sequence representations on separate simulated amplicon and shotgun datasets.","summary_source_ids":["pc-mer-2024"],"summary_source_locator":"Methods §3.1; Results/Conclusion; cached text lines 23–24, 48, 58","sections":[{"title":"Evaluation methodology","body":"Amplicon and shotgun collections derived from RDP version 11 update 5 (30 September 2016), with balanced taxonomic subsets and a separate unbalanced Qiita collection. The taxonomic label level and sequence-generation method must remain attached to each evaluation. Ten-fold cross-validation is reported for the benchmark experiments. Classification accuracy at multiple ranks; the genus task must be distinguished from easier higher-rank outcomes. Conventional classifiers using PC-mer are compared with CNN/DBN approaches using alternative encodings. The study evaluates supervised classifiers with ten-fold cross-validation across the amplicon and shotgun collections. Sections 2.2 and 3.1–3.5 do not specify genome-, species- or homology-grouped fold assignment; ten-fold evaluation alone does not establish unseen-species generalization.","source_ids":["pc-mer-2024"],"source_locator":"Results 3.1 Datasets; Methods §3.1; Results/Conclusion; cached text lines 23–24, 48, 58; Methods 2.2 Learning unit; Results 3.1 Datasets and 3.5.1 Training and testing procedure"}],"facts":[{"label":"Datasets","value":"Amplicon and shotgun collections derived from RDP version 11 update 5 (30 September 2016), with balanced taxonomic subsets and a separate unbalanced Qiita collection. The taxonomic label level and sequence-generation method must remain attached to each evaluation.","status":"source_checked","source_ids":["pc-mer-2024"],"source_locator":"Results 3.1 Datasets"},{"label":"Splits","value":"Ten-fold cross-validation is reported for the benchmark experiments.","status":"source_checked","source_ids":["pc-mer-2024"],"source_locator":"Methods §3.1; Results/Conclusion; cached text lines 23–24, 48, 58"},{"label":"Metrics","value":"Classification accuracy at multiple ranks; the genus task must be distinguished from easier higher-rank outcomes.","status":"source_checked","source_ids":["pc-mer-2024"],"source_locator":"Methods §3.1; Results/Conclusion; cached text lines 23–24, 48, 58"},{"label":"Baselines","value":"Conventional classifiers using PC-mer are compared with CNN/DBN approaches using alternative encodings.","status":"source_checked","source_ids":["pc-mer-2024"],"source_locator":"Methods §3.1; Results/Conclusion; cached text lines 23–24, 48, 58"},{"label":"Leakage controls","value":"The study evaluates supervised classifiers with ten-fold cross-validation across the amplicon and shotgun collections. Sections 2.2 and 3.1–3.5 do not specify genome-, species- or homology-grouped fold assignment; ten-fold evaluation alone does not establish unseen-species generalization.","status":"unreported","source_ids":["pc-mer-2024"],"source_locator":"Methods 2.2 Learning unit; Results 3.1 Datasets and 3.5.1 Training and testing procedure"},{"label":"Uncertainty","value":"The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","status":"unreported","source_ids":["pc-mer-2024"],"source_locator":"Methods §3.1; Results/Conclusion; cached text lines 23–24, 48, 58"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["pc-mer-2024"],"source_locator":"Methods §3.1; Results/Conclusion; cached text lines 23–24, 48, 58"},{"label":"Organisms","value":"Simulated microbial amplicon and shotgun communities.","status":"source_checked","source_ids":["pc-mer-2024"],"source_locator":"Methods §3.1; Results/Conclusion; cached text lines 23–24, 48, 58"},{"label":"Assays","value":"Genus-level taxonomic labels.","status":"source_checked","source_ids":["pc-mer-2024"],"source_locator":"Methods §3.1; Results/Conclusion; cached text lines 23–24, 48, 58"},{"label":"Allowed inputs","value":"Sequence-derived PC-mer or alternative feature encodings.","status":"source_checked","source_ids":["pc-mer-2024"],"source_locator":"Methods §3.1; Results/Conclusion; cached text lines 23–24, 48, 58"},{"label":"Adaptation","value":"Supervised classifiers evaluated through ten-fold cross-validation.","status":"source_checked","source_ids":["pc-mer-2024"],"source_locator":"Methods §3.1; Results/Conclusion; cached text lines 23–24, 48, 58"}],"strengths":[],"limitations":[{"text":"Ten-fold classification is not evidence of unseen-genome or unseen-species generalization without grouped partition information. Hardware also differs across cited runtime comparisons.","source_ids":["pc-mer-2024"],"source_locator":"Results 3.1 Datasets; Methods 2.2 Learning unit; Results 3.1 Datasets and 3.5.1 Training and testing procedure"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Sequence-derived PC-mer or alternative feature encodings.","Evaluation: Ten-fold cross-validation is reported for the benchmark experiments.","Readout: Classification accuracy at multiple ranks; the genus task must be distinguished from easier higher-rank outcomes."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["pc-mer-2024"],"source_locator":"Methods §3.1; Results/Conclusion; cached text lines 23–24, 48, 58"},"coverage":"limited","gaps":["Leakage controls: The study evaluates supervised classifiers with ten-fold cross-validation across the amplicon and shotgun collections. Sections 2.2 and 3.1–3.5 do not specify genome-, species- or homology-grouped fold assignment; ten-fold evaluation alone does not establish unseen-species generalization.","Uncertainty: The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Targeted full-paper and supplement review of the outstanding task fields, with original dataset metadata checked where accessible. Source-scoped omissions are explicit; no independent benchmark reproduction or numerical-result change."}}}}
{"id":"reported-task-45105e1c486251","kind":"benchmark","name":"CAMI II phylum read classification","description":"","status":"needs_review","facets":{"areas":["microbes-communities"],"tasks":["CAMI II phylum read classification"]},"source_ids":["ncd-metagenomics-2026"],"links":[{"relation":"dataset","target_id":"reported-dataset-beb4f5da29da0a"}],"attributes":{"entity_level":"task","version":null,"task":"CAMI II phylum read classification","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"CAMI II read classification evaluates a computationally limited subsample, with taxonomic-rank-specific interpretation.","summary_source_ids":["ncd-metagenomics-2026"],"summary_source_locator":"Methods: Evaluation protocol; Results: CAMI II; cached text lines 67–68, 113–116","sections":[{"title":"Evaluation methodology","body":"A 10,000-read subset of CAMI II Sample_0 from the human-microbiome collection. Queries are classified against the paper’s separately assembled metagenomic reference/training data. Macro-F1 is the arithmetic mean of classwise F1; only classes present in the evaluated subset enter the macro calculation. NCD-gzip and Kraken2. The CAMI II test uses 10,000 reads from Sample_0 and the paper’s RefSeq metagenomic training collection. A disjoint genome split is described for the separate simulated RefSeq experiment; the CAMI passage does not establish reference-genome or homology exclusion against this external sample.","source_ids":["ncd-metagenomics-2026"],"source_locator":"Methods: Evaluation protocol; Results: CAMI II; cached text lines 67–68, 113–116; Methods: Datasets/Metagenomic reads; Results: CAMI dataset"}],"facts":[{"label":"Datasets","value":"A 10,000-read subset of CAMI II Sample_0 from the human-microbiome collection.","status":"source_checked","source_ids":["ncd-metagenomics-2026"],"source_locator":"Methods: Evaluation protocol; Results: CAMI II; cached text lines 67–68, 113–116"},{"label":"Splits","value":"Queries are classified against the paper’s separately assembled metagenomic reference/training data.","status":"source_checked","source_ids":["ncd-metagenomics-2026"],"source_locator":"Methods: Evaluation protocol; Results: CAMI II; cached text lines 67–68, 113–116"},{"label":"Metrics","value":"Macro-F1 is the arithmetic mean of classwise F1; only classes present in the evaluated subset enter the macro calculation.","status":"source_checked","source_ids":["ncd-metagenomics-2026"],"source_locator":"Methods: Evaluation protocol; Results: CAMI II; cached text lines 67–68, 113–116"},{"label":"Baselines","value":"NCD-gzip and Kraken2.","status":"source_checked","source_ids":["ncd-metagenomics-2026"],"source_locator":"Methods: Evaluation protocol; Results: CAMI II; cached text lines 67–68, 113–116"},{"label":"Leakage controls","value":"The CAMI II test uses 10,000 reads from Sample_0 and the paper’s RefSeq metagenomic training collection. A disjoint genome split is described for the separate simulated RefSeq experiment; the CAMI passage does not establish reference-genome or homology exclusion against this external sample.","status":"unreported","source_ids":["ncd-metagenomics-2026"],"source_locator":"Methods: Datasets/Metagenomic reads; Results: CAMI dataset"},{"label":"Uncertainty","value":"Table 5 describes a single 10,000-read CAMI II subsample. It does not provide confidence intervals, repeated-subsample variation or random-seed uncertainty for the phylum metrics.","status":"unreported","source_ids":["ncd-metagenomics-2026"],"source_locator":"Results: CAMI dataset; Table 5"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["ncd-metagenomics-2026"],"source_locator":"Methods: Evaluation protocol; Results: CAMI II; cached text lines 67–68, 113–116"},{"label":"Organisms","value":"CAMI II human-microbiome community taxa.","status":"source_checked","source_ids":["ncd-metagenomics-2026"],"source_locator":"Methods: Evaluation protocol; Results: CAMI II; cached text lines 67–68, 113–116"},{"label":"Assays","value":"Read-origin taxonomy at the selected evaluation rank.","status":"source_checked","source_ids":["ncd-metagenomics-2026"],"source_locator":"Methods: Evaluation protocol; Results: CAMI II; cached text lines 67–68, 113–116"},{"label":"Allowed inputs","value":"DNA reads and a separately assembled reference/training collection.","status":"source_checked","source_ids":["ncd-metagenomics-2026"],"source_locator":"Methods: Evaluation protocol; Results: CAMI II; cached text lines 67–68, 113–116"},{"label":"Adaptation","value":"Reference-based classification using NCD-gzip or Kraken2.","status":"source_checked","source_ids":["ncd-metagenomics-2026"],"source_locator":"Methods: Evaluation protocol; Results: CAMI II; cached text lines 67–68, 113–116"}],"strengths":[],"limitations":[{"text":"The subsample contains only Bacteria at superkingdom level, so that result does not establish broad multiclass discrimination. Phylum and superkingdom are separate catalogue outcomes.","source_ids":["ncd-metagenomics-2026"],"source_locator":"Methods: Evaluation protocol; Results: CAMI II; cached text lines 67–68, 113–116"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: DNA reads and a separately assembled reference/training collection.","Evaluation: Queries are classified against the paper’s separately assembled metagenomic reference/training data.","Readout: Macro-F1 is the arithmetic mean of classwise F1; only classes present in the evaluated subset enter the macro calculation."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["ncd-metagenomics-2026"],"source_locator":"Methods: Evaluation protocol; Results: CAMI II; cached text lines 67–68, 113–116"},"coverage":"limited","gaps":["Leakage controls: The CAMI II test uses 10,000 reads from Sample_0 and the paper’s RefSeq metagenomic training collection. A disjoint genome split is described for the separate simulated RefSeq experiment; the CAMI passage does not establish reference-genome or homology exclusion against this external sample.","Uncertainty: Table 5 describes a single 10,000-read CAMI II subsample. It does not provide confidence intervals, repeated-subsample variation or random-seed uncertainty for the phylum metrics."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Targeted full-paper and supplement review of the outstanding task fields, with original dataset metadata checked where accessible. Source-scoped omissions are explicit; no independent benchmark reproduction or numerical-result change."}}}}
{"id":"reported-task-45ead9a1eddf8d","kind":"benchmark","name":"antigen-antibody HIV neutralization prediction","description":"","status":"needs_review","facets":{"areas":["molecular-interactions"],"tasks":["antigen-antibody HIV neutralization prediction"]},"source_ids":["deepinteraware-2025"],"links":[{"relation":"dataset","target_id":"reported-dataset-50f0bdb7cf9ca4"}],"attributes":{"entity_level":"task","version":null,"task":"antigen-antibody HIV neutralization prediction","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Antibody neutralization classification is one of several distinct evaluations in the DeepInterAware paper.","summary_source_ids":["deepinteraware-2025"],"summary_source_locator":"Experimental Setting—Datasets; Model Evaluation Metrics; cached text lines 85–93; matching task comparison table/ablation captions","sections":[{"title":"Evaluation methodology","body":"The HIV dataset is used for neutralization prediction; other datasets address binding, binding sites and free-energy changes. The HIV neutralization table separates antibody-unseen evaluation; exact partition construction is delegated to Supporting Information §S1.1. AUROC, AUPRC and MCC are primary metrics for binding and neutralization classification. The HIV neutralization table compares DrugBAN, ESM2AbLang, ESM2AntiBERTy, ResPPI, PIPR, AbAgIntPre, MasonsCNN and DeepAAI. HIV pairs with antigen and antibody homology above 0.9 are filtered. Ab Unseen holds out 15% of antibodies; Ag Unseen holds out 20% of antigens. Ag&Ab Unseen additionally removes from training all pairs involving test antibodies. The other partner is deliberately shared in each single-partner holdout.","source_ids":["deepinteraware-2025","evidence-task-final-a-deepinteraware-2025-advs-12-2412533-s001-pdf"],"source_locator":"Experimental Setting—Datasets; Model Evaluation Metrics; cached text lines 85–93; matching task comparison table/ablation captions; Supporting Information S1.1, HIV dataset and three neutralization scenarios"}],"facts":[{"label":"Datasets","value":"The HIV dataset is used for neutralization prediction; other datasets address binding, binding sites and free-energy changes.","status":"source_checked","source_ids":["deepinteraware-2025"],"source_locator":"Experimental Setting—Datasets; Model Evaluation Metrics; cached text lines 85–93; matching task comparison table/ablation captions"},{"label":"Splits","value":"The HIV neutralization table separates antibody-unseen evaluation; exact partition construction is delegated to Supporting Information §S1.1.","status":"source_checked","source_ids":["deepinteraware-2025"],"source_locator":"Experimental Setting—Datasets; Model Evaluation Metrics; cached text lines 85–93; matching task comparison table/ablation captions"},{"label":"Metrics","value":"AUROC, AUPRC and MCC are primary metrics for binding and neutralization classification.","status":"source_checked","source_ids":["deepinteraware-2025"],"source_locator":"Experimental Setting—Datasets; Model Evaluation Metrics; cached text lines 85–93; matching task comparison table/ablation captions"},{"label":"Baselines","value":"The HIV neutralization table compares DrugBAN, ESM2AbLang, ESM2AntiBERTy, ResPPI, PIPR, AbAgIntPre, MasonsCNN and DeepAAI.","status":"source_checked","source_ids":["deepinteraware-2025"],"source_locator":"Experimental Setting—Datasets; Model Evaluation Metrics; cached text lines 85–93; matching task comparison table/ablation captions"},{"label":"Leakage controls","value":"HIV pairs with antigen and antibody homology above 0.9 are filtered. Ab Unseen holds out 15% of antibodies; Ag Unseen holds out 20% of antigens. Ag&Ab Unseen additionally removes from training all pairs involving test antibodies. The other partner is deliberately shared in each single-partner holdout.","status":"source_checked","source_ids":["deepinteraware-2025","evidence-task-final-a-deepinteraware-2025-advs-12-2412533-s001-pdf"],"source_locator":"Supporting Information S1.1, HIV dataset and three neutralization scenarios"},{"label":"Uncertainty","value":"Each HIV scenario is repeated with five distinct random seeds and dataset splits. Supporting Information S1.1 calls the reported summaries means and variances; this wording is preserved rather than silently interpreting the table’s ± terms as 95% confidence intervals.","status":"source_checked","source_ids":["deepinteraware-2025","evidence-task-final-a-deepinteraware-2025-advs-12-2412533-s001-pdf"],"source_locator":"Supporting Information S1.1 HIV; Table 2"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["deepinteraware-2025"],"source_locator":"Experimental Setting—Datasets; Model Evaluation Metrics; cached text lines 85–93; matching task comparison table/ablation captions"},{"label":"Organisms","value":"HIV neutralization collection for this task.","status":"source_checked","source_ids":["deepinteraware-2025"],"source_locator":"Experimental Setting—Datasets; Model Evaluation Metrics; cached text lines 85–93; matching task comparison table/ablation captions"},{"label":"Assays","value":"Antibody neutralization outcomes; other paper tasks use distinct binding/free-energy labels.","status":"source_checked","source_ids":["deepinteraware-2025"],"source_locator":"Experimental Setting—Datasets; Model Evaluation Metrics; cached text lines 85–93; matching task comparison table/ablation captions"},{"label":"Allowed inputs","value":"Antibody–antigen pair representations.","status":"source_checked","source_ids":["deepinteraware-2025"],"source_locator":"Experimental Setting—Datasets; Model Evaluation Metrics; cached text lines 85–93; matching task comparison table/ablation captions"},{"label":"Adaptation","value":"Supervised antibody–antigen neutralization classification; the antibody-unseen test is identified separately from binding and transfer tasks.","status":"source_checked","source_ids":["deepinteraware-2025"],"source_locator":"Experimental Setting—Datasets; Model Evaluation Metrics; cached text lines 85–93; matching task comparison table/ablation captions"}],"strengths":[],"limitations":[{"text":"Ab Unseen, Ag Unseen and joint holdouts ask different transfer questions. Partner sharing is intentional in the single-partner scenarios and must remain visible in comparisons.","source_ids":["deepinteraware-2025","evidence-task-final-a-deepinteraware-2025-advs-12-2412533-s001-pdf"],"source_locator":"Supporting Information S1.1, HIV dataset and three neutralization scenarios; Supporting Information S1.1 HIV; Table 2"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Antibody–antigen pair representations.","Evaluation: The HIV neutralization table separates antibody-unseen evaluation; exact partition construction is delegated to Supporting Information §S1.1.","Readout: AUROC, AUPRC and MCC are primary metrics for binding and neutralization classification."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["deepinteraware-2025"],"source_locator":"Experimental Setting—Datasets; Model Evaluation Metrics; cached text lines 85–93; matching task comparison table/ablation captions"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Targeted full-paper and supplement review of the outstanding task fields, with original dataset metadata checked where accessible. Source-scoped omissions are explicit; no independent benchmark reproduction or numerical-result change."}}}}
{"id":"reported-task-46e927bea10702","kind":"benchmark","name":"miRNA-mRNA interaction prediction","description":"","status":"needs_review","facets":{"areas":["rna-transcriptomes"],"tasks":["miRNA-mRNA interaction prediction"]},"source_ids":["rnaret-2026"],"links":[{"relation":"dataset","target_id":"reported-dataset-99afd0c86b2954"}],"attributes":{"entity_level":"task","version":null,"task":"miRNA-mRNA interaction prediction","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"miRNA–mRNA interaction classification evaluates paired sequence representations on a benchmark split and an independent collection.","summary_source_ids":["rnaret-2026"],"summary_source_locator":"Results: RNA–RNA interaction task; Methods: RNA–RNA interaction prediction; cached text lines 17, 79; task metric definitions and corresponding results table; matching task comparison table/ablation captions","sections":[{"title":"Evaluation methodology","body":"MirTarRAW positive/negative miRNA–mRNA pairs, plus DeepMirTarLeft. MirTarRAW is partitioned 72:8:20 for training, validation and testing; DeepMirTarLeft is an additional independent dataset. F1, precision, recall, accuracy and AUC for miRNA–mRNA interaction classification. DeepMirTar, RNA-FM, RNABERT, RNA-MSM and RNAErnie in the interaction comparison; other RNA-task baselines are separate. MirTarRAW is divided 72%/8%/20% for training/validation/test, with DeepMirTarLeft as an additional test set. The RNAret methods do not specify a miRNA-identity, transcript-identity or homology-grouped holdout for this task. The paper’s explicit RNAStrAlign/ArchiveII overlap exclusion belongs to secondary-structure prediction and must not be transferred here.","source_ids":["rnaret-2026"],"source_locator":"Results: RNA–RNA interaction task; Methods: RNA–RNA interaction prediction; cached text lines 17, 79; task metric definitions and corresponding results table; matching task comparison table/ablation captions; Results: miRNA-mRNA interaction prediction; Methods: Statistics and reproducibility"}],"facts":[{"label":"Datasets","value":"MirTarRAW positive/negative miRNA–mRNA pairs, plus DeepMirTarLeft.","status":"source_checked","source_ids":["rnaret-2026"],"source_locator":"Results: RNA–RNA interaction task; Methods: RNA–RNA interaction prediction; cached text lines 17, 79; task metric definitions and corresponding results table; matching task comparison table/ablation captions"},{"label":"Splits","value":"MirTarRAW is partitioned 72:8:20 for training, validation and testing; DeepMirTarLeft is an additional independent dataset.","status":"source_checked","source_ids":["rnaret-2026"],"source_locator":"Results: RNA–RNA interaction task; Methods: RNA–RNA interaction prediction; cached text lines 17, 79; task metric definitions and corresponding results table; matching task comparison table/ablation captions"},{"label":"Metrics","value":"F1, precision, recall, accuracy and AUC for miRNA–mRNA interaction classification.","status":"source_checked","source_ids":["rnaret-2026"],"source_locator":"Results: RNA–RNA interaction task; Methods: RNA–RNA interaction prediction; cached text lines 17, 79; task metric definitions and corresponding results table; matching task comparison table/ablation captions"},{"label":"Baselines","value":"DeepMirTar, RNA-FM, RNABERT, RNA-MSM and RNAErnie in the interaction comparison; other RNA-task baselines are separate.","status":"source_checked","source_ids":["rnaret-2026"],"source_locator":"Results: RNA–RNA interaction task; Methods: RNA–RNA interaction prediction; cached text lines 17, 79; task metric definitions and corresponding results table; matching task comparison table/ablation captions"},{"label":"Leakage controls","value":"MirTarRAW is divided 72%/8%/20% for training/validation/test, with DeepMirTarLeft as an additional test set. The RNAret methods do not specify a miRNA-identity, transcript-identity or homology-grouped holdout for this task. The paper’s explicit RNAStrAlign/ArchiveII overlap exclusion belongs to secondary-structure prediction and must not be transferred here.","status":"unreported","source_ids":["rnaret-2026"],"source_locator":"Results: miRNA-mRNA interaction prediction; Methods: Statistics and reproducibility"},{"label":"Uncertainty","value":"Table 1 reports point classification metrics. RNAret’s Statistics and reproducibility section and Supplementary Information do not specify repeated-run intervals or an uncertainty estimator for miRNA–mRNA prediction. Intervals reported in the original miTAR study would not quantify RNAret’s results.","status":"unreported","source_ids":["rnaret-2026","evidence-task-final-a-rnaret-2026-42003-2026-9757-moesm2-esm-pdf"],"source_locator":"Table 1; Methods: Statistics and reproducibility; Supplementary Information"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["rnaret-2026"],"source_locator":"Results: RNA–RNA interaction task; Methods: RNA–RNA interaction prediction; cached text lines 17, 79; task metric definitions and corresponding results table; matching task comparison table/ablation captions"},{"label":"Organisms","value":"Human miRNA–target pairs. The original miTAR study identifies its DeepMirTar and miRAW inputs as human datasets; MirTarRAW combines portions of these two datasets and DeepMirTarLeft is the withheld remainder of DeepMirTar.","status":"source_checked","source_ids":["rnaret-2026","evidence-task-final-a-pmc7912887-xml"],"source_locator":"RNAret: miRNA-mRNA interaction prediction; miTAR: Abstract and Methods/Datasets"},{"label":"Assays","value":"miRNA–mRNA interaction labels.","status":"source_checked","source_ids":["rnaret-2026"],"source_locator":"Results: RNA–RNA interaction task; Methods: RNA–RNA interaction prediction; cached text lines 17, 79; task metric definitions and corresponding results table; matching task comparison table/ablation captions"},{"label":"Allowed inputs","value":"Paired miRNA and mRNA sequences.","status":"source_checked","source_ids":["rnaret-2026"],"source_locator":"Results: RNA–RNA interaction task; Methods: RNA–RNA interaction prediction; cached text lines 17, 79; task metric definitions and corresponding results table; matching task comparison table/ablation captions"},{"label":"Adaptation","value":"Supervised pair classification on the training portion with validation and independent testing.","status":"source_checked","source_ids":["rnaret-2026"],"source_locator":"Results: RNA–RNA interaction task; Methods: RNA–RNA interaction prediction; cached text lines 17, 79; task metric definitions and corresponding results table; matching task comparison table/ablation captions"}],"strengths":[],"limitations":[{"text":"A pair-level partition does not establish unseen-miRNA or unseen-transcript performance. RNAret’s secondary-structure overlap exclusions and another method’s uncertainty estimates do not apply to this interaction result.","source_ids":["rnaret-2026","evidence-task-final-a-rnaret-2026-42003-2026-9757-moesm2-esm-pdf","evidence-task-final-a-pmc7912887-xml"],"source_locator":"Results: miRNA-mRNA interaction prediction; Methods: Statistics and reproducibility; Table 1; Methods: Statistics and reproducibility; Supplementary Information; RNAret: miRNA-mRNA interaction prediction; miTAR: Abstract and Methods/Datasets"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Paired miRNA and mRNA sequences.","Evaluation: MirTarRAW is partitioned 72:8:20 for training, validation and testing; DeepMirTarLeft is an additional independent dataset.","Readout: F1, precision, recall, accuracy and AUC for miRNA–mRNA interaction classification."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["rnaret-2026"],"source_locator":"Results: RNA–RNA interaction task; Methods: RNA–RNA interaction prediction; cached text lines 17, 79; task metric definitions and corresponding results table; matching task comparison table/ablation captions"},"coverage":"limited","gaps":["Leakage controls: MirTarRAW is divided 72%/8%/20% for training/validation/test, with DeepMirTarLeft as an additional test set. The RNAret methods do not specify a miRNA-identity, transcript-identity or homology-grouped holdout for this task. The paper’s explicit RNAStrAlign/ArchiveII overlap exclusion belongs to secondary-structure prediction and must not be transferred here.","Uncertainty: Table 1 reports point classification metrics. RNAret’s Statistics and reproducibility section and Supplementary Information do not specify repeated-run intervals or an uncertainty estimator for miRNA–mRNA prediction. Intervals reported in the original miTAR study would not quantify RNAret’s results."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Targeted full-paper and supplement review of the outstanding task fields, with original dataset metadata checked where accessible. Source-scoped omissions are explicit; no independent benchmark reproduction or numerical-result change."}}}}
{"id":"reported-task-47465954d606e6","kind":"benchmark","name":"vaccine-antigen candidate prediction","description":"","status":"needs_review","facets":{"areas":["proteins-complexes"],"tasks":["vaccine-antigen candidate prediction"]},"source_ids":["vaxign-esm-2024"],"links":[{"relation":"dataset","target_id":"reported-dataset-b91c871eb7740a"}],"attributes":{"entity_level":"task","version":null,"task":"vaccine-antigen candidate prediction","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Vaccine-antigen classification tests transfer by withholding one organism’s labeled protein examples.","summary_source_ids":["vaxign-esm-2024"],"summary_source_locator":"Methods: leave-one-pathogen-out validation; Performance evaluation; Results: long-protein coverage; cached text lines 18, 25–26, 39","sections":[{"title":"Evaluation methodology","body":"Previously curated vaccine-antigen positive/negative datasets used by Vaxign-ML and Vaxign-DL. Leave-one-pathogen-out evaluation trains on the remaining organisms and predicts the withheld organism’s labels. Accuracy, sensitivity, specificity, weighted F1, MCC, AUROC and AUPRC. Vaxign-ML and Vaxign-DL are the referenced baseline datasets/methods. Dataset curation removes positive-protein homologs above 30% sequence similarity and selects negative proteins below 30% similarity to positives. The LOPOV experiment additionally withholds one of ten pathogens. These controls do not establish that sequences were absent from ESM-1b pretraining.","source_ids":["vaxign-esm-2024"],"source_locator":"Methods: leave-one-pathogen-out validation; Performance evaluation; Results: long-protein coverage; cached text lines 18, 25–26, 39; Methods: Collection of Positive and Negative protein sequences; Leave-one-pathogen-out validation"}],"facts":[{"label":"Datasets","value":"Previously curated vaccine-antigen positive/negative datasets used by Vaxign-ML and Vaxign-DL.","status":"source_checked","source_ids":["vaxign-esm-2024"],"source_locator":"Methods: leave-one-pathogen-out validation; Performance evaluation; Results: long-protein coverage; cached text lines 18, 25–26, 39"},{"label":"Splits","value":"Leave-one-pathogen-out evaluation trains on the remaining organisms and predicts the withheld organism’s labels.","status":"source_checked","source_ids":["vaxign-esm-2024"],"source_locator":"Methods: leave-one-pathogen-out validation; Performance evaluation; Results: long-protein coverage; cached text lines 18, 25–26, 39"},{"label":"Metrics","value":"Accuracy, sensitivity, specificity, weighted F1, MCC, AUROC and AUPRC.","status":"source_checked","source_ids":["vaxign-esm-2024"],"source_locator":"Methods: leave-one-pathogen-out validation; Performance evaluation; Results: long-protein coverage; cached text lines 18, 25–26, 39"},{"label":"Baselines","value":"Vaxign-ML and Vaxign-DL are the referenced baseline datasets/methods.","status":"source_checked","source_ids":["vaxign-esm-2024"],"source_locator":"Methods: leave-one-pathogen-out validation; Performance evaluation; Results: long-protein coverage; cached text lines 18, 25–26, 39"},{"label":"Leakage controls","value":"Dataset curation removes positive-protein homologs above 30% sequence similarity and selects negative proteins below 30% similarity to positives. The LOPOV experiment additionally withholds one of ten pathogens. These controls do not establish that sequences were absent from ESM-1b pretraining.","status":"source_checked","source_ids":["vaxign-esm-2024"],"source_locator":"Methods: Collection of Positive and Negative protein sequences; Leave-one-pathogen-out validation"},{"label":"Uncertainty","value":"Tables 1–3 include ± terms for most metrics but not AUROC. Their captions and the Deep learning pipeline section do not define whether those terms are standard deviations, standard errors or confidence intervals, or specify their repetition count.","status":"unreported","source_ids":["vaxign-esm-2024"],"source_locator":"Methods: Deep learning pipeline; Tables 1–3"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["vaxign-esm-2024"],"source_locator":"Methods: leave-one-pathogen-out validation; Performance evaluation; Results: long-protein coverage; cached text lines 18, 25–26, 39"},{"label":"Organisms","value":"Multiple pathogen species in the curated antigen evaluation.","status":"source_checked","source_ids":["vaxign-esm-2024"],"source_locator":"Methods: leave-one-pathogen-out validation; Performance evaluation; Results: long-protein coverage; cached text lines 18, 25–26, 39"},{"label":"Assays","value":"Vaccine-antigen positive/negative annotations.","status":"source_checked","source_ids":["vaxign-esm-2024"],"source_locator":"Methods: leave-one-pathogen-out validation; Performance evaluation; Results: long-protein coverage; cached text lines 18, 25–26, 39"},{"label":"Allowed inputs","value":"Protein sequence representations.","status":"source_checked","source_ids":["vaxign-esm-2024"],"source_locator":"Methods: leave-one-pathogen-out validation; Performance evaluation; Results: long-protein coverage; cached text lines 18, 25–26, 39"},{"label":"Adaptation","value":"Supervised candidate classification with leave-one-pathogen-out evaluation.","status":"source_checked","source_ids":["vaxign-esm-2024"],"source_locator":"Methods: leave-one-pathogen-out validation; Performance evaluation; Results: long-protein coverage; cached text lines 18, 25–26, 39"}],"strengths":[],"limitations":[{"text":"Long proteins may be omitted or truncated, changing evaluation coverage.","source_ids":["vaxign-esm-2024"],"source_locator":"Methods: leave-one-pathogen-out validation; Performance evaluation; Results: long-protein coverage; cached text lines 18, 25–26, 39"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Protein sequence representations.","Evaluation: Leave-one-pathogen-out evaluation trains on the remaining organisms and predicts the withheld organism’s labels.","Readout: Accuracy, sensitivity, specificity, weighted F1, MCC, AUROC and AUPRC."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["vaxign-esm-2024"],"source_locator":"Methods: leave-one-pathogen-out validation; Performance evaluation; Results: long-protein coverage; cached text lines 18, 25–26, 39"},"coverage":"limited","gaps":["Uncertainty: Tables 1–3 include ± terms for most metrics but not AUROC. Their captions and the Deep learning pipeline section do not define whether those terms are standard deviations, standard errors or confidence intervals, or specify their repetition count."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Targeted full-paper and supplement review of the outstanding task fields, with original dataset metadata checked where accessible. Source-scoped omissions are explicit; no independent benchmark reproduction or numerical-result change."}}}}
{"id":"reported-task-4a54ce01b5a855","kind":"benchmark","name":"unseen-species genus classification","description":"","status":"needs_review","facets":{"areas":["dna-genomes"],"tasks":["unseen-species genus classification"]},"source_ids":["barcodebert-2026"],"links":[{"relation":"dataset","target_id":"reported-dataset-bc127dc9c441fe"}],"attributes":{"entity_level":"task","version":null,"task":"unseen-species genus classification","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Genus classification tests whether frozen DNA-barcode embeddings transfer to species absent from the reference partition.","summary_source_ids":["barcodebert-2026"],"summary_source_locator":"Experimental setup §§4.1.1–4.1.4; Table 1; cached text lines 35–50, 60","sections":[{"title":"Evaluation methodology","body":"DNA barcodes organized into Seen and Unseen species partitions, with genus labels. One-nearest-neighbour genus prediction uses the Seen training subset as reference and the Unseen species partition as queries. Genus-level classification accuracy with cosine-similarity nearest-neighbour retrieval. BLAST and several DNA language-model embeddings are compared in the paper. Unseen species are excluded from the reference partition; this task is distinct from fine-tuned closed-world species classification. The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","source_ids":["barcodebert-2026"],"source_locator":"Experimental setup §§4.1.1–4.1.4; Table 1; cached text lines 35–50, 60"}],"facts":[{"label":"Datasets","value":"DNA barcodes organized into Seen and Unseen species partitions, with genus labels.","status":"source_checked","source_ids":["barcodebert-2026"],"source_locator":"Experimental setup §§4.1.1–4.1.4; Table 1; cached text lines 35–50, 60"},{"label":"Splits","value":"One-nearest-neighbour genus prediction uses the Seen training subset as reference and the Unseen species partition as queries.","status":"source_checked","source_ids":["barcodebert-2026"],"source_locator":"Experimental setup §§4.1.1–4.1.4; Table 1; cached text lines 35–50, 60"},{"label":"Metrics","value":"Genus-level classification accuracy with cosine-similarity nearest-neighbour retrieval.","status":"source_checked","source_ids":["barcodebert-2026"],"source_locator":"Experimental setup §§4.1.1–4.1.4; Table 1; cached text lines 35–50, 60"},{"label":"Baselines","value":"BLAST and several DNA language-model embeddings are compared in the paper.","status":"source_checked","source_ids":["barcodebert-2026"],"source_locator":"Experimental setup §§4.1.1–4.1.4; Table 1; cached text lines 35–50, 60"},{"label":"Leakage controls","value":"Unseen species are excluded from the reference partition; this task is distinct from fine-tuned closed-world species classification.","status":"source_checked","source_ids":["barcodebert-2026"],"source_locator":"Experimental setup §§4.1.1–4.1.4; Table 1; cached text lines 35–50, 60"},{"label":"Uncertainty","value":"The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","status":"unreported","source_ids":["barcodebert-2026"],"source_locator":"Experimental setup §§4.1.1–4.1.4; Table 1; cached text lines 35–50, 60"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["barcodebert-2026"],"source_locator":"Experimental setup §§4.1.1–4.1.4; Table 1; cached text lines 35–50, 60"},{"label":"Organisms","value":"Seen and Unseen species partitions with genus-level labels.","status":"source_checked","source_ids":["barcodebert-2026"],"source_locator":"Experimental setup §§4.1.1–4.1.4; Table 1; cached text lines 35–50, 60"},{"label":"Assays","value":"DNA barcode sequence/taxonomic annotations.","status":"source_checked","source_ids":["barcodebert-2026"],"source_locator":"Experimental setup §§4.1.1–4.1.4; Table 1; cached text lines 35–50, 60"},{"label":"Allowed inputs","value":"DNA barcodes represented as frozen model embeddings.","status":"source_checked","source_ids":["barcodebert-2026"],"source_locator":"Experimental setup §§4.1.1–4.1.4; Table 1; cached text lines 35–50, 60"},{"label":"Adaptation","value":"One-nearest-neighbour retrieval from Seen training examples; no task-head training for this genus task.","status":"source_checked","source_ids":["barcodebert-2026"],"source_locator":"Experimental setup §§4.1.1–4.1.4; Table 1; cached text lines 35–50, 60"}],"strengths":[{"text":"Species holdout directly tests genus retrieval beyond the reference species.","source_ids":["barcodebert-2026"],"source_locator":"Experimental setup §§4.1.1–4.1.4; Table 1; cached text lines 35–50, 60"}],"limitations":[{"text":"Species exclusion establishes the task split, not complete exclusion from every comparator’s pretraining corpus.","source_ids":["barcodebert-2026"],"source_locator":"Experimental setup §§4.1.1–4.1.4; Table 1; cached text lines 35–50, 60"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: DNA barcodes represented as frozen model embeddings.","Evaluation: One-nearest-neighbour retrieval from Seen training examples; no task-head training for this genus task.","Readout: Genus-level classification accuracy with cosine-similarity nearest-neighbour retrieval."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["barcodebert-2026"],"source_locator":"Experimental setup §§4.1.1–4.1.4; Table 1; cached text lines 35–50, 60"},"coverage":"limited","gaps":["Species exclusion establishes the task split, not complete exclusion from every comparator’s pretraining corpus."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Task-specific computational methodology and field context checked in the cited primary-source artifact. Source-backed fields, inapplicable evaluator dimensions and unresolved details are distinguished. Numerical results were not reproduced."}}}}
{"id":"reported-task-4df1fb456d3deb","kind":"benchmark","name":"Cell-type structure in frozen embeddings","description":"","status":"needs_review","facets":{"areas":["cells-tissues"],"tasks":["Cell-type structure in frozen embeddings"]},"source_ids":["genept-2024"],"links":[{"relation":"dataset","target_id":"reported-dataset-eaa2965545c87b"}],"attributes":{"entity_level":"task","version":null,"task":"Cell-type structure in frozen embeddings","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Frozen cell embeddings are examined for cell structure and patient-associated batch patterns.","summary_source_ids":["genept-2024"],"summary_source_locator":"Results: cell clustering and batch effects; Discussion; cached text lines 57, 61","sections":[{"title":"Evaluation methodology","body":"The Aorta single-cell dataset is one analyzed collection, with patient and phenotype annotations. The frozen-embedding clustering analysis does not define a supervised train/test split. Random holdout used for phenotype-classifier experiments belongs to a different evaluation. Adjusted Rand index compares embedding-derived clusters with patient and phenotype labels. Geneformer, GenePT-s, scGPT and original expression measurements. The clustering comparison measures cell and patient associations in embeddings; it does not establish an independent unseen-patient supervised test.","source_ids":["genept-2024"],"source_locator":"Results: cell clustering and batch effects; Discussion; cached text lines 57, 61"}],"facts":[{"label":"Datasets","value":"The Aorta single-cell dataset is one analyzed collection, with patient and phenotype annotations.","status":"source_checked","source_ids":["genept-2024"],"source_locator":"Results: cell clustering and batch effects; Discussion; cached text lines 57, 61"},{"label":"Splits","value":"The frozen-embedding clustering analysis does not define a supervised train/test split. Random holdout used for phenotype-classifier experiments belongs to a different evaluation.","status":"unreported","source_ids":["genept-2024"],"source_locator":"Results: cell clustering and batch effects; Discussion; cached text lines 57, 61"},{"label":"Metrics","value":"Adjusted Rand index compares embedding-derived clusters with patient and phenotype labels.","status":"source_checked","source_ids":["genept-2024"],"source_locator":"Results: cell clustering and batch effects; Discussion; cached text lines 57, 61"},{"label":"Baselines","value":"Geneformer, GenePT-s, scGPT and original expression measurements.","status":"source_checked","source_ids":["genept-2024"],"source_locator":"Results: cell clustering and batch effects; Discussion; cached text lines 57, 61"},{"label":"Leakage controls","value":"The clustering comparison measures cell and patient associations in embeddings; it does not establish an independent unseen-patient supervised test.","status":"unreported","source_ids":["genept-2024"],"source_locator":"Results: cell clustering and batch effects; Discussion; cached text lines 57, 61"},{"label":"Uncertainty","value":"The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","status":"unreported","source_ids":["genept-2024"],"source_locator":"Results: cell clustering and batch effects; Discussion; cached text lines 57, 61"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["genept-2024"],"source_locator":"Results: cell clustering and batch effects; Discussion; cached text lines 57, 61"},{"label":"Organisms","value":"Human tissue cohorts are explicitly established for the aorta, arterial plaque, osteoarthritis cartilage and multiple-sclerosis brain datasets by their original studies. GenePT also evaluates pancreas and tumor-infiltrating myeloid collections. A complete organism-by-subset census is not supplied in the GenePT cell-task description.","status":"source_checked","source_ids":["genept-2024","evidence-task-final-a-pmc9556750-xml","evidence-task-final-a-pmc7331607-xml","evidence-task-final-a-pmc6731122-xml"],"source_locator":"GenePT: Methods, Cell-level tasks and references 39–43; Alsaigh et al. Abstract; Chou et al. Results; Schirmer et al. Methods/Human tissue samples"},{"label":"Assays","value":"Single-cell expression with cell-type, patient and phenotype annotations.","status":"source_checked","source_ids":["genept-2024"],"source_locator":"Results: cell clustering and batch effects; Discussion; cached text lines 57, 61"},{"label":"Allowed inputs","value":"Gene-expression-derived cell embeddings and GenePT representations.","status":"source_checked","source_ids":["genept-2024"],"source_locator":"Results: cell clustering and batch effects; Discussion; cached text lines 57, 61"},{"label":"Adaptation","value":"Frozen-embedding geometry assessment; no task-classifier result is implied by cell separation alone.","status":"source_checked","source_ids":["genept-2024"],"source_locator":"Results: cell clustering and batch effects; Discussion; cached text lines 57, 61"}],"strengths":[],"limitations":[{"text":"Clustering agreement is not donor-held-out classification. The number of clusters is chosen to match annotated classes, and text-derived gene representations depend on the available descriptions.","source_ids":["genept-2024","evidence-task-final-a-pmc9556750-xml","evidence-task-final-a-pmc7331607-xml","evidence-task-final-a-pmc6731122-xml"],"source_locator":"GenePT: Methods, Cell-level tasks and references 39–43; Alsaigh et al. Abstract; Chou et al. Results; Schirmer et al. Methods/Human tissue samples"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Gene-expression-derived cell embeddings and GenePT representations.","Evaluation: The frozen-embedding clustering analysis does not define a supervised train/test split. Random holdout used for phenotype-classifier experiments belongs to a different evaluation.","Readout: Adjusted Rand index compares embedding-derived clusters with patient and phenotype labels."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["genept-2024"],"source_locator":"Results: cell clustering and batch effects; Discussion; cached text lines 57, 61"},"coverage":"limited","gaps":["Splits: The frozen-embedding clustering analysis does not define a supervised train/test split. Random holdout used for phenotype-classifier experiments belongs to a different evaluation.","Leakage controls: The clustering comparison measures cell and patient associations in embeddings; it does not establish an independent unseen-patient supervised test.","Uncertainty: The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Targeted full-paper and supplement review of the outstanding task fields, with original dataset metadata checked where accessible. Source-scoped omissions are explicit; no independent benchmark reproduction or numerical-result change."}}}}
{"id":"reported-task-53506fe386e4a1","kind":"benchmark","name":"human-versus-viral protein classification","description":"","status":"needs_review","facets":{"areas":["proteins-complexes"],"tasks":["human-versus-viral protein classification"]},"source_ids":["viral-immune-mimicry-2025"],"links":[{"relation":"dataset","target_id":"reported-dataset-43f24c4dfb7351"}],"attributes":{"entity_level":"task","version":null,"task":"human-versus-viral protein classification","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"This evaluation asks whether protein representations distinguish human and viral sequence labels. It is a classification task, not a direct test of immune function.","summary_source_ids":["viral-immune-mimicry-2025"],"summary_source_locator":"Abstract; Methods 2.1 Protein Datasets and 2.6 Model Performance; Table 1","sections":[{"title":"What the evaluation establishes","body":"The source uses reviewed protein records and separates training and test data by sequence clusters. That reduces direct overlap between related examples under the stated clustering rule. It reports classification metrics for different representations; classification errors and biological explanations of those errors are separate claims.","source_ids":["viral-immune-mimicry-2025"],"source_locator":"Abstract; Methods 2.1 Protein Datasets and 2.6 Model Performance; Table 1"}],"facts":[{"label":"Entity type","value":"Paper-specific evaluation task; this profile is a descriptive evidence summary.","status":"source_checked","source_ids":["viral-immune-mimicry-2025"],"source_locator":"Abstract; Methods 2.1 Protein Datasets and 2.6 Model Performance; Table 1"},{"label":"Datasets","value":"Reviewed human and vertebrate-host viral protein records from Swiss-Prot/UniProtKB, with redundancy filtering described in Methods 2.1.","status":"source_checked","source_ids":["viral-immune-mimicry-2025"],"source_locator":"Abstract; Methods 2.1 Protein Datasets and 2.6 Model Performance; Table 1"},{"label":"Organisms","value":"Human proteins and proteins from viruses with a known vertebrate host.","status":"source_checked","source_ids":["viral-immune-mimicry-2025"],"source_locator":"Abstract; Methods 2.1 Protein Datasets and 2.6 Model Performance; Table 1"},{"label":"Assays","value":"Sequence-origin labels from curated database records. This classification endpoint is not an experimental immune-response assay.","status":"source_checked","source_ids":["viral-immune-mimicry-2025"],"source_locator":"Abstract; Methods 2.1 Protein Datasets and 2.6 Model Performance; Table 1"},{"label":"Splits","value":"Methods 2.1 assigns whole UniRef50 clusters to training or test sets. This profile records the split principle, not a verified membership manifest.","status":"source_checked","source_ids":["viral-immune-mimicry-2025"],"source_locator":"Abstract; Methods 2.1 Protein Datasets and 2.6 Model Performance; Table 1"},{"label":"Allowed inputs","value":"Protein sequence representations paired with the study’s human/viral labels.","status":"source_checked","source_ids":["viral-immune-mimicry-2025"],"source_locator":"Abstract; Methods 2.1 Protein Datasets and 2.6 Model Performance; Table 1"},{"label":"Adaptation","value":"Pretrained protein representations are evaluated through a study-specific classifier; a backbone name alone does not identify the full fitted pipeline.","status":"source_checked","source_ids":["viral-immune-mimicry-2025"],"source_locator":"Abstract; Methods 2.1 Protein Datasets and 2.6 Model Performance; Table 1"},{"label":"Metrics","value":"AUROC, log loss, accuracy, precision and recall are described; precision and recall use macro averaging. The linked result retains its original percentage unit.","status":"source_checked","source_ids":["viral-immune-mimicry-2025"],"source_locator":"Abstract; Methods 2.1 Protein Datasets and 2.6 Model Performance; Table 1"},{"label":"Baselines","value":"Table 1 compares the reported protein-representation configurations. They are classification comparators, not experimental immune-function controls.","status":"source_checked","source_ids":["viral-immune-mimicry-2025"],"source_locator":"Abstract; Methods 2.1 Protein Datasets and 2.6 Model Performance; Table 1"}],"strengths":[{"text":"Cluster separation makes the treatment of related sequences visible when interpreting classification accuracy.","source_ids":["viral-immune-mimicry-2025"],"source_locator":"Abstract; Methods 2.1 Protein Datasets and 2.6 Model Performance; Table 1"}],"limitations":[{"text":"Distinguishing sequence origin does not establish an immune mechanism. Database selection, similarity filtering and label composition constrain generalisation.","source_ids":["viral-immune-mimicry-2025"],"source_locator":"Abstract; Methods 2.1 Protein Datasets and 2.6 Model Performance; Table 1"}],"diagram":{"title":"Conceptual assessment outline","steps":["Curated sequence-origin labels","Keep sequence clusters separate","Assess held-out classifications","Report classification metrics"],"caption":"Conceptual overview of the published statistical assessment.","source_ids":["viral-immune-mimicry-2025"],"source_locator":"Abstract; Methods 2.1 Protein Datasets and 2.6 Model Performance; Table 1"},"coverage":"limited","gaps":["This profile reviews the evaluation question, data type and statistical readout. It does not reconstruct an executable protocol, verify exact split membership or independently reproduce the experiment."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected the cited primary abstract and descriptive computational-evaluation sections. Review covers the descriptive claims shown; no executable protocol was reconstructed. Original numerical records retain their prior transcription review."}}}}
{"id":"reported-task-53e3d216eef6db","kind":"benchmark","name":"Simulated prophage-contig detection","description":"","status":"needs_review","facets":{"areas":["microbes-communities"],"tasks":["Simulated prophage-contig detection"]},"source_ids":["viral-contig-simulation-2021"],"links":[{"relation":"dataset","target_id":"reported-dataset-cd51026cdb6a7a"}],"attributes":{"entity_level":"task","version":null,"task":"Simulated prophage-contig detection","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Simulated metagenome assessment compares viral-contig and prophage identification across varying sample conditions.","summary_source_ids":["viral-contig-simulation-2021"],"summary_source_locator":"Discussion and Conclusion; cached text lines 71–74","sections":[{"title":"Evaluation methodology","body":"Thirty simulated metagenome collections. Simulation scenarios vary taxonomic complexity and contig characteristics; exact reference withholding remains unextracted. Precision, F1 and prophage/genus recall are discussed as different outcomes. Nine viral-identification tools, including Vibrant, VirSorter and Kraken2. Low- and medium-complexity simulation collections use nonoverlapping selected reference genomes. That condition does not establish exclusion from every identification tool’s database. The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","source_ids":["viral-contig-simulation-2021"],"source_locator":"Discussion and Conclusion; cached text lines 71–74"}],"facts":[{"label":"Datasets","value":"Thirty simulated metagenome collections.","status":"source_checked","source_ids":["viral-contig-simulation-2021"],"source_locator":"Discussion and Conclusion; cached text lines 71–74"},{"label":"Splits","value":"Simulation scenarios vary taxonomic complexity and contig characteristics; exact reference withholding remains unextracted.","status":"source_checked","source_ids":["viral-contig-simulation-2021"],"source_locator":"Discussion and Conclusion; cached text lines 71–74"},{"label":"Metrics","value":"Precision, F1 and prophage/genus recall are discussed as different outcomes.","status":"source_checked","source_ids":["viral-contig-simulation-2021"],"source_locator":"Discussion and Conclusion; cached text lines 71–74"},{"label":"Baselines","value":"Nine viral-identification tools, including Vibrant, VirSorter and Kraken2.","status":"source_checked","source_ids":["viral-contig-simulation-2021"],"source_locator":"Discussion and Conclusion; cached text lines 71–74"},{"label":"Leakage controls","value":"Low- and medium-complexity simulation collections use nonoverlapping selected reference genomes. That condition does not establish exclusion from every identification tool’s database.","status":"source_checked","source_ids":["viral-contig-simulation-2021"],"source_locator":"Discussion and Conclusion; cached text lines 71–74"},{"label":"Uncertainty","value":"The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","status":"unreported","source_ids":["viral-contig-simulation-2021"],"source_locator":"Discussion and Conclusion; cached text lines 71–74"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["viral-contig-simulation-2021"],"source_locator":"Discussion and Conclusion; cached text lines 71–74"},{"label":"Organisms","value":"Simulated microbial/viral communities.","status":"source_checked","source_ids":["viral-contig-simulation-2021"],"source_locator":"Discussion and Conclusion; cached text lines 71–74"},{"label":"Assays","value":"Known-origin simulated contigs.","status":"source_checked","source_ids":["viral-contig-simulation-2021"],"source_locator":"Discussion and Conclusion; cached text lines 71–74"},{"label":"Allowed inputs","value":"Assembled contig sequence.","status":"source_checked","source_ids":["viral-contig-simulation-2021"],"source_locator":"Discussion and Conclusion; cached text lines 71–74"},{"label":"Adaptation","value":"Existing identification tools are applied across controlled simulation conditions.","status":"source_checked","source_ids":["viral-contig-simulation-2021"],"source_locator":"Discussion and Conclusion; cached text lines 71–74"}],"strengths":[],"limitations":[{"text":"Contig length and taxonomic complexity change performance. The simulation result does not establish generalization to arbitrary real metagenomes or unknown reference families.","source_ids":["viral-contig-simulation-2021"],"source_locator":"Discussion and Conclusion; cached text lines 71–74"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Assembled contig sequence.","Evaluation: Existing identification tools are applied across controlled simulation conditions.","Readout: Precision, F1 and prophage/genus recall are discussed as different outcomes."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["viral-contig-simulation-2021"],"source_locator":"Discussion and Conclusion; cached text lines 71–74"},"coverage":"limited","gaps":["Contig length and taxonomic complexity change performance. The simulation result does not establish generalization to arbitrary real metagenomes or unknown reference families."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Task-specific computational methodology and field context checked in the cited primary-source artifact. Source-backed fields, inapplicable evaluator dimensions and unresolved details are distinguished. Numerical results were not reproduced."}}}}
{"id":"reported-task-5693847493f19f","kind":"benchmark","name":"mRNA half-life prediction","description":"","status":"needs_review","facets":{"areas":["rna-transcriptomes"],"tasks":["mRNA half-life prediction"]},"source_ids":["mrna-lm-2025"],"links":[{"relation":"dataset","target_id":"reported-dataset-52f00ccaabf0d9"}],"attributes":{"entity_level":"task","version":null,"task":"mRNA half-life prediction","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"mRNA half-life prediction is evaluated with cross-validation on a compiled human annotation dataset.","summary_source_ids":["mrna-lm-2025"],"summary_source_locator":"Methods: Datasets; Results: comparison; cached text lines 38–40, 51","sections":[{"title":"Evaluation methodology","body":"Human transcript half-life measurements compiled from multiple publications and samples. Ten-fold cross-validation for half-life, distinct from the five-fold settings used for other paper tasks. Spearman correlation, summarized across folds. RNA-FM and Saluki; the cited Saluki comparison is a literature-reported ablation without a splice-site feature. The half-life experiment inherits the original Saluki dataset’s predefined ten-fold split. The supplement does not specify an additional homology filter or a pretraining-versus-test sequence audit for mRNA-LM. Its random five-fold rule applies to the other datasets, not half-life.","source_ids":["mrna-lm-2025","evidence-task-final-a-mrna-lm-2025-gkaf044-supplemental-file-pdf"],"source_locator":"Methods: Datasets; Results: comparison; cached text lines 38–40, 51; Supplement: Section 1 Sequence extraction and data processing"}],"facts":[{"label":"Datasets","value":"Human transcript half-life measurements compiled from multiple publications and samples.","status":"source_checked","source_ids":["mrna-lm-2025"],"source_locator":"Methods: Datasets; Results: comparison; cached text lines 38–40, 51"},{"label":"Splits","value":"Ten-fold cross-validation for half-life, distinct from the five-fold settings used for other paper tasks.","status":"source_checked","source_ids":["mrna-lm-2025"],"source_locator":"Methods: Datasets; Results: comparison; cached text lines 38–40, 51"},{"label":"Metrics","value":"Spearman correlation, summarized across folds.","status":"source_checked","source_ids":["mrna-lm-2025"],"source_locator":"Methods: Datasets; Results: comparison; cached text lines 38–40, 51"},{"label":"Baselines","value":"RNA-FM and Saluki; the cited Saluki comparison is a literature-reported ablation without a splice-site feature.","status":"source_checked","source_ids":["mrna-lm-2025"],"source_locator":"Methods: Datasets; Results: comparison; cached text lines 38–40, 51"},{"label":"Leakage controls","value":"The half-life experiment inherits the original Saluki dataset’s predefined ten-fold split. The supplement does not specify an additional homology filter or a pretraining-versus-test sequence audit for mRNA-LM. Its random five-fold rule applies to the other datasets, not half-life.","status":"source_checked","source_ids":["mrna-lm-2025","evidence-task-final-a-mrna-lm-2025-gkaf044-supplemental-file-pdf"],"source_locator":"Supplement: Section 1 Sequence extraction and data processing"},{"label":"Uncertainty","value":"Fold-wise distributions are shown in supplementary box plots.","status":"source_checked","source_ids":["mrna-lm-2025"],"source_locator":"Methods: Datasets; Results: comparison; cached text lines 38–40, 51"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["mrna-lm-2025"],"source_locator":"Methods: Datasets; Results: comparison; cached text lines 38–40, 51"},{"label":"Organisms","value":"Human.","status":"source_checked","source_ids":["mrna-lm-2025"],"source_locator":"Methods: Datasets; Results: comparison; cached text lines 38–40, 51"},{"label":"Assays","value":"Published transcript half-life measurements.","status":"source_checked","source_ids":["mrna-lm-2025"],"source_locator":"Methods: Datasets; Results: comparison; cached text lines 38–40, 51"},{"label":"Allowed inputs","value":"mRNA sequence representations.","status":"source_checked","source_ids":["mrna-lm-2025"],"source_locator":"Methods: Datasets; Results: comparison; cached text lines 38–40, 51"},{"label":"Adaptation","value":"Supervised half-life prediction with ten-fold cross-validation.","status":"source_checked","source_ids":["mrna-lm-2025"],"source_locator":"Methods: Datasets; Results: comparison; cached text lines 38–40, 51"}],"strengths":[],"limitations":[{"text":"The inherited ten-fold half-life split and the random folds of other tasks are different protocols. The quoted Saluki ablation omits splice-site features and is not the full Saluki configuration.","source_ids":["mrna-lm-2025","evidence-task-final-a-mrna-lm-2025-gkaf044-supplemental-file-pdf"],"source_locator":"Supplement: Section 1 Sequence extraction and data processing"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: mRNA sequence representations.","Evaluation: Ten-fold cross-validation for half-life, distinct from the five-fold settings used for other paper tasks.","Readout: Spearman correlation, summarized across folds."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["mrna-lm-2025"],"source_locator":"Methods: Datasets; Results: comparison; cached text lines 38–40, 51"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Targeted full-paper and supplement review of the outstanding task fields, with original dataset metadata checked where accessible. Source-scoped omissions are explicit; no independent benchmark reproduction or numerical-result change."}}}}
{"id":"reported-task-571f0a2e7faed3","kind":"benchmark","name":"Strain-level abundance quantification","description":"","status":"needs_review","facets":{"areas":["microbes-communities"],"tasks":["Strain-level abundance quantification"]},"source_ids":["cammiq-2022"],"links":[{"relation":"dataset","target_id":"reported-dataset-72e837e5b97041"}],"attributes":{"entity_level":"task","version":null,"task":"Strain-level abundance quantification","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Strain-level taxonomic abundance estimation is compared against known compositions in simulated human-gut read collections.","summary_source_ids":["cammiq-2022"],"summary_source_locator":"Results: Datasets; Performance of CAMMiQ at the strain level; Table 5; cached text lines 19–22, 55–57","sections":[{"title":"Evaluation methodology","body":"Queries derived from a reference index of human-gut bacterial genomes; Table 5 reports four strain-level query sets. An index-and-query benchmark rather than a supervised train/test split; the relevant evaluation is the strain-level collection. Correctly identified strains and L1/L2 distances between predicted and true relative abundance vectors. Kraken2, KrakenUniq, CLARK, Centrifuge and MetaPhlAn2. The four strain-level queries are simulated from the same 614-genome reference collection used to build the custom indices. A genome without unique or doubly unique 100-mers is excluded from query generation. This is identification and abundance estimation against an indexed collection, not a strain-held-out generalization test.","source_ids":["cammiq-2022","evidence-task-final-a-cammiq-2022-41467-2022-33869-moesm1-esm-pdf"],"source_locator":"Results: Datasets; Performance of CAMMiQ at the strain level; Table 5; cached text lines 19–22, 55–57; Supplementary Notes 5.2 Strain-Level Index Dataset and 5.5 Strain-Level Queries"}],"facts":[{"label":"Datasets","value":"Queries derived from a reference index of human-gut bacterial genomes; Table 5 reports four strain-level query sets.","status":"source_checked","source_ids":["cammiq-2022"],"source_locator":"Results: Datasets; Performance of CAMMiQ at the strain level; Table 5; cached text lines 19–22, 55–57"},{"label":"Splits","value":"An index-and-query benchmark rather than a supervised train/test split; the relevant evaluation is the strain-level collection.","status":"source_checked","source_ids":["cammiq-2022"],"source_locator":"Results: Datasets; Performance of CAMMiQ at the strain level; Table 5; cached text lines 19–22, 55–57"},{"label":"Metrics","value":"Correctly identified strains and L1/L2 distances between predicted and true relative abundance vectors.","status":"source_checked","source_ids":["cammiq-2022"],"source_locator":"Results: Datasets; Performance of CAMMiQ at the strain level; Table 5; cached text lines 19–22, 55–57"},{"label":"Baselines","value":"Kraken2, KrakenUniq, CLARK, Centrifuge and MetaPhlAn2.","status":"source_checked","source_ids":["cammiq-2022"],"source_locator":"Results: Datasets; Performance of CAMMiQ at the strain level; Table 5; cached text lines 19–22, 55–57"},{"label":"Leakage controls","value":"The four strain-level queries are simulated from the same 614-genome reference collection used to build the custom indices. A genome without unique or doubly unique 100-mers is excluded from query generation. This is identification and abundance estimation against an indexed collection, not a strain-held-out generalization test.","status":"source_checked","source_ids":["cammiq-2022","evidence-task-final-a-cammiq-2022-41467-2022-33869-moesm1-esm-pdf"],"source_locator":"Supplementary Notes 5.2 Strain-Level Index Dataset and 5.5 Strain-Level Queries"},{"label":"Uncertainty","value":"Table 5 reports one identification count and L1/L2 abundance error for each of four different query constructions. Supplementary Note 5.5 does not define these four constructions as replicate draws for a confidence interval, and no uncertainty interval accompanies the table values.","status":"unreported","source_ids":["cammiq-2022","evidence-task-final-a-cammiq-2022-41467-2022-33869-moesm1-esm-pdf"],"source_locator":"Table 5; Supplementary Note 5.5 Strain-Level Queries"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["cammiq-2022"],"source_locator":"Results: Datasets; Performance of CAMMiQ at the strain level; Table 5; cached text lines 19–22, 55–57"},{"label":"Organisms","value":"Human-gut bacterial strains.","status":"source_checked","source_ids":["cammiq-2022"],"source_locator":"Results: Datasets; Performance of CAMMiQ at the strain level; Table 5; cached text lines 19–22, 55–57"},{"label":"Assays","value":"Synthetic sequence queries with known strain abundances.","status":"source_checked","source_ids":["cammiq-2022"],"source_locator":"Results: Datasets; Performance of CAMMiQ at the strain level; Table 5; cached text lines 19–22, 55–57"},{"label":"Allowed inputs","value":"Metagenomic reads and a reference genome index.","status":"source_checked","source_ids":["cammiq-2022"],"source_locator":"Results: Datasets; Performance of CAMMiQ at the strain level; Table 5; cached text lines 19–22, 55–57"},{"label":"Adaptation","value":"Reference-index construction and query classification; no assay-label supervised fitting is prescribed.","status":"source_checked","source_ids":["cammiq-2022"],"source_locator":"Results: Datasets; Performance of CAMMiQ at the strain level; Table 5; cached text lines 19–22, 55–57"}],"strengths":[{"text":"Known query composition provides an explicit reference for strain-abundance error.","source_ids":["cammiq-2022"],"source_locator":"Results: Datasets; Performance of CAMMiQ at the strain level; Table 5; cached text lines 19–22, 55–57"}],"limitations":[{"text":"Reference-index inclusion and simulated query composition define the scope; no unseen-strain generalization guarantee is inferred.","source_ids":["cammiq-2022"],"source_locator":"Results: Datasets; Performance of CAMMiQ at the strain level; Table 5; cached text lines 19–22, 55–57"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Metagenomic reads and a reference genome index.","Evaluation: An index-and-query benchmark rather than a supervised train/test split; the relevant evaluation is the strain-level collection.","Readout: Correctly identified strains and L1/L2 distances between predicted and true relative abundance vectors."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["cammiq-2022"],"source_locator":"Results: Datasets; Performance of CAMMiQ at the strain level; Table 5; cached text lines 19–22, 55–57"},"coverage":"limited","gaps":["Uncertainty: Table 5 reports one identification count and L1/L2 abundance error for each of four different query constructions. Supplementary Note 5.5 does not define these four constructions as replicate draws for a confidence interval, and no uncertainty interval accompanies the table values."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Targeted full-paper and supplement review of the outstanding task fields, with original dataset metadata checked where accessible. Source-scoped omissions are explicit; no independent benchmark reproduction or numerical-result change."}}}}
{"id":"reported-task-57dc3dcdb67a81","kind":"benchmark","name":"Mean ribosome load from MPRA","description":"","status":"needs_review","facets":{"areas":["rna-transcriptomes"],"tasks":["Mean ribosome load from MPRA"]},"source_ids":["mrnabench-2025"],"links":[{"relation":"dataset","target_id":"reported-dataset-3a3e3880a3fed0"}],"attributes":{"entity_level":"task","version":null,"task":"Mean ribosome load from MPRA","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Mean ribosome-load prediction includes a dedicated check of compositional generalization beyond the default split.","summary_source_ids":["mrnabench-2025"],"summary_source_locator":"Benchmarking Tasks; compositional generalization analysis; cached text lines 24, 26–27, 190–193; default-split results table caption and compositional-split results; matching task comparison table/ablation captions","sections":[{"title":"Evaluation methodology","body":"MRL-MPRA sequence/measurement dataset within mRNABench. The standard MRL-MPRA evaluation uses a naive random split. A separate compositional-generalization experiment holds out combinations of upstream AUG and Kozak features; neither should be described as the homology split used for other mRNABench tasks. Pearson correlation and the change between split conditions. Naive sequence-feature baseline, randomly initialized Naive Mamba, supervised CNN and multiple frozen foundation-model representations. The standard MRL-MPRA split is random. A separate compositional test holds out selected combinations of upstream AUG and Kozak features. Homology-based splits described for other benchmark tasks are not evidence of a homology-disjoint MRL-MPRA split.","source_ids":["mrnabench-2025"],"source_locator":"Benchmarking Tasks; compositional generalization analysis; cached text lines 24, 26–27, 190–193; default-split results table caption and compositional-split results; matching task comparison table/ablation captions; Methods: Data Splitting Strategies; Appendix: Compositional Generalization"}],"facts":[{"label":"Datasets","value":"MRL-MPRA sequence/measurement dataset within mRNABench.","status":"source_checked","source_ids":["mrnabench-2025"],"source_locator":"Benchmarking Tasks; compositional generalization analysis; cached text lines 24, 26–27, 190–193; default-split results table caption and compositional-split results; matching task comparison table/ablation captions"},{"label":"Splits","value":"The standard MRL-MPRA evaluation uses a naive random split. A separate compositional-generalization experiment holds out combinations of upstream AUG and Kozak features; neither should be described as the homology split used for other mRNABench tasks.","status":"source_checked","source_ids":["mrnabench-2025"],"source_locator":"Methods: Data Splitting Strategies; Appendix: Compositional Generalization"},{"label":"Metrics","value":"Pearson correlation and the change between split conditions.","status":"source_checked","source_ids":["mrnabench-2025"],"source_locator":"Benchmarking Tasks; compositional generalization analysis; cached text lines 24, 26–27, 190–193; default-split results table caption and compositional-split results; matching task comparison table/ablation captions"},{"label":"Baselines","value":"Naive sequence-feature baseline, randomly initialized Naive Mamba, supervised CNN and multiple frozen foundation-model representations.","status":"source_checked","source_ids":["mrnabench-2025"],"source_locator":"Benchmarking Tasks; compositional generalization analysis; cached text lines 24, 26–27, 190–193; default-split results table caption and compositional-split results; matching task comparison table/ablation captions"},{"label":"Leakage controls","value":"The standard MRL-MPRA split is random. A separate compositional test holds out selected combinations of upstream AUG and Kozak features. Homology-based splits described for other benchmark tasks are not evidence of a homology-disjoint MRL-MPRA split.","status":"source_checked","source_ids":["mrnabench-2025"],"source_locator":"Methods: Data Splitting Strategies; Appendix: Compositional Generalization"},{"label":"Uncertainty","value":"The default split results are averaged over ten random splits with 95% confidence intervals; the compositional split is reported separately.","status":"source_checked","source_ids":["mrnabench-2025"],"source_locator":"Benchmarking Tasks; compositional generalization analysis; cached text lines 24, 26–27, 190–193; default-split results table caption and compositional-split results; matching task comparison table/ablation captions"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["mrnabench-2025"],"source_locator":"Benchmarking Tasks; compositional generalization analysis; cached text lines 24, 26–27, 190–193; default-split results table caption and compositional-split results; matching task comparison table/ablation captions"},{"label":"Organisms","value":"The MRL-MPRA task uses synthetic or designed 5′ UTR reporter libraries assayed in human cells. The reporter’s experimental host is distinct from a natural source organism for each synthetic UTR.","status":"source_checked","source_ids":["mrnabench-2025"],"source_locator":"Benchmarking Tasks: Local Tasks, MRL-MPRA"},{"label":"Assays","value":"Massively parallel reporter measurements of mean ribosome load.","status":"source_checked","source_ids":["mrnabench-2025"],"source_locator":"Benchmarking Tasks; compositional generalization analysis; cached text lines 24, 26–27, 190–193; default-split results table caption and compositional-split results; matching task comparison table/ablation captions"},{"label":"Allowed inputs","value":"mRNA/UTR sequence representations.","status":"source_checked","source_ids":["mrnabench-2025"],"source_locator":"Benchmarking Tasks; compositional generalization analysis; cached text lines 24, 26–27, 190–193; default-split results table caption and compositional-split results; matching task comparison table/ablation captions"},{"label":"Adaptation","value":"Embedding-based supervised prediction under default and compositional split conditions.","status":"source_checked","source_ids":["mrnabench-2025"],"source_locator":"Benchmarking Tasks; compositional generalization analysis; cached text lines 24, 26–27, 190–193; default-split results table caption and compositional-split results; matching task comparison table/ablation captions"}],"strengths":[],"limitations":[{"text":"Compositional testing changes the feature combinations represented in training and testing.","source_ids":["mrnabench-2025"],"source_locator":"Benchmarking Tasks; compositional generalization analysis; cached text lines 24, 26–27, 190–193; default-split results table caption and compositional-split results; matching task comparison table/ablation captions"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: mRNA/UTR sequence representations.","Evaluation: The standard MRL-MPRA evaluation uses a naive random split. A separate compositional-generalization experiment holds out combinations of upstream AUG and Kozak features; neither should be described as the homology split used for other mRNABench tasks.","Readout: Pearson correlation and the change between split conditions."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["mrnabench-2025"],"source_locator":"Benchmarking Tasks; compositional generalization analysis; cached text lines 24, 26–27, 190–193; default-split results table caption and compositional-split results; matching task comparison table/ablation captions; Methods: Data Splitting Strategies; Appendix: Compositional Generalization"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Targeted full-paper and supplement review of the outstanding task fields, with original dataset metadata checked where accessible. Source-scoped omissions are explicit; no independent benchmark reproduction or numerical-result change."}}}}
{"id":"reported-task-5b929593eefc76","kind":"benchmark","name":"Cell-type identification","description":"","status":"needs_review","facets":{"areas":["cells-tissues"],"tasks":["Cell-type identification"]},"source_ids":["single-cell-peft-2024"],"links":[{"relation":"dataset","target_id":"reported-dataset-488d5de6bb9c1b"}],"attributes":{"entity_level":"task","version":null,"task":"Cell-type identification","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Cell-type identification compares parameter-efficient adaptation with conventional fine-tuning on external annotation datasets.","summary_source_ids":["single-cell-peft-2024"],"summary_source_locator":"Methods: Finetuning and evaluation settings; Data preparation; cached text lines 36–41","sections":[{"title":"Evaluation methodology","body":"M.S., Zheng68k, NSCLC and COVID-19 single-cell datasets. Original study splits and preprocessing are retained for reused benchmarks; validation loss selects checkpoints. Accuracy, precision, recall and weighted F1; silhouette is an additional embedding diagnostic. Prompt-based adaptation and traditional fine-tuning of single-cell models. The source states evaluated datasets were not used in the assessed models’ pretraining; an independent corpus audit remains outstanding. The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","source_ids":["single-cell-peft-2024"],"source_locator":"Methods: Finetuning and evaluation settings; Data preparation; cached text lines 36–41"}],"facts":[{"label":"Datasets","value":"M.S., Zheng68k, NSCLC and COVID-19 single-cell datasets.","status":"source_checked","source_ids":["single-cell-peft-2024"],"source_locator":"Methods: Finetuning and evaluation settings; Data preparation; cached text lines 36–41"},{"label":"Splits","value":"Original study splits and preprocessing are retained for reused benchmarks; validation loss selects checkpoints.","status":"source_checked","source_ids":["single-cell-peft-2024"],"source_locator":"Methods: Finetuning and evaluation settings; Data preparation; cached text lines 36–41"},{"label":"Metrics","value":"Accuracy, precision, recall and weighted F1; silhouette is an additional embedding diagnostic.","status":"source_checked","source_ids":["single-cell-peft-2024"],"source_locator":"Methods: Finetuning and evaluation settings; Data preparation; cached text lines 36–41"},{"label":"Baselines","value":"Prompt-based adaptation and traditional fine-tuning of single-cell models.","status":"source_checked","source_ids":["single-cell-peft-2024"],"source_locator":"Methods: Finetuning and evaluation settings; Data preparation; cached text lines 36–41"},{"label":"Leakage controls","value":"The source states evaluated datasets were not used in the assessed models’ pretraining; an independent corpus audit remains outstanding.","status":"source_checked","source_ids":["single-cell-peft-2024"],"source_locator":"Methods: Finetuning and evaluation settings; Data preparation; cached text lines 36–41"},{"label":"Uncertainty","value":"The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","status":"unreported","source_ids":["single-cell-peft-2024"],"source_locator":"Methods: Finetuning and evaluation settings; Data preparation; cached text lines 36–41"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["single-cell-peft-2024"],"source_locator":"Methods: Finetuning and evaluation settings; Data preparation; cached text lines 36–41"},{"label":"Organisms","value":"Dataset-specific cell collections including human disease datasets.","status":"source_checked","source_ids":["single-cell-peft-2024"],"source_locator":"Methods: Finetuning and evaluation settings; Data preparation; cached text lines 36–41"},{"label":"Assays","value":"Single-cell expression and cell-type annotations.","status":"source_checked","source_ids":["single-cell-peft-2024"],"source_locator":"Methods: Finetuning and evaluation settings; Data preparation; cached text lines 36–41"},{"label":"Allowed inputs","value":"Single-cell gene-expression representations.","status":"source_checked","source_ids":["single-cell-peft-2024"],"source_locator":"Methods: Finetuning and evaluation settings; Data preparation; cached text lines 36–41"},{"label":"Adaptation","value":"Prompt-based parameter-efficient adaptation versus conventional model fine-tuning.","status":"source_checked","source_ids":["single-cell-peft-2024"],"source_locator":"Methods: Finetuning and evaluation settings; Data preparation; cached text lines 36–41"}],"strengths":[],"limitations":[{"text":"Frozen-backbone prompting and full fine-tuning use different adaptation budgets. Exact donor grouping remain unextracted.","source_ids":["single-cell-peft-2024"],"source_locator":"Methods: Finetuning and evaluation settings; Data preparation; cached text lines 36–41"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Single-cell gene-expression representations.","Evaluation: Prompt-based parameter-efficient adaptation versus conventional model fine-tuning.","Readout: Accuracy, precision, recall and weighted F1; silhouette is an additional embedding diagnostic."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["single-cell-peft-2024"],"source_locator":"Methods: Finetuning and evaluation settings; Data preparation; cached text lines 36–41"},"coverage":"limited","gaps":["Frozen-backbone prompting and full fine-tuning use different adaptation budgets. Exact donor grouping remain unextracted."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Task-specific computational methodology and field context checked in the cited primary-source artifact. Source-backed fields, inapplicable evaluator dimensions and unresolved details are distinguished. Numerical results were not reproduced."}}}}
{"id":"reported-task-5ec7581b246ea6","kind":"benchmark","name":"RNA secondary structure","description":"","status":"needs_review","facets":{"areas":["rna-transcriptomes"],"tasks":["RNA secondary structure"]},"source_ids":["tu-fold-2025"],"links":[{"relation":"dataset","target_id":"reported-dataset-f2e729f333a333"}],"attributes":{"entity_level":"task","version":null,"task":"RNA secondary structure","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"RNA secondary-structure evaluation includes an explicit held-out-family generalization setting.","summary_source_ids":["tu-fold-2025"],"summary_source_locator":"Methods: Evaluation method; Results: cross-RNA-family evaluation; cached text lines 23–25, 75–77","sections":[{"title":"Evaluation methodology","body":"RNA8F combines RNAStrAlign, ArchiveII, bpRNA and CRW2, retaining sequences no longer than 500 nucleotides from eight RNA families. Table 1 gives family-by-length counts for training, validation and test subsets. Cross-family experiments train on the other families and test the held-out family. F1 and interaction network fidelity, with overall averages across family evaluations. CONTRAfold, RNAstructure, LinearFold and MXfold2; TU-Fold is also compared with its knowledge-merge variant. Family holdout is separated from ordinary sequence-level evaluation.","source_ids":["tu-fold-2025"],"source_locator":"Materials and methods: Experimental settings; Table 1; Methods: Evaluation method; Results: cross-RNA-family evaluation; cached text lines 23–25, 75–77"}],"facts":[{"label":"Datasets","value":"RNA8F combines RNAStrAlign, ArchiveII, bpRNA and CRW2, retaining sequences no longer than 500 nucleotides from eight RNA families. Table 1 gives family-by-length counts for training, validation and test subsets.","status":"source_checked","source_ids":["tu-fold-2025"],"source_locator":"Materials and methods: Experimental settings; Table 1"},{"label":"Splits","value":"Cross-family experiments train on the other families and test the held-out family.","status":"source_checked","source_ids":["tu-fold-2025"],"source_locator":"Methods: Evaluation method; Results: cross-RNA-family evaluation; cached text lines 23–25, 75–77"},{"label":"Metrics","value":"F1 and interaction network fidelity, with overall averages across family evaluations.","status":"source_checked","source_ids":["tu-fold-2025"],"source_locator":"Methods: Evaluation method; Results: cross-RNA-family evaluation; cached text lines 23–25, 75–77"},{"label":"Baselines","value":"CONTRAfold, RNAstructure, LinearFold and MXfold2; TU-Fold is also compared with its knowledge-merge variant.","status":"source_checked","source_ids":["tu-fold-2025"],"source_locator":"Methods: Evaluation method; Results: cross-RNA-family evaluation; cached text lines 23–25, 75–77"},{"label":"Leakage controls","value":"Family holdout is separated from ordinary sequence-level evaluation.","status":"source_checked","source_ids":["tu-fold-2025"],"source_locator":"Methods: Evaluation method; Results: cross-RNA-family evaluation; cached text lines 23–25, 75–77"},{"label":"Uncertainty","value":"The sequence-length-stratified tables report means and standard deviations over three folds. This does not establish a separate interval for every cross-family experiment.","status":"source_checked","source_ids":["tu-fold-2025"],"source_locator":"Methods: Evaluation method; Results: cross-RNA-family evaluation; cached text lines 23–25, 75–77"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["tu-fold-2025"],"source_locator":"Methods: Evaluation method; Results: cross-RNA-family evaluation; cached text lines 23–25, 75–77"},{"label":"Organisms","value":"RNA8F combines RNAStrAlign, ArchiveII, bpRNA and CRW2 into eight RNA families. Table 1 stratifies by family and sequence length; the paper does not provide the organism distribution of the processed train/validation/test subsets. Family names are not organism labels.","status":"unreported","source_ids":["tu-fold-2025"],"source_locator":"Materials and methods: Experimental settings; Table 1"},{"label":"Assays","value":"RNA secondary-structure annotations.","status":"source_checked","source_ids":["tu-fold-2025"],"source_locator":"Methods: Evaluation method; Results: cross-RNA-family evaluation; cached text lines 23–25, 75–77"},{"label":"Allowed inputs","value":"RNA sequence.","status":"source_checked","source_ids":["tu-fold-2025"],"source_locator":"Methods: Evaluation method; Results: cross-RNA-family evaluation; cached text lines 23–25, 75–77"},{"label":"Adaptation","value":"Supervised structure learning; cross-family runs train on other families and hold one family out.","status":"source_checked","source_ids":["tu-fold-2025"],"source_locator":"Methods: Evaluation method; Results: cross-RNA-family evaluation; cached text lines 23–25, 75–77"}],"strengths":[],"limitations":[{"text":"RNA family and sequence-length imbalance affect pooled performance. Within-family and held-out-family tests are separate regimes; output constraints are part of the prediction procedure.","source_ids":["tu-fold-2025"],"source_locator":"Materials and methods: Experimental settings; Table 1"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: RNA sequence.","Evaluation: Cross-family experiments train on the other families and test the held-out family.","Readout: F1 and interaction network fidelity, with overall averages across family evaluations."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["tu-fold-2025"],"source_locator":"Methods: Evaluation method; Results: cross-RNA-family evaluation; cached text lines 23–25, 75–77"},"coverage":"limited","gaps":["Organisms: RNA8F combines RNAStrAlign, ArchiveII, bpRNA and CRW2 into eight RNA families. Table 1 stratifies by family and sequence length; the paper does not provide the organism distribution of the processed train/validation/test subsets. Family names are not organism labels."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Targeted full-paper and supplement review of the outstanding task fields, with original dataset metadata checked where accessible. Source-scoped omissions are explicit; no independent benchmark reproduction or numerical-result change."}}}}
{"id":"reported-task-6243658a1bc215","kind":"benchmark","name":"Zero-shot substitution mutation effects: stability","description":"","status":"needs_review","facets":{"areas":["proteins-complexes"],"tasks":["Zero-shot substitution mutation effects: stability"]},"source_ids":["proteingym-2023"],"links":[{"relation":"dataset","target_id":"reported-dataset-7cec655cd742f3"}],"attributes":{"entity_level":"task","version":null,"task":"Zero-shot substitution mutation effects: stability","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"The stability subset evaluates zero-shot prediction of substitution effects using experimentally measured variant phenotypes.","summary_source_ids":["proteingym-2023"],"summary_source_locator":"Methods: zero-shot DMS evaluation and supervised protocols; cached text lines 54, 67, 78","sections":[{"title":"Evaluation methodology","body":"Stability-labelled substitution assays within the ProteinGym DMS collection. Zero-shot evaluation does not train on the evaluated assay labels; supervised mutation splits are a separate regime. Spearman correlation is central; the suite also defines binarized and ranking metrics. Aggregate suite scores first group related functions. For the zero-shot stability category, Table A7 includes site-independent frequencies, WaveNet, EVmutation, DeepSequence, EVE and GEMME; protein language models; hybrids; and inverse-folding models such as ESM-IF1, MIF-ST and ProteinMPNN. The supervised category has separate embedding and one-hot baselines, including one-hot without augmentation, in Table A12. Their label access and structural inputs differ. Absence of assay labels during adaptation does not establish absence of sequence families from pretraining.","source_ids":["proteingym-2023"],"source_locator":"Methods: zero-shot DMS evaluation and supervised protocols; cached text lines 54, 67, 78; Appendix Tables A7 and A12; Evaluation framework"}],"facts":[{"label":"Datasets","value":"Stability-labelled substitution assays within the ProteinGym DMS collection.","status":"source_checked","source_ids":["proteingym-2023"],"source_locator":"Methods: zero-shot DMS evaluation and supervised protocols; cached text lines 54, 67, 78"},{"label":"Splits","value":"Zero-shot evaluation does not train on the evaluated assay labels; supervised mutation splits are a separate regime.","status":"source_checked","source_ids":["proteingym-2023"],"source_locator":"Methods: zero-shot DMS evaluation and supervised protocols; cached text lines 54, 67, 78"},{"label":"Metrics","value":"Spearman correlation is central; the suite also defines binarized and ranking metrics. Aggregate suite scores first group related functions.","status":"source_checked","source_ids":["proteingym-2023"],"source_locator":"Methods: zero-shot DMS evaluation and supervised protocols; cached text lines 54, 67, 78"},{"label":"Baselines","value":"For the zero-shot stability category, Table A7 includes site-independent frequencies, WaveNet, EVmutation, DeepSequence, EVE and GEMME; protein language models; hybrids; and inverse-folding models such as ESM-IF1, MIF-ST and ProteinMPNN. The supervised category has separate embedding and one-hot baselines, including one-hot without augmentation, in Table A12. Their label access and structural inputs differ.","status":"source_checked","source_ids":["proteingym-2023"],"source_locator":"Appendix Tables A7 and A12; Evaluation framework"},{"label":"Leakage controls","value":"Absence of assay labels during adaptation does not establish absence of sequence families from pretraining.","status":"source_checked","source_ids":["proteingym-2023"],"source_locator":"Methods: zero-shot DMS evaluation and supervised protocols; cached text lines 54, 67, 78"},{"label":"Uncertainty","value":"The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","status":"unreported","source_ids":["proteingym-2023"],"source_locator":"Methods: zero-shot DMS evaluation and supervised protocols; cached text lines 54, 67, 78"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["proteingym-2023"],"source_locator":"Methods: zero-shot DMS evaluation and supervised protocols; cached text lines 54, 67, 78"},{"label":"Organisms","value":"Proteins represented in the stability subset; exact taxon membership is assay-specific.","status":"source_checked","source_ids":["proteingym-2023"],"source_locator":"Methods: zero-shot DMS evaluation and supervised protocols; cached text lines 54, 67, 78"},{"label":"Assays","value":"DMS substitution stability measurements.","status":"source_checked","source_ids":["proteingym-2023"],"source_locator":"Methods: zero-shot DMS evaluation and supervised protocols; cached text lines 54, 67, 78"},{"label":"Allowed inputs","value":"Variant/target protein sequences, with method-specific evolutionary or structural resources.","status":"source_checked","source_ids":["proteingym-2023"],"source_locator":"Methods: zero-shot DMS evaluation and supervised protocols; cached text lines 54, 67, 78"},{"label":"Adaptation","value":"Zero-shot evaluation without fitting on the evaluated assay labels.","status":"source_checked","source_ids":["proteingym-2023"],"source_locator":"Methods: zero-shot DMS evaluation and supervised protocols; cached text lines 54, 67, 78"}],"strengths":[],"limitations":[{"text":"The stability subgroup and the equal-functional-group suite average are different aggregates. Sequence-only, alignment-based and structure-dependent comparators do not receive identical information.","source_ids":["proteingym-2023"],"source_locator":"Appendix Tables A7 and A12; Evaluation framework"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Variant/target protein sequences, with method-specific evolutionary or structural resources.","Evaluation: Zero-shot evaluation does not train on the evaluated assay labels; supervised mutation splits are a separate regime.","Readout: Spearman correlation is central; the suite also defines binarized and ranking metrics. Aggregate suite scores first group related functions."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["proteingym-2023"],"source_locator":"Methods: zero-shot DMS evaluation and supervised protocols; cached text lines 54, 67, 78"},"coverage":"limited","gaps":["Uncertainty: The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Targeted full-paper and supplement review of the outstanding task fields, with original dataset metadata checked where accessible. Source-scoped omissions are explicit; no independent benchmark reproduction or numerical-result change."}}}}
{"id":"reported-task-6312c8a7ac045e","kind":"benchmark","name":"cell-type annotation","description":"","status":"needs_review","facets":{"areas":["cells-tissues"],"tasks":["cell-type annotation"]},"source_ids":["gremln-2026"],"links":[{"relation":"dataset","target_id":"reported-dataset-59def895fbdbb4"}],"attributes":{"entity_level":"task","version":null,"task":"cell-type annotation","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Cell-type annotation evaluates representations on datasets selected to be separate from pretraining data.","summary_source_ids":["gremln-2026"],"summary_source_locator":"Evaluation Datasets; Transcriptomic Landscape Learning & Cell Type Annotation; Bayesian Graph Integration; cached text lines 39–50","sections":[{"title":"Evaluation methodology","body":"Healthy immune-cell data, cancer-infiltrating immune-cell data and non-immune-cell datasets. The source states that evaluation datasets do not overlap with pretraining datasets. Table 1 reports precision, recall and macro F1 separately for human immune cells and held-out non-immune cells. The table describes its ± terms as confidence intervals across three random initializations of test splits; the caption does not name a confidence level. The human immune-cell experiment compares GREmLN with scGPT, scFoundation and Geneformer. scGPT is excluded from the held-out non-immune-cell comparison because its CELLxGENE pretraining prevents that evaluation from being zero-shot. Test-cell regulatory graphs combine training-derived graphs using a training-fitted classifier rather than true test-cell labels.","source_ids":["gremln-2026"],"source_locator":"Evaluation Datasets; Transcriptomic Landscape Learning & Cell Type Annotation; Bayesian Graph Integration; cached text lines 39–50; Table 1 caption and column headings; Validation Experiments: Training cell type classifier; Table 1"}],"facts":[{"label":"Datasets","value":"Healthy immune-cell data, cancer-infiltrating immune-cell data and non-immune-cell datasets.","status":"source_checked","source_ids":["gremln-2026"],"source_locator":"Evaluation Datasets; Transcriptomic Landscape Learning & Cell Type Annotation; Bayesian Graph Integration; cached text lines 39–50"},{"label":"Splits","value":"The source states that evaluation datasets do not overlap with pretraining datasets.","status":"source_checked","source_ids":["gremln-2026"],"source_locator":"Evaluation Datasets; Transcriptomic Landscape Learning & Cell Type Annotation; Bayesian Graph Integration; cached text lines 39–50"},{"label":"Metrics","value":"Table 1 reports precision, recall and macro F1 separately for human immune cells and held-out non-immune cells. The table describes its ± terms as confidence intervals across three random initializations of test splits; the caption does not name a confidence level.","status":"source_checked","source_ids":["gremln-2026"],"source_locator":"Table 1 caption and column headings"},{"label":"Baselines","value":"The human immune-cell experiment compares GREmLN with scGPT, scFoundation and Geneformer. scGPT is excluded from the held-out non-immune-cell comparison because its CELLxGENE pretraining prevents that evaluation from being zero-shot.","status":"source_checked","source_ids":["gremln-2026"],"source_locator":"Validation Experiments: Training cell type classifier; Table 1"},{"label":"Leakage controls","value":"Test-cell regulatory graphs combine training-derived graphs using a training-fitted classifier rather than true test-cell labels.","status":"source_checked","source_ids":["gremln-2026"],"source_locator":"Evaluation Datasets; Transcriptomic Landscape Learning & Cell Type Annotation; Bayesian Graph Integration; cached text lines 39–50"},{"label":"Uncertainty","value":"The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","status":"unreported","source_ids":["gremln-2026"],"source_locator":"Evaluation Datasets; Transcriptomic Landscape Learning & Cell Type Annotation; Bayesian Graph Integration; cached text lines 39–50"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["gremln-2026"],"source_locator":"Evaluation Datasets; Transcriptomic Landscape Learning & Cell Type Annotation; Bayesian Graph Integration; cached text lines 39–50"},{"label":"Organisms","value":"Human immune cells and separately held-out non-immune cell collections.","status":"source_checked","source_ids":["gremln-2026"],"source_locator":"Evaluation Datasets; Transcriptomic Landscape Learning & Cell Type Annotation; Bayesian Graph Integration; cached text lines 39–50"},{"label":"Assays","value":"Single-cell expression with cell-type annotations.","status":"source_checked","source_ids":["gremln-2026"],"source_locator":"Evaluation Datasets; Transcriptomic Landscape Learning & Cell Type Annotation; Bayesian Graph Integration; cached text lines 39–50"},{"label":"Allowed inputs","value":"Single-cell expression representations.","status":"source_checked","source_ids":["gremln-2026"],"source_locator":"Evaluation Datasets; Transcriptomic Landscape Learning & Cell Type Annotation; Bayesian Graph Integration; cached text lines 39–50"},{"label":"Adaptation","value":"Zero-shot cell-type annotation is evaluated on held-out non-immune cells; fine-tuned perturbation-label prediction is a separate task.","status":"source_checked","source_ids":["gremln-2026"],"source_locator":"Evaluation Datasets; Transcriptomic Landscape Learning & Cell Type Annotation; Bayesian Graph Integration; cached text lines 39–50"}],"strengths":[],"limitations":[{"text":"The immune-cell and held-out non-immune-cell settings have different comparator eligibility. scGPT’s absence from the latter is a pretraining-overlap control, not a failed prediction.","source_ids":["gremln-2026"],"source_locator":"Table 1 caption and column headings; Validation Experiments: Training cell type classifier; Table 1"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Single-cell expression representations.","Evaluation: The source states that evaluation datasets do not overlap with pretraining datasets.","Readout: Table 1 reports precision, recall and macro F1 separately for human immune cells and held-out non-immune cells. The table describes its ± terms as confidence intervals across three random initializations of test splits; the caption does not name a confidence level."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["gremln-2026"],"source_locator":"Evaluation Datasets; Transcriptomic Landscape Learning & Cell Type Annotation; Bayesian Graph Integration; cached text lines 39–50; Table 1 caption and column headings"},"coverage":"limited","gaps":["Uncertainty: The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Targeted full-paper and supplement review of the outstanding task fields, with original dataset metadata checked where accessible. Source-scoped omissions are explicit; no independent benchmark reproduction or numerical-result change."}}}}
{"id":"reported-task-6330d593980b5b","kind":"benchmark","name":"Long-read taxonomic profiling","description":"","status":"needs_review","facets":{"areas":["microbes-communities"],"tasks":["Long-read taxonomic profiling"]},"source_ids":["lemur-magnet-2024"],"links":[{"relation":"dataset","target_id":"reported-dataset-768a7ff5bac414"}],"attributes":{"entity_level":"task","version":null,"task":"Long-read taxonomic profiling","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Long-read taxonomic profiling measures both taxon detection and relative-abundance agreement.","summary_source_ids":["lemur-magnet-2024"],"summary_source_locator":"Methods: Synthetic and simulated datasets; Method evaluation; cached text lines 75–77, 97–99; uncertainty/repeat-run/statistical-comparison passages","sections":[{"title":"Evaluation methodology","body":"Synthetic and simulated communities with known reference composition. Known-composition simulated or mock-community samples are profiled against reference resources. The Dilthey simulation has RefSeq species representatives for 94 of 96 strains; the additional simulation selects RefSeq-represented species. Genus/species precision, recall and F1 after abundance thresholding; normalized L1 loss or Spearman correlation for abundance. Lemur v1.0.1 is evaluated alone and with Magnet against Centrifuger v1.0.0, Kraken 2 v2.1.3, Melon v0.1.0, MetaMaps commit 633d2e0 and Sourmash v4.8.2. Melon lacks fungal references; the paper also reports bacterial-only comparisons for fungal-containing datasets. Reference availability is part of this identification task: 94 of 96 strains in the first simulation have a corresponding RefSeq species representative. The additional metagenome simulation deliberately selects species with RefSeq representative genomes and available MAGs. These settings do not establish novel-species generalization.","source_ids":["lemur-magnet-2024"],"source_locator":"Methods: Synthetic and simulated datasets; Method evaluation; cached text lines 75–77, 97–99; uncertainty/repeat-run/statistical-comparison passages; Methods: Synthetic and simulated datasets; Methods: Method Comparison; Methods: Simulated data from Dilthey et al. 2019; Simulated metagenome"}],"facts":[{"label":"Datasets","value":"Synthetic and simulated communities with known reference composition.","status":"source_checked","source_ids":["lemur-magnet-2024"],"source_locator":"Methods: Synthetic and simulated datasets; Method evaluation; cached text lines 75–77, 97–99; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Splits","value":"Known-composition simulated or mock-community samples are profiled against reference resources. The Dilthey simulation has RefSeq species representatives for 94 of 96 strains; the additional simulation selects RefSeq-represented species.","status":"source_checked","source_ids":["lemur-magnet-2024"],"source_locator":"Methods: Synthetic and simulated datasets"},{"label":"Metrics","value":"Genus/species precision, recall and F1 after abundance thresholding; normalized L1 loss or Spearman correlation for abundance.","status":"source_checked","source_ids":["lemur-magnet-2024"],"source_locator":"Methods: Synthetic and simulated datasets; Method evaluation; cached text lines 75–77, 97–99; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Baselines","value":"Lemur v1.0.1 is evaluated alone and with Magnet against Centrifuger v1.0.0, Kraken 2 v2.1.3, Melon v0.1.0, MetaMaps commit 633d2e0 and Sourmash v4.8.2. Melon lacks fungal references; the paper also reports bacterial-only comparisons for fungal-containing datasets.","status":"source_checked","source_ids":["lemur-magnet-2024"],"source_locator":"Methods: Method Comparison"},{"label":"Leakage controls","value":"Reference availability is part of this identification task: 94 of 96 strains in the first simulation have a corresponding RefSeq species representative. The additional metagenome simulation deliberately selects species with RefSeq representative genomes and available MAGs. These settings do not establish novel-species generalization.","status":"source_checked","source_ids":["lemur-magnet-2024"],"source_locator":"Methods: Simulated data from Dilthey et al. 2019; Simulated metagenome"},{"label":"Uncertainty","value":"Mean and standard deviation across five replicate runs are reported for the Zymo EVEN and LOG comparisons; read subsampling also uses repeated seeds.","status":"source_checked","source_ids":["lemur-magnet-2024"],"source_locator":"Methods: Synthetic and simulated datasets; Method evaluation; cached text lines 75–77, 97–99; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["lemur-magnet-2024"],"source_locator":"Methods: Synthetic and simulated datasets; Method evaluation; cached text lines 75–77, 97–99; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Organisms","value":"Synthetic and simulated microbial communities.","status":"source_checked","source_ids":["lemur-magnet-2024"],"source_locator":"Methods: Synthetic and simulated datasets; Method evaluation; cached text lines 75–77, 97–99; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Assays","value":"Known taxonomic composition of sequence mixtures.","status":"source_checked","source_ids":["lemur-magnet-2024"],"source_locator":"Methods: Synthetic and simulated datasets; Method evaluation; cached text lines 75–77, 97–99; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Allowed inputs","value":"Long sequencing reads.","status":"source_checked","source_ids":["lemur-magnet-2024"],"source_locator":"Methods: Synthetic and simulated datasets; Method evaluation; cached text lines 75–77, 97–99; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Adaptation","value":"Lemur and the comparator profilers use their reference resources; Magnet additionally aligns reads to cluster-representative genomes with minimap2. This is reference-based taxonomic profiling, rather than fitting a classifier on labeled train/test folds.","status":"source_checked","source_ids":["lemur-magnet-2024"],"source_locator":"Methods: Competitive read alignment with Magnet; Method Comparison"}],"strengths":[],"limitations":[{"text":"Spearman is computed only over taxa present in both truth and prediction, so it omits false-positive and false-negative taxa. Presence/absence thresholds vary for the Zymo EVEN setting.","source_ids":["lemur-magnet-2024"],"source_locator":"Methods: Synthetic and simulated datasets; Method evaluation; cached text lines 75–77, 97–99; uncertainty/repeat-run/statistical-comparison passages"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Long sequencing reads.","Evaluation: Known-composition simulated or mock-community samples are profiled against reference resources. The Dilthey simulation has RefSeq species representatives for 94 of 96 strains; the additional simulation selects RefSeq-represented species.","Readout: Genus/species precision, recall and F1 after abundance thresholding; normalized L1 loss or Spearman correlation for abundance."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["lemur-magnet-2024"],"source_locator":"Methods: Synthetic and simulated datasets; Method evaluation; cached text lines 75–77, 97–99; uncertainty/repeat-run/statistical-comparison passages; Methods: Synthetic and simulated datasets"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Targeted full-paper and supplement review of the outstanding task fields, with original dataset metadata checked where accessible. Source-scoped omissions are explicit; no independent benchmark reproduction or numerical-result change."}}}}
{"id":"reported-task-64607443a9ba15","kind":"benchmark","name":"enhancer prediction","description":"","status":"needs_review","facets":{"areas":["dna-genomes"],"tasks":["enhancer prediction"]},"source_ids":["enhancer-position-encoding-2024"],"links":[{"relation":"dataset","target_id":"reported-dataset-a03b8e9efde37b"}],"attributes":{"entity_level":"task","version":null,"task":"enhancer prediction","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Enhancer classification uses species-specific VISTA sequence collections with separate validation and test partitions.","summary_source_ids":["enhancer-position-encoding-2024"],"summary_source_locator":"Methods: Datasets; Model performance evaluation metrics; cached text lines 60–62, 87–89; matching task comparison table/ablation captions","sections":[{"title":"Evaluation methodology","body":"Human hg19 and mouse mm9 positive/negative VISTA enhancer sequences. Each species dataset is divided into training, validation and testing in an 80:10:10 ratio. Sensitivity, specificity, accuracy, MCC and ROC-AUC. CNN, random forest, logistic regression, KNN, SVM and XGBoost under the compared feature encodings. The authors report CD-HIT-based removal of highly similar sequences before splitting. The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","source_ids":["enhancer-position-encoding-2024"],"source_locator":"Methods: Datasets; Model performance evaluation metrics; cached text lines 60–62, 87–89; matching task comparison table/ablation captions"}],"facts":[{"label":"Datasets","value":"Human hg19 and mouse mm9 positive/negative VISTA enhancer sequences.","status":"source_checked","source_ids":["enhancer-position-encoding-2024"],"source_locator":"Methods: Datasets; Model performance evaluation metrics; cached text lines 60–62, 87–89; matching task comparison table/ablation captions"},{"label":"Splits","value":"Each species dataset is divided into training, validation and testing in an 80:10:10 ratio.","status":"source_checked","source_ids":["enhancer-position-encoding-2024"],"source_locator":"Methods: Datasets; Model performance evaluation metrics; cached text lines 60–62, 87–89; matching task comparison table/ablation captions"},{"label":"Metrics","value":"Sensitivity, specificity, accuracy, MCC and ROC-AUC.","status":"source_checked","source_ids":["enhancer-position-encoding-2024"],"source_locator":"Methods: Datasets; Model performance evaluation metrics; cached text lines 60–62, 87–89; matching task comparison table/ablation captions"},{"label":"Baselines","value":"CNN, random forest, logistic regression, KNN, SVM and XGBoost under the compared feature encodings.","status":"source_checked","source_ids":["enhancer-position-encoding-2024"],"source_locator":"Methods: Datasets; Model performance evaluation metrics; cached text lines 60–62, 87–89; matching task comparison table/ablation captions"},{"label":"Leakage controls","value":"The authors report CD-HIT-based removal of highly similar sequences before splitting.","status":"source_checked","source_ids":["enhancer-position-encoding-2024"],"source_locator":"Methods: Datasets; Model performance evaluation metrics; cached text lines 60–62, 87–89; matching task comparison table/ablation captions"},{"label":"Uncertainty","value":"The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","status":"unreported","source_ids":["enhancer-position-encoding-2024"],"source_locator":"Methods: Datasets; Model performance evaluation metrics; cached text lines 60–62, 87–89; matching task comparison table/ablation captions"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["enhancer-position-encoding-2024"],"source_locator":"Methods: Datasets; Model performance evaluation metrics; cached text lines 60–62, 87–89; matching task comparison table/ablation captions"},{"label":"Organisms","value":"Human hg19 and mouse mm9.","status":"source_checked","source_ids":["enhancer-position-encoding-2024"],"source_locator":"Methods: Datasets; Model performance evaluation metrics; cached text lines 60–62, 87–89; matching task comparison table/ablation captions"},{"label":"Assays","value":"VISTA enhancer annotations.","status":"source_checked","source_ids":["enhancer-position-encoding-2024"],"source_locator":"Methods: Datasets; Model performance evaluation metrics; cached text lines 60–62, 87–89; matching task comparison table/ablation captions"},{"label":"Allowed inputs","value":"DNA enhancer and negative sequences.","status":"source_checked","source_ids":["enhancer-position-encoding-2024"],"source_locator":"Methods: Datasets; Model performance evaluation metrics; cached text lines 60–62, 87–89; matching task comparison table/ablation captions"},{"label":"Adaptation","value":"Supervised classification with separate train/validation/test partitions for each species.","status":"source_checked","source_ids":["enhancer-position-encoding-2024"],"source_locator":"Methods: Datasets; Model performance evaluation metrics; cached text lines 60–62, 87–89; matching task comparison table/ablation captions"}],"strengths":[],"limitations":[{"text":"The exact identity-filter implementation, comparator settings remain unextracted; sequence filtering alone does not establish chromosome-held-out evaluation.","source_ids":["enhancer-position-encoding-2024"],"source_locator":"Methods: Datasets; Model performance evaluation metrics; cached text lines 60–62, 87–89; matching task comparison table/ablation captions"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: DNA enhancer and negative sequences.","Evaluation: Supervised classification with separate train/validation/test partitions for each species.","Readout: Sensitivity, specificity, accuracy, MCC and ROC-AUC."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["enhancer-position-encoding-2024"],"source_locator":"Methods: Datasets; Model performance evaluation metrics; cached text lines 60–62, 87–89; matching task comparison table/ablation captions"},"coverage":"limited","gaps":["The exact identity-filter implementation, comparator settings remain unextracted; sequence filtering alone does not establish chromosome-held-out evaluation."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Task-specific computational methodology and field context checked in the cited primary-source artifact. Source-backed fields, inapplicable evaluator dimensions and unresolved details are distinguished. Numerical results were not reproduced."}}}}
{"id":"reported-task-660753ec94e631","kind":"benchmark","name":"Cell-type annotation","description":"","status":"needs_review","facets":{"areas":["cells-tissues"],"tasks":["Cell-type annotation"]},"source_ids":["scelmo-2025"],"links":[{"relation":"dataset","target_id":"reported-dataset-8f123f006964ad"}],"attributes":{"entity_level":"task","version":null,"task":"Cell-type annotation","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Cell-type annotation separates cross-batch prediction from a within-study random split.","summary_source_ids":["scelmo-2025"],"summary_source_locator":"Results: Cell-type annotation; Methods: evaluation and baselines; cached text lines 23, 81, 96","sections":[{"title":"Evaluation methodology","body":"hPancreas, PBMC and Aorta annotated single-cell datasets. hPancreas/PBMC withhold one batch; Aorta uses an 80:20 within-study split. Accuracy, precision, recall and F1. GPT-based classifiers, scGPT, Geneformer, GPTCelltype, MLP and PCA-derived representations. Batch-held-out evaluation is explicit for two datasets; it is not the split used for Aorta.","source_ids":["scelmo-2025"],"source_locator":"Results: Cell-type annotation; Methods: evaluation and baselines; cached text lines 23, 81, 96"}],"facts":[{"label":"Datasets","value":"hPancreas, PBMC and Aorta annotated single-cell datasets.","status":"source_checked","source_ids":["scelmo-2025"],"source_locator":"Results: Cell-type annotation; Methods: evaluation and baselines; cached text lines 23, 81, 96"},{"label":"Splits","value":"hPancreas/PBMC withhold one batch; Aorta uses an 80:20 within-study split.","status":"source_checked","source_ids":["scelmo-2025"],"source_locator":"Results: Cell-type annotation; Methods: evaluation and baselines; cached text lines 23, 81, 96"},{"label":"Metrics","value":"Accuracy, precision, recall and F1.","status":"source_checked","source_ids":["scelmo-2025"],"source_locator":"Results: Cell-type annotation; Methods: evaluation and baselines; cached text lines 23, 81, 96"},{"label":"Baselines","value":"GPT-based classifiers, scGPT, Geneformer, GPTCelltype, MLP and PCA-derived representations.","status":"source_checked","source_ids":["scelmo-2025"],"source_locator":"Results: Cell-type annotation; Methods: evaluation and baselines; cached text lines 23, 81, 96"},{"label":"Leakage controls","value":"Batch-held-out evaluation is explicit for two datasets; it is not the split used for Aorta.","status":"source_checked","source_ids":["scelmo-2025"],"source_locator":"Results: Cell-type annotation; Methods: evaluation and baselines; cached text lines 23, 81, 96"},{"label":"Uncertainty","value":"Table 1 reports point metrics for annotation and identifies some rows as copied from GenePT. The cell-annotation section and table do not give repeated-run uncertainty or confidence intervals for the scELMo rows; copied comparator results must not be counted as independent replications.","status":"unreported","source_ids":["scelmo-2025"],"source_locator":"Results: scELMo for cell-type annotation; Table 1 caption"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["scelmo-2025"],"source_locator":"Results: Cell-type annotation; Methods: evaluation and baselines; cached text lines 23, 81, 96"},{"label":"Organisms","value":"Human pancreas, PBMC and Aorta datasets.","status":"source_checked","source_ids":["scelmo-2025"],"source_locator":"Results: Cell-type annotation; Methods: evaluation and baselines; cached text lines 23, 81, 96"},{"label":"Assays","value":"Single-cell expression and cell-type labels.","status":"source_checked","source_ids":["scelmo-2025"],"source_locator":"Results: Cell-type annotation; Methods: evaluation and baselines; cached text lines 23, 81, 96"},{"label":"Allowed inputs","value":"Expression-derived representations and GPT-based gene information.","status":"source_checked","source_ids":["scelmo-2025"],"source_locator":"Results: Cell-type annotation; Methods: evaluation and baselines; cached text lines 23, 81, 96"},{"label":"Adaptation","value":"Supervised annotation and classifier comparisons, with batch holdout where available.","status":"source_checked","source_ids":["scelmo-2025"],"source_locator":"Results: Cell-type annotation; Methods: evaluation and baselines; cached text lines 23, 81, 96"}],"strengths":[{"text":"Batch holdouts test cross-batch annotation where the dataset supports them.","source_ids":["scelmo-2025"],"source_locator":"Results: Cell-type annotation; Methods: evaluation and baselines; cached text lines 23, 81, 96"}],"limitations":[{"text":"Frozen embeddings with a kNN classifier and embeddings with a trained neural adapter are different regimes. Some comparator scores are copied from GenePT and retain that evidence origin.","source_ids":["scelmo-2025"],"source_locator":"Results: scELMo for cell-type annotation; Table 1 caption"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Expression-derived representations and GPT-based gene information.","Evaluation: hPancreas/PBMC withhold one batch; Aorta uses an 80:20 within-study split.","Readout: Accuracy, precision, recall and F1."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["scelmo-2025"],"source_locator":"Results: Cell-type annotation; Methods: evaluation and baselines; cached text lines 23, 81, 96"},"coverage":"limited","gaps":["Uncertainty: Table 1 reports point metrics for annotation and identifies some rows as copied from GenePT. The cell-annotation section and table do not give repeated-run uncertainty or confidence intervals for the scELMo rows; copied comparator results must not be counted as independent replications."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Targeted full-paper and supplement review of the outstanding task fields, with original dataset metadata checked where accessible. Source-scoped omissions are explicit; no independent benchmark reproduction or numerical-result change."}}}}
{"id":"reported-task-6e54c7452b2b81","kind":"benchmark","name":"human protein-protein interaction prediction","description":"","status":"needs_review","facets":{"areas":["proteins-complexes"],"tasks":["human protein-protein interaction prediction"]},"source_ids":["mulan-2025"],"links":[{"relation":"dataset","target_id":"reported-dataset-38151fa548e291"}],"attributes":{"entity_level":"task","version":null,"task":"human protein-protein interaction prediction","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Human protein-interaction prediction evaluates structural augmentation of pretrained protein representations.","summary_source_ids":["mulan-2025"],"summary_source_locator":"Methods §2.5; Results: HumanPPI; cached text lines 31–33, 43; task metric definitions and corresponding results table; uncertainty/repeat-run/statistical-comparison passages","sections":[{"title":"Evaluation methodology","body":"HumanPPI downstream task; exact dataset release remains unextracted. Downstream hyperparameters and checkpoint selection use validation metrics. Area under the ROC curve for binary protein-interaction classification. MULAN-augmented models are compared with their underlying ESM2 representations. Original downstream dataset splits are retained; this alone does not certify protein-pair identity independence or absence from backbone pretraining. Downstream results average five initialization seeds; corresponding standard deviations are provided in the supplement.","source_ids":["mulan-2025"],"source_locator":"Methods §2.5; Results: HumanPPI; cached text lines 31–33, 43; task metric definitions and corresponding results table; uncertainty/repeat-run/statistical-comparison passages"}],"facts":[{"label":"Datasets","value":"HumanPPI downstream task; exact dataset release remains unextracted.","status":"source_checked","source_ids":["mulan-2025"],"source_locator":"Methods §2.5; Results: HumanPPI; cached text lines 31–33, 43; task metric definitions and corresponding results table; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Splits","value":"Downstream hyperparameters and checkpoint selection use validation metrics.","status":"source_checked","source_ids":["mulan-2025"],"source_locator":"Methods §2.5; Results: HumanPPI; cached text lines 31–33, 43; task metric definitions and corresponding results table; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Metrics","value":"Area under the ROC curve for binary protein-interaction classification.","status":"source_checked","source_ids":["mulan-2025"],"source_locator":"Methods §2.5; Results: HumanPPI; cached text lines 31–33, 43; task metric definitions and corresponding results table; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Baselines","value":"MULAN-augmented models are compared with their underlying ESM2 representations.","status":"source_checked","source_ids":["mulan-2025"],"source_locator":"Methods §2.5; Results: HumanPPI; cached text lines 31–33, 43; task metric definitions and corresponding results table; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Leakage controls","value":"Original downstream dataset splits are retained; this alone does not certify protein-pair identity independence or absence from backbone pretraining.","status":"source_checked","source_ids":["mulan-2025"],"source_locator":"Methods §2.5; Results: HumanPPI; cached text lines 31–33, 43; task metric definitions and corresponding results table; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Uncertainty","value":"Downstream results average five initialization seeds; corresponding standard deviations are provided in the supplement.","status":"source_checked","source_ids":["mulan-2025"],"source_locator":"Methods §2.5; Results: HumanPPI; cached text lines 31–33, 43; task metric definitions and corresponding results table; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["mulan-2025"],"source_locator":"Methods §2.5; Results: HumanPPI; cached text lines 31–33, 43; task metric definitions and corresponding results table; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Organisms","value":"Human.","status":"source_checked","source_ids":["mulan-2025"],"source_locator":"Methods §2.5; Results: HumanPPI; cached text lines 31–33, 43; task metric definitions and corresponding results table; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Assays","value":"HumanPPI protein-interaction annotations.","status":"source_checked","source_ids":["mulan-2025"],"source_locator":"Methods §2.5; Results: HumanPPI; cached text lines 31–33, 43; task metric definitions and corresponding results table; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Allowed inputs","value":"Protein pairs represented using sequence and structure-informed embeddings.","status":"source_checked","source_ids":["mulan-2025"],"source_locator":"Methods §2.5; Results: HumanPPI; cached text lines 31–33, 43; task metric definitions and corresponding results table; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Adaptation","value":"Supervised downstream prediction with validation-based tuning/checkpoint selection.","status":"source_checked","source_ids":["mulan-2025"],"source_locator":"Methods §2.5; Results: HumanPPI; cached text lines 31–33, 43; task metric definitions and corresponding results table; uncertainty/repeat-run/statistical-comparison passages"}],"strengths":[],"limitations":[{"text":"Protein-pair grouping, negative sampling, metric implementation remain unextracted. Gains in other downstream tasks do not specify this task’s protocol.","source_ids":["mulan-2025"],"source_locator":"Methods §2.5; Results: HumanPPI; cached text lines 31–33, 43; task metric definitions and corresponding results table; uncertainty/repeat-run/statistical-comparison passages"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Protein pairs represented using sequence and structure-informed embeddings.","Evaluation: Supervised downstream prediction with validation-based tuning/checkpoint selection.","Readout: Area under the ROC curve for binary protein-interaction classification."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["mulan-2025"],"source_locator":"Methods §2.5; Results: HumanPPI; cached text lines 31–33, 43; task metric definitions and corresponding results table; uncertainty/repeat-run/statistical-comparison passages"},"coverage":"limited","gaps":["Protein-pair grouping, negative sampling, metric implementation remain unextracted. Gains in other downstream tasks do not specify this task’s protocol."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Task-specific computational methodology and field context checked in the cited primary-source artifact. Source-backed fields, inapplicable evaluator dimensions and unresolved details are distinguished. Numerical results were not reproduced."}}}}
{"id":"reported-task-7621fa1be55362","kind":"benchmark","name":"protein localization classification","description":"","status":"needs_review","facets":{"areas":["cells-tissues"],"tasks":["protein localization classification"]},"source_ids":["cell-dino-2025"],"links":[{"relation":"dataset","target_id":"reported-dataset-d356eac961cb69"}],"attributes":{"entity_level":"task","version":null,"task":"protein localization classification","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Protein-localization classification evaluates frozen microscopy-image representations using Human Protein Atlas tasks.","summary_source_ids":["cell-dino-2025"],"summary_source_locator":"Methods: Training and evaluation protocol on HPA datasets; Kaggle evaluation protocol; cached text lines 46, 86–94; task metric definitions and corresponding results table","sections":[{"title":"Evaluation methodology","body":"HPA field-of-view and single-cell image tasks. The Kaggle field-of-view evaluation uses five folds to choose and average class thresholds. F1 for protein-localization classification; cell-line accuracy and Cell Painting metrics are separate tasks. The paper includes nearest-neighbour, linear and MLP-based evaluations; these are distinct classifier settings. Routine HPA train/validation analysis and Kaggle public-test generalization are different settings; per-class thresholds are selected on validation folds rather than inferred from the public score.","source_ids":["cell-dino-2025"],"source_locator":"Methods: Training and evaluation protocol on HPA datasets; Kaggle evaluation protocol; cached text lines 46, 86–94; task metric definitions and corresponding results table"}],"facts":[{"label":"Datasets","value":"HPA field-of-view and single-cell image tasks.","status":"source_checked","source_ids":["cell-dino-2025"],"source_locator":"Methods: Training and evaluation protocol on HPA datasets; Kaggle evaluation protocol; cached text lines 46, 86–94; task metric definitions and corresponding results table"},{"label":"Splits","value":"The Kaggle field-of-view evaluation uses five folds to choose and average class thresholds.","status":"source_checked","source_ids":["cell-dino-2025"],"source_locator":"Methods: Training and evaluation protocol on HPA datasets; Kaggle evaluation protocol; cached text lines 46, 86–94; task metric definitions and corresponding results table"},{"label":"Metrics","value":"F1 for protein-localization classification; cell-line accuracy and Cell Painting metrics are separate tasks.","status":"source_checked","source_ids":["cell-dino-2025"],"source_locator":"Methods: Training and evaluation protocol on HPA datasets; Kaggle evaluation protocol; cached text lines 46, 86–94; task metric definitions and corresponding results table"},{"label":"Baselines","value":"The paper includes nearest-neighbour, linear and MLP-based evaluations; these are distinct classifier settings.","status":"source_checked","source_ids":["cell-dino-2025"],"source_locator":"Methods: Training and evaluation protocol on HPA datasets; Kaggle evaluation protocol; cached text lines 46, 86–94; task metric definitions and corresponding results table"},{"label":"Leakage controls","value":"Routine HPA train/validation analysis and Kaggle public-test generalization are different settings; per-class thresholds are selected on validation folds rather than inferred from the public score.","status":"source_checked","source_ids":["cell-dino-2025"],"source_locator":"Methods: Training and evaluation protocol on HPA datasets; Kaggle evaluation protocol; cached text lines 46, 86–94; task metric definitions and corresponding results table"},{"label":"Uncertainty","value":"The protein-localization tables report point F1 or challenge scores. The HPA evaluation methods and S1 Text do not define repeated-seed confidence intervals or an uncertainty estimator for these values; comparisons between protocols retain their distinct classifiers and label access.","status":"unreported","source_ids":["cell-dino-2025","evidence-task-final-a-cell-dino-2025-pcbi-1013828-s001-pdf"],"source_locator":"Methods: Training and evaluation protocol on HPA datasets; Tables 1–3 and 6; S1 Text"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["cell-dino-2025"],"source_locator":"Methods: Training and evaluation protocol on HPA datasets; Kaggle evaluation protocol; cached text lines 46, 86–94; task metric definitions and corresponding results table"},{"label":"Organisms","value":"Human Protein Atlas cells.","status":"source_checked","source_ids":["cell-dino-2025"],"source_locator":"Methods: Training and evaluation protocol on HPA datasets; Kaggle evaluation protocol; cached text lines 46, 86–94; task metric definitions and corresponding results table"},{"label":"Assays","value":"Microscopy images with protein-localization labels.","status":"source_checked","source_ids":["cell-dino-2025"],"source_locator":"Methods: Training and evaluation protocol on HPA datasets; Kaggle evaluation protocol; cached text lines 46, 86–94; task metric definitions and corresponding results table"},{"label":"Allowed inputs","value":"Field-of-view or single-cell images, depending on the task setting.","status":"source_checked","source_ids":["cell-dino-2025"],"source_locator":"Methods: Training and evaluation protocol on HPA datasets; Kaggle evaluation protocol; cached text lines 46, 86–94; task metric definitions and corresponding results table"},{"label":"Adaptation","value":"Frozen-feature nearest-neighbour/linear/MLP evaluation regimes are distinct; threshold selection uses folds.","status":"source_checked","source_ids":["cell-dino-2025"],"source_locator":"Methods: Training and evaluation protocol on HPA datasets; Kaggle evaluation protocol; cached text lines 46, 86–94; task metric definitions and corresponding results table"}],"strengths":[],"limitations":[{"text":"Field-of-view challenge ensembling, single-cell classification and nearest-neighbor evaluation use different classifiers and protocols. Their scores should not be merged into one interchangeable result.","source_ids":["cell-dino-2025","evidence-task-final-a-cell-dino-2025-pcbi-1013828-s001-pdf"],"source_locator":"Methods: Training and evaluation protocol on HPA datasets; Tables 1–3 and 6; S1 Text"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Field-of-view or single-cell images, depending on the task setting.","Evaluation: The Kaggle field-of-view evaluation uses five folds to choose and average class thresholds.","Readout: F1 for protein-localization classification; cell-line accuracy and Cell Painting metrics are separate tasks."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["cell-dino-2025"],"source_locator":"Methods: Training and evaluation protocol on HPA datasets; Kaggle evaluation protocol; cached text lines 46, 86–94; task metric definitions and corresponding results table"},"coverage":"limited","gaps":["Uncertainty: The protein-localization tables report point F1 or challenge scores. The HPA evaluation methods and S1 Text do not define repeated-seed confidence intervals or an uncertainty estimator for these values; comparisons between protocols retain their distinct classifiers and label access."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Targeted full-paper and supplement review of the outstanding task fields, with original dataset metadata checked where accessible. Source-scoped omissions are explicit; no independent benchmark reproduction or numerical-result change."}}}}
{"id":"reported-task-77a32496ce8fe6","kind":"benchmark","name":"protein-ligand binding affinity prediction","description":"","status":"needs_review","facets":{"areas":["molecular-interactions"],"tasks":["protein-ligand binding affinity prediction"]},"source_ids":["ligand-affinity-meta-model-2024"],"links":[{"relation":"dataset","target_id":"reported-dataset-17132fbabd7683"}],"attributes":{"entity_level":"task","version":null,"task":"protein-ligand binding affinity prediction","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Affinity prediction combines refined-set training with core-set and separately filtered general-set evaluations.","summary_source_ids":["ligand-affinity-meta-model-2024"],"summary_source_locator":"Dataset Selection: PDBbind; GeneralSet Benchmark; model training; cached text lines 16, 38, 55–56; task metric definitions and corresponding results table","sections":[{"title":"Evaluation methodology","body":"PDBbind v2020 refined/core collections and BindingDB-trained component models. CASF2016 core complexes are excluded from refined-set training; a filtered general-set subset excludes refined/core entries. RMSE and Pearson correlation coefficient for binding-affinity evaluation; target-specific ranking collections are reported separately. Table 4 compares the docking and sequence-model components with HAC-Net, FAST’s pretrained SG-CNN and KDeep. Training sets and affinity units differ: HAC-Net is retrained, SG-CNN retains its original training, and KDeep uses limited retraining or its server default. PerSpect ML is discussed from published results rather than rerun. PDBbind fine-tuning excludes the CoreSet, and the filtered GeneralSet excludes both RefinedSet and CoreSet. The separate three-target ranking test additionally enforces less than 30% protein similarity. These distinct controls do not establish a BindingDB-overlap audit for the CASF-2016 or GeneralSet rows.","source_ids":["ligand-affinity-meta-model-2024"],"source_locator":"Dataset Selection: PDBbind; GeneralSet Benchmark; model training; cached text lines 16, 38, 55–56; task metric definitions and corresponding results table; Methods: Comparison to Other Tools; Results: Comparison with Structure-Based Tools; Table 4; Methods: Fine-Tuning of the BDB-Trained Models Using PDBbind; GeneralSet Benchmark; Ranking Benchmark"}],"facts":[{"label":"Datasets","value":"PDBbind v2020 refined/core collections and BindingDB-trained component models.","status":"source_checked","source_ids":["ligand-affinity-meta-model-2024"],"source_locator":"Dataset Selection: PDBbind; GeneralSet Benchmark; model training; cached text lines 16, 38, 55–56; task metric definitions and corresponding results table"},{"label":"Splits","value":"CASF2016 core complexes are excluded from refined-set training; a filtered general-set subset excludes refined/core entries.","status":"source_checked","source_ids":["ligand-affinity-meta-model-2024"],"source_locator":"Dataset Selection: PDBbind; GeneralSet Benchmark; model training; cached text lines 16, 38, 55–56; task metric definitions and corresponding results table"},{"label":"Metrics","value":"RMSE and Pearson correlation coefficient for binding-affinity evaluation; target-specific ranking collections are reported separately.","status":"source_checked","source_ids":["ligand-affinity-meta-model-2024"],"source_locator":"Dataset Selection: PDBbind; GeneralSet Benchmark; model training; cached text lines 16, 38, 55–56; task metric definitions and corresponding results table"},{"label":"Baselines","value":"Table 4 compares the docking and sequence-model components with HAC-Net, FAST’s pretrained SG-CNN and KDeep. Training sets and affinity units differ: HAC-Net is retrained, SG-CNN retains its original training, and KDeep uses limited retraining or its server default. PerSpect ML is discussed from published results rather than rerun.","status":"source_checked","source_ids":["ligand-affinity-meta-model-2024"],"source_locator":"Methods: Comparison to Other Tools; Results: Comparison with Structure-Based Tools; Table 4"},{"label":"Leakage controls","value":"PDBbind fine-tuning excludes the CoreSet, and the filtered GeneralSet excludes both RefinedSet and CoreSet. The separate three-target ranking test additionally enforces less than 30% protein similarity. These distinct controls do not establish a BindingDB-overlap audit for the CASF-2016 or GeneralSet rows.","status":"source_checked","source_ids":["ligand-affinity-meta-model-2024"],"source_locator":"Methods: Fine-Tuning of the BDB-Trained Models Using PDBbind; GeneralSet Benchmark; Ranking Benchmark"},{"label":"Uncertainty","value":"One PDBbind model group uses repeated five-fold cross-validation.","status":"source_checked","source_ids":["ligand-affinity-meta-model-2024"],"source_locator":"Dataset Selection: PDBbind; GeneralSet Benchmark; model training; cached text lines 16, 38, 55–56; task metric definitions and corresponding results table"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["ligand-affinity-meta-model-2024"],"source_locator":"Dataset Selection: PDBbind; GeneralSet Benchmark; model training; cached text lines 16, 38, 55–56; task metric definitions and corresponding results table"},{"label":"Organisms","value":"CASF-2016 and the filtered PDBbind2020 GeneralSet pool protein–ligand complexes. The dataset and benchmark methods do not provide a species census for the scored subsets; target accession identities, rather than an assumed single organism, define their biological context.","status":"unreported","source_ids":["ligand-affinity-meta-model-2024"],"source_locator":"Methods: GeneralSet Benchmark and data preparation; Table 4"},{"label":"Assays","value":"Protein–ligand affinity and structural labels.","status":"source_checked","source_ids":["ligand-affinity-meta-model-2024"],"source_locator":"Dataset Selection: PDBbind; GeneralSet Benchmark; model training; cached text lines 16, 38, 55–56; task metric definitions and corresponding results table"},{"label":"Allowed inputs","value":"Protein–ligand representations used by component scoring models.","status":"source_checked","source_ids":["ligand-affinity-meta-model-2024"],"source_locator":"Dataset Selection: PDBbind; GeneralSet Benchmark; model training; cached text lines 16, 38, 55–56; task metric definitions and corresponding results table"},{"label":"Adaptation","value":"Supervised meta-model/component fitting; CASF core examples are excluded from the refined training collection.","status":"source_checked","source_ids":["ligand-affinity-meta-model-2024"],"source_locator":"Dataset Selection: PDBbind; GeneralSet Benchmark; model training; cached text lines 16, 38, 55–56; task metric definitions and corresponding results table"}],"strengths":[],"limitations":[{"text":"Failed docking predictions are excluded from benchmark subsets, changing test coverage. BindingDB-trained and PDBbind-trained models have different training provenance.","source_ids":["ligand-affinity-meta-model-2024"],"source_locator":"Dataset Selection: PDBbind; GeneralSet Benchmark; model training; cached text lines 16, 38, 55–56; task metric definitions and corresponding results table"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Protein–ligand representations used by component scoring models.","Evaluation: CASF2016 core complexes are excluded from refined-set training; a filtered general-set subset excludes refined/core entries.","Readout: RMSE and Pearson correlation coefficient for binding-affinity evaluation; target-specific ranking collections are reported separately."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["ligand-affinity-meta-model-2024"],"source_locator":"Dataset Selection: PDBbind; GeneralSet Benchmark; model training; cached text lines 16, 38, 55–56; task metric definitions and corresponding results table"},"coverage":"limited","gaps":["Organisms: CASF-2016 and the filtered PDBbind2020 GeneralSet pool protein–ligand complexes. The dataset and benchmark methods do not provide a species census for the scored subsets; target accession identities, rather than an assumed single organism, define their biological context."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Targeted full-paper and supplement review of the outstanding task fields, with original dataset metadata checked where accessible. Source-scoped omissions are explicit; no independent benchmark reproduction or numerical-result change."}}}}
{"id":"reported-task-786c09824e9bf5","kind":"benchmark","name":"clathrin protein classification","description":"","status":"needs_review","facets":{"areas":["proteins-complexes"],"tasks":["clathrin protein classification"]},"source_ids":["clathrin-plm-2025"],"links":[{"relation":"dataset","target_id":"reported-dataset-0aab382ca2c063"}],"attributes":{"entity_level":"task","version":null,"task":"clathrin protein classification","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Clathrin classification uses cross-validation and multiple independent benchmark datasets with different redundancy filters.","summary_source_ids":["clathrin-plm-2025"],"summary_source_locator":"Methods: Dataset construction; Performance evaluation; Results: independent test datasets; cached text lines 9–10, 24–25, 54","sections":[{"title":"Evaluation methodology","body":"Le2019 and Zhang2020-derived clathrin/non-clathrin sequence collections. Ten-fold cross-validation on training data plus named independent CLA test sets. Accuracy, AUC, Matthews correlation coefficient, F1, sensitivity and specificity are reported. The Performance evaluation subsection describes ten-fold evaluation and early stopping but does not specify a universal probability threshold for every comparator. deep-clathrin includes literature-reported scores; DeepCLA is reimplemented for comparison. The Zhang2020-derived collection uses BLAST redundancy filtering at 0.7, and the new dataset uses CD-HIT at 0.6. These dataset filters do not establish absence from the pretrained protein encoders’ corpora. Feature subsets are evaluated on both cross-validation and independent-test performance, which limits treating that test as untouched model selection.","source_ids":["clathrin-plm-2025"],"source_locator":"Methods: Dataset construction; Performance evaluation; Results: independent test datasets; cached text lines 9–10, 24–25, 54; Materials and methods: Performance evaluation; Materials and methods: Dataset construction and Overall framework of PLM-CLA; Results: The effect of feature selection methods on the predictive performance"}],"facts":[{"label":"Datasets","value":"Le2019 and Zhang2020-derived clathrin/non-clathrin sequence collections.","status":"source_checked","source_ids":["clathrin-plm-2025"],"source_locator":"Methods: Dataset construction; Performance evaluation; Results: independent test datasets; cached text lines 9–10, 24–25, 54"},{"label":"Splits","value":"Ten-fold cross-validation on training data plus named independent CLA test sets.","status":"source_checked","source_ids":["clathrin-plm-2025"],"source_locator":"Methods: Dataset construction; Performance evaluation; Results: independent test datasets; cached text lines 9–10, 24–25, 54"},{"label":"Metrics","value":"Accuracy, AUC, Matthews correlation coefficient, F1, sensitivity and specificity are reported. The Performance evaluation subsection describes ten-fold evaluation and early stopping but does not specify a universal probability threshold for every comparator.","status":"source_checked","source_ids":["clathrin-plm-2025"],"source_locator":"Materials and methods: Performance evaluation"},{"label":"Baselines","value":"deep-clathrin includes literature-reported scores; DeepCLA is reimplemented for comparison.","status":"source_checked","source_ids":["clathrin-plm-2025"],"source_locator":"Methods: Dataset construction; Performance evaluation; Results: independent test datasets; cached text lines 9–10, 24–25, 54"},{"label":"Leakage controls","value":"The Zhang2020-derived collection uses BLAST redundancy filtering at 0.7, and the new dataset uses CD-HIT at 0.6. These dataset filters do not establish absence from the pretrained protein encoders’ corpora. Feature subsets are evaluated on both cross-validation and independent-test performance, which limits treating that test as untouched model selection.","status":"source_checked","source_ids":["clathrin-plm-2025"],"source_locator":"Materials and methods: Dataset construction and Overall framework of PLM-CLA; Results: The effect of feature selection methods on the predictive performance"},{"label":"Uncertainty","value":"The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","status":"unreported","source_ids":["clathrin-plm-2025"],"source_locator":"Methods: Dataset construction; Performance evaluation; Results: independent test datasets; cached text lines 9–10, 24–25, 54"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["clathrin-plm-2025"],"source_locator":"Methods: Dataset construction; Performance evaluation; Results: independent test datasets; cached text lines 9–10, 24–25, 54"},{"label":"Organisms","value":"The paper does not enumerate organisms. Its pinned released CSVs contain sequence identifiers and amino-acid sequences, without organism columns; one dataset uses synthetic row identifiers. Consequently, a complete source-species inventory is not established by the supplied benchmark metadata.","status":"unreported","source_ids":["clathrin-plm-2025","evidence-task-final-a-clathrin-dataset-clathrin0-6-csv","evidence-task-final-a-clathrin-dataset-clathrin0-7-csv","evidence-task-final-a-clathrin-dataset-clathrin1-0-csv","evidence-task-final-a-clathrin-readme-md"],"source_locator":"Dataset construction; pinned repository README and Dataset/Clathrin0.6.csv, Clathrin0.7.csv and Clathrin1.0.csv headers"},{"label":"Assays","value":"Clathrin/non-clathrin sequence labels.","status":"source_checked","source_ids":["clathrin-plm-2025"],"source_locator":"Methods: Dataset construction; Performance evaluation; Results: independent test datasets; cached text lines 9–10, 24–25, 54"},{"label":"Allowed inputs","value":"Protein amino-acid sequences represented by protein language models.","status":"source_checked","source_ids":["clathrin-plm-2025"],"source_locator":"Methods: Dataset construction; Performance evaluation; Results: independent test datasets; cached text lines 9–10, 24–25, 54"},{"label":"Adaptation","value":"Supervised classification with cross-validation and independent test collections.","status":"source_checked","source_ids":["clathrin-plm-2025"],"source_locator":"Methods: Dataset construction; Performance evaluation; Results: independent test datasets; cached text lines 9–10, 24–25, 54"}],"strengths":[],"limitations":[{"text":"Literature scores and newly executed comparisons have different provenance. The several CLA datasets must not be treated as interchangeable.","source_ids":["clathrin-plm-2025"],"source_locator":"Methods: Dataset construction; Performance evaluation; Results: independent test datasets; cached text lines 9–10, 24–25, 54"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Protein amino-acid sequences represented by protein language models.","Evaluation: Ten-fold cross-validation on training data plus named independent CLA test sets.","Readout: Accuracy, AUC, Matthews correlation coefficient, F1, sensitivity and specificity are reported. The Performance evaluation subsection describes ten-fold evaluation and early stopping but does not specify a universal probability threshold for every comparator."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["clathrin-plm-2025"],"source_locator":"Methods: Dataset construction; Performance evaluation; Results: independent test datasets; cached text lines 9–10, 24–25, 54; Materials and methods: Performance evaluation"},"coverage":"limited","gaps":["Uncertainty: The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","Organisms: The paper does not enumerate organisms. Its pinned released CSVs contain sequence identifiers and amino-acid sequences, without organism columns; one dataset uses synthetic row identifiers. Consequently, a complete source-species inventory is not established by the supplied benchmark metadata."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Targeted full-paper and supplement review of the outstanding task fields, with original dataset metadata checked where accessible. Source-scoped omissions are explicit; no independent benchmark reproduction or numerical-result change."}}}}
{"id":"reported-task-7efe245cc94ee5","kind":"benchmark","name":"Combinatorial cell-label classification","description":"","status":"needs_review","facets":{"areas":["cells-tissues"],"tasks":["Combinatorial cell-label classification"]},"source_ids":["cell2sentence-2024"],"links":[{"relation":"dataset","target_id":"reported-dataset-87e91d9d6e6f4b"}],"attributes":{"entity_level":"task","version":null,"task":"Combinatorial cell-label classification","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Combinatorial cell-label classification tests exact labels and partial correctness across single-cell and bulk-expression datasets.","summary_source_ids":["cell2sentence-2024"],"summary_source_locator":"Experiment 2: cell label classification; Table 3; Fine-Tuning Datasets; cached text lines 47–53, 77–78, 113","sections":[{"title":"Evaluation methodology","body":"Cytokine stimulation, L1000 and GTEx datasets with multipart metadata labels. The cytokine label-classification setting includes all label combinations during training; held-out combinations belong to a separate generation task. Accuracy and AUROC are reported separately for exact full-label matching and partial-label credit. k-nearest-neighbour, XGBoost, Geneformer and scGPT comparators. For combinatorial label classification, all combinations of cytokine perturbations are included in training. Holding out 10 of 140 cytokine combinations belongs to the separate perturbed-cell-generation experiment, not this classifier. L1000 and GTEx provide bulk-expression evaluation outside the single-cell fine-tuning distribution.","source_ids":["cell2sentence-2024"],"source_locator":"Experiment 2: cell label classification; Table 3; Fine-Tuning Datasets; cached text lines 47–53, 77–78, 113; Fine-Tuning Datasets; Experiment 2: combinatorial label classification"}],"facts":[{"label":"Datasets","value":"Cytokine stimulation, L1000 and GTEx datasets with multipart metadata labels.","status":"source_checked","source_ids":["cell2sentence-2024"],"source_locator":"Experiment 2: cell label classification; Table 3; Fine-Tuning Datasets; cached text lines 47–53, 77–78, 113"},{"label":"Splits","value":"The cytokine label-classification setting includes all label combinations during training; held-out combinations belong to a separate generation task.","status":"source_checked","source_ids":["cell2sentence-2024"],"source_locator":"Experiment 2: cell label classification; Table 3; Fine-Tuning Datasets; cached text lines 47–53, 77–78, 113"},{"label":"Metrics","value":"Accuracy and AUROC are reported separately for exact full-label matching and partial-label credit.","status":"source_checked","source_ids":["cell2sentence-2024"],"source_locator":"Experiment 2: cell label classification; Table 3; Fine-Tuning Datasets; cached text lines 47–53, 77–78, 113"},{"label":"Baselines","value":"k-nearest-neighbour, XGBoost, Geneformer and scGPT comparators.","status":"source_checked","source_ids":["cell2sentence-2024"],"source_locator":"Experiment 2: cell label classification; Table 3; Fine-Tuning Datasets; cached text lines 47–53, 77–78, 113"},{"label":"Leakage controls","value":"For combinatorial label classification, all combinations of cytokine perturbations are included in training. Holding out 10 of 140 cytokine combinations belongs to the separate perturbed-cell-generation experiment, not this classifier. L1000 and GTEx provide bulk-expression evaluation outside the single-cell fine-tuning distribution.","status":"source_checked","source_ids":["cell2sentence-2024"],"source_locator":"Fine-Tuning Datasets; Experiment 2: combinatorial label classification"},{"label":"Uncertainty","value":"Three experimental repeats are reported for each dataset.","status":"source_checked","source_ids":["cell2sentence-2024"],"source_locator":"Experiment 2: cell label classification; Table 3; Fine-Tuning Datasets; cached text lines 47–53, 77–78, 113"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["cell2sentence-2024"],"source_locator":"Experiment 2: cell label classification; Table 3; Fine-Tuning Datasets; cached text lines 47–53, 77–78, 113"},{"label":"Organisms","value":"The cytokine-stimulation arm explicitly uses human peripheral blood mononuclear cells. The two additional arms use L1000 and GTEx bulk-expression collections, with their cell-line and tissue identities retained in combinatorial labels rather than treated as the same PBMC population.","status":"source_checked","source_ids":["cell2sentence-2024"],"source_locator":"Experiments: Experiment 2/cell label prediction/Methodology; Fine-Tuning Datasets"},{"label":"Assays","value":"Expression measurements with multipart cell/tissue/stimulation labels.","status":"source_checked","source_ids":["cell2sentence-2024"],"source_locator":"Experiment 2: cell label classification; Table 3; Fine-Tuning Datasets; cached text lines 47–53, 77–78, 113"},{"label":"Allowed inputs","value":"Gene-expression-derived cell representations.","status":"source_checked","source_ids":["cell2sentence-2024"],"source_locator":"Experiment 2: cell label classification; Table 3; Fine-Tuning Datasets; cached text lines 47–53, 77–78, 113"},{"label":"Adaptation","value":"Supervised label classification includes all label combinations in training; compositional generation is a different experiment.","status":"source_checked","source_ids":["cell2sentence-2024"],"source_locator":"Experiment 2: cell label classification; Table 3; Fine-Tuning Datasets; cached text lines 47–53, 77–78, 113"}],"strengths":[],"limitations":[{"text":"The combinatorial classifier includes every cytokine-label combination during training. A held-out-combination claim belongs to the separate generation task; exact and partial label credit are also different metrics.","source_ids":["cell2sentence-2024"],"source_locator":"Fine-Tuning Datasets; Experiment 2: combinatorial label classification; Experiments: Experiment 2/cell label prediction/Methodology; Fine-Tuning Datasets"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Gene-expression-derived cell representations.","Evaluation: The cytokine label-classification setting includes all label combinations during training; held-out combinations belong to a separate generation task.","Readout: Accuracy and AUROC are reported separately for exact full-label matching and partial-label credit."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["cell2sentence-2024"],"source_locator":"Experiment 2: cell label classification; Table 3; Fine-Tuning Datasets; cached text lines 47–53, 77–78, 113"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Targeted full-paper and supplement review of the outstanding task fields, with original dataset metadata checked where accessible. Source-scoped omissions are explicit; no independent benchmark reproduction or numerical-result change."}}}}
{"id":"reported-task-82fc7843f07324","kind":"benchmark","name":"human RNA 2-prime-O-methylation site prediction","description":"","status":"needs_review","facets":{"areas":["rna-transcriptomes"],"tasks":["human RNA 2-prime-O-methylation site prediction"]},"source_ids":["2ome-lm-2025"],"links":[{"relation":"dataset","target_id":"reported-dataset-bd3d8e7d6cd196"}],"attributes":{"entity_level":"task","version":null,"task":"human RNA 2-prime-O-methylation site prediction","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Human RNA methylation-site classification is evaluated on a curated, balanced dataset and an independent test partition.","summary_source_ids":["2ome-lm-2025"],"summary_source_locator":"Methods §§2.1, 2.4; Results §§3.1–3.2; cached text lines 11–14, 35–36, 43–48","sections":[{"title":"Evaluation methodology","body":"RMBase v3.0 and multiple experimentally annotated RNA modification datasets supply positive and negative examples. An 80:20 training/test partition is followed by five-fold cross-validation within training data. Accuracy, F1, precision, recall, AUROC, AUPR and MCC; cross-validation results are averaged across folds. GloVe/word2vec neural baselines; NmRF, H2Opred and Meta-2OM web servers; BERT2OME retrained on the benchmark. The checked dataset/split passages specify sample partitioning but do not specify a gene-, donor- or overlapping-window exclusion rule. The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","source_ids":["2ome-lm-2025"],"source_locator":"Methods §§2.1, 2.4; Results §§3.1–3.2; cached text lines 11–14, 35–36, 43–48"}],"facts":[{"label":"Datasets","value":"RMBase v3.0 and multiple experimentally annotated RNA modification datasets supply positive and negative examples.","status":"source_checked","source_ids":["2ome-lm-2025"],"source_locator":"Methods §§2.1, 2.4; Results §§3.1–3.2; cached text lines 11–14, 35–36, 43–48"},{"label":"Splits","value":"An 80:20 training/test partition is followed by five-fold cross-validation within training data.","status":"source_checked","source_ids":["2ome-lm-2025"],"source_locator":"Methods §§2.1, 2.4; Results §§3.1–3.2; cached text lines 11–14, 35–36, 43–48"},{"label":"Metrics","value":"Accuracy, F1, precision, recall, AUROC, AUPR and MCC; cross-validation results are averaged across folds.","status":"source_checked","source_ids":["2ome-lm-2025"],"source_locator":"Methods §§2.1, 2.4; Results §§3.1–3.2; cached text lines 11–14, 35–36, 43–48"},{"label":"Baselines","value":"GloVe/word2vec neural baselines; NmRF, H2Opred and Meta-2OM web servers; BERT2OME retrained on the benchmark.","status":"source_checked","source_ids":["2ome-lm-2025"],"source_locator":"Methods §§2.1, 2.4; Results §§3.1–3.2; cached text lines 11–14, 35–36, 43–48"},{"label":"Leakage controls","value":"The checked dataset/split passages specify sample partitioning but do not specify a gene-, donor- or overlapping-window exclusion rule.","status":"unreported","source_ids":["2ome-lm-2025"],"source_locator":"Methods §§2.1, 2.4; Results §§3.1–3.2; cached text lines 11–14, 35–36, 43–48"},{"label":"Uncertainty","value":"The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","status":"unreported","source_ids":["2ome-lm-2025"],"source_locator":"Methods §§2.1, 2.4; Results §§3.1–3.2; cached text lines 11–14, 35–36, 43–48"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["2ome-lm-2025"],"source_locator":"Methods §§2.1, 2.4; Results §§3.1–3.2; cached text lines 11–14, 35–36, 43–48"},{"label":"Organisms","value":"Human.","status":"source_checked","source_ids":["2ome-lm-2025"],"source_locator":"Methods §§2.1, 2.4; Results §§3.1–3.2; cached text lines 11–14, 35–36, 43–48"},{"label":"Assays","value":"RNA modification-site annotations from RMBase and experimental datasets.","status":"source_checked","source_ids":["2ome-lm-2025"],"source_locator":"Methods §§2.1, 2.4; Results §§3.1–3.2; cached text lines 11–14, 35–36, 43–48"},{"label":"Allowed inputs","value":"RNA sequence windows around candidate modification sites.","status":"source_checked","source_ids":["2ome-lm-2025"],"source_locator":"Methods §§2.1, 2.4; Results §§3.1–3.2; cached text lines 11–14, 35–36, 43–48"},{"label":"Adaptation","value":"Supervised site classifier; comparator servers and retrained models have distinct training provenance.","status":"source_checked","source_ids":["2ome-lm-2025"],"source_locator":"Methods §§2.1, 2.4; Results §§3.1–3.2; cached text lines 11–14, 35–36, 43–48"}],"strengths":[],"limitations":[{"text":"Web-server comparisons and retrained comparators do not have identical training provenance; overlap with their training corpora has not been audited here.","source_ids":["2ome-lm-2025"],"source_locator":"Methods §§2.1, 2.4; Results §§3.1–3.2; cached text lines 11–14, 35–36, 43–48"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: RNA sequence windows around candidate modification sites.","Evaluation: Supervised site classifier; comparator servers and retrained models have distinct training provenance.","Readout: Accuracy, F1, precision, recall, AUROC, AUPR and MCC; cross-validation results are averaged across folds."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["2ome-lm-2025"],"source_locator":"Methods §§2.1, 2.4; Results §§3.1–3.2; cached text lines 11–14, 35–36, 43–48"},"coverage":"limited","gaps":["Web-server comparisons and retrained comparators do not have identical training provenance; overlap with their training corpora has not been audited here."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Task-specific computational methodology and field context checked in the cited primary-source artifact. Source-backed fields, inapplicable evaluator dimensions and unresolved details are distinguished. Numerical results were not reproduced."}}}}
{"id":"reported-task-83be0998084c91","kind":"benchmark","name":"protein variant-effect classification","description":"","status":"needs_review","facets":{"areas":["proteins-complexes"],"tasks":["protein variant-effect classification"]},"source_ids":["structure-informed-plm-2025"],"links":[{"relation":"dataset","target_id":"reported-dataset-2eaa2a051d45ee"}],"attributes":{"entity_level":"task","version":null,"task":"protein variant-effect classification","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Structure-informed variant-effect evaluation compares ranking and binary classification against experimental protein measurements.","summary_source_ids":["evidence-benchmark-structure-informed-html"],"summary_source_locator":"Results: Structure-Informed pLMs Predict Variant Fitness Robustly; Tables 2 and 4; Methods: Datasets, Mutation effect datasets, Competing methods","sections":[{"title":"Evaluation methodology","body":"DeepSequence-derived missense-variant DMS datasets, excluding tRNA and viral-family sets; available family sequences and structures support adaptation. A labeled validation subset selects the structural weighting and model; labels are not used to fit the model parameters. The exact held-out membership after selection remains unextracted. Spearman correlation evaluates ranking; AUROC and AUPRC evaluate high/low labels thresholded relative to the wild type. Table 4 compares sequence, structure and combined score variants. PSSM, EVmutation, DeepSequence, Wavenet and several protein language models; competing scores are obtained from the ESM-1v repository. Family-specific adaptation and validation-label model selection are part of the protocol; it should not be represented as selection without labeled data. One-sided paired Wilcoxon signed-rank comparisons across assay performances are reported for the main method comparison.","source_ids":["evidence-benchmark-structure-informed-html"],"source_locator":"Results: Structure-Informed pLMs Predict Variant Fitness Robustly; Tables 2 and 4; Methods: Datasets, Mutation effect datasets, Competing methods"}],"facts":[{"label":"Datasets","value":"DeepSequence-derived missense-variant DMS datasets, excluding tRNA and viral-family sets; available family sequences and structures support adaptation.","status":"source_checked","source_ids":["evidence-benchmark-structure-informed-html"],"source_locator":"Results: Structure-Informed pLMs Predict Variant Fitness Robustly; Tables 2 and 4; Methods: Datasets, Mutation effect datasets, Competing methods"},{"label":"Splits","value":"A labeled validation subset selects the structural weighting and model; labels are not used to fit the model parameters. The exact held-out membership after selection remains unextracted.","status":"source_checked","source_ids":["evidence-benchmark-structure-informed-html"],"source_locator":"Results: Structure-Informed pLMs Predict Variant Fitness Robustly; Tables 2 and 4; Methods: Datasets, Mutation effect datasets, Competing methods"},{"label":"Metrics","value":"Spearman correlation evaluates ranking; AUROC and AUPRC evaluate high/low labels thresholded relative to the wild type. Table 4 compares sequence, structure and combined score variants.","status":"source_checked","source_ids":["evidence-benchmark-structure-informed-html"],"source_locator":"Results: Structure-Informed pLMs Predict Variant Fitness Robustly; Tables 2 and 4; Methods: Datasets, Mutation effect datasets, Competing methods"},{"label":"Baselines","value":"PSSM, EVmutation, DeepSequence, Wavenet and several protein language models; competing scores are obtained from the ESM-1v repository.","status":"source_checked","source_ids":["evidence-benchmark-structure-informed-html"],"source_locator":"Results: Structure-Informed pLMs Predict Variant Fitness Robustly; Tables 2 and 4; Methods: Datasets, Mutation effect datasets, Competing methods"},{"label":"Leakage controls","value":"Family-specific adaptation and validation-label model selection are part of the protocol; it should not be represented as selection without labeled data.","status":"source_checked","source_ids":["evidence-benchmark-structure-informed-html"],"source_locator":"Results: Structure-Informed pLMs Predict Variant Fitness Robustly; Tables 2 and 4; Methods: Datasets, Mutation effect datasets, Competing methods"},{"label":"Uncertainty","value":"One-sided paired Wilcoxon signed-rank comparisons across assay performances are reported for the main method comparison.","status":"source_checked","source_ids":["evidence-benchmark-structure-informed-html"],"source_locator":"Results: Structure-Informed pLMs Predict Variant Fitness Robustly; Tables 2 and 4; Methods: Datasets, Mutation effect datasets, Competing methods"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["evidence-benchmark-structure-informed-html"],"source_locator":"Results: Structure-Informed pLMs Predict Variant Fitness Robustly; Tables 2 and 4; Methods: Datasets, Mutation effect datasets, Competing methods"},{"label":"Organisms","value":"Nonviral protein families in the selected DMS collection.","status":"source_checked","source_ids":["evidence-benchmark-structure-informed-html"],"source_locator":"Results: Structure-Informed pLMs Predict Variant Fitness Robustly; Tables 2 and 4; Methods: Datasets, Mutation effect datasets, Competing methods"},{"label":"Assays","value":"DMS missense-variant measurements.","status":"source_checked","source_ids":["evidence-benchmark-structure-informed-html"],"source_locator":"Results: Structure-Informed pLMs Predict Variant Fitness Robustly; Tables 2 and 4; Methods: Datasets, Mutation effect datasets, Competing methods"},{"label":"Allowed inputs","value":"Protein sequence; available family sequences/structures are used during adaptation.","status":"source_checked","source_ids":["evidence-benchmark-structure-informed-html"],"source_locator":"Results: Structure-Informed pLMs Predict Variant Fitness Robustly; Tables 2 and 4; Methods: Datasets, Mutation effect datasets, Competing methods"},{"label":"Adaptation","value":"Unsupervised family adaptation plus labeled-validation model selection; no fitness-label fitting of model parameters.","status":"source_checked","source_ids":["evidence-benchmark-structure-informed-html"],"source_locator":"Results: Structure-Informed pLMs Predict Variant Fitness Robustly; Tables 2 and 4; Methods: Datasets, Mutation effect datasets, Competing methods"}],"strengths":[{"text":"The paper explicitly distinguishes parameter fitting from validation-label model selection.","source_ids":["evidence-benchmark-structure-informed-html"],"source_locator":"Results: Structure-Informed pLMs Predict Variant Fitness Robustly; Tables 2 and 4; Methods: Datasets, Mutation effect datasets, Competing methods"}],"limitations":[{"text":"The final journal Table 4 scoring variant is distinct from the main Table 2 configuration. Exact test membership, scoring configuration and source version must remain attached to each reported result.","source_ids":["evidence-benchmark-structure-informed-html"],"source_locator":"Results: Structure-Informed pLMs Predict Variant Fitness Robustly; Tables 2 and 4; Methods: Datasets, Mutation effect datasets, Competing methods"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Protein sequence; available family sequences/structures are used during adaptation.","Evaluation: Unsupervised family adaptation plus labeled-validation model selection; no fitness-label fitting of model parameters.","Readout: Spearman correlation evaluates ranking; AUROC and AUPRC evaluate high/low labels thresholded relative to the wild type. Table 4 compares sequence, structure and combined score variants."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["evidence-benchmark-structure-informed-html"],"source_locator":"Results: Structure-Informed pLMs Predict Variant Fitness Robustly; Tables 2 and 4; Methods: Datasets, Mutation effect datasets, Competing methods"},"coverage":"limited","gaps":["The final journal Table 4 scoring variant is distinct from the main Table 2 configuration. Exact test membership, scoring configuration and source version must remain attached to each reported result."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Task-specific computational methodology and field context checked in the cited primary-source artifact. Source-backed fields, inapplicable evaluator dimensions and unresolved details are distinguished. Numerical results were not reproduced."}}}}
{"id":"reported-task-8406b6aabfb8c0","kind":"benchmark","name":"Mock-community MAG taxonomy classification","description":"","status":"needs_review","facets":{"areas":["microbes-communities"],"tasks":["Mock-community MAG taxonomy classification"]},"source_ids":["kmetashot-2025"],"links":[{"relation":"dataset","target_id":"reported-dataset-8cad416ddc80dc"}],"attributes":{"entity_level":"task","version":null,"task":"Mock-community MAG taxonomy classification","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Metagenome-assembled-genome taxonomy is evaluated on simulated and experimentally sequenced mock communities.","summary_source_ids":["kmetashot-2025"],"summary_source_locator":"Methods: CAMI datasets; real dataset; cached text lines 36–42","sections":[{"title":"Evaluation methodology","body":"CAMI II community datasets and a separately sequenced mock community with known reference genomes. Community-specific evaluations compare inferred MAG classifications against reference labels; these are not supervised train/test partitions. Sensitivity, precision, false-discovery rate, balanced accuracy and F1. GTDBtk and CAMITAX; two assembly methods are also assessed for the real mock community. The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","source_ids":["kmetashot-2025"],"source_locator":"Methods: CAMI datasets; real dataset; cached text lines 36–42"}],"facts":[{"label":"Datasets","value":"CAMI II community datasets and a separately sequenced mock community with known reference genomes.","status":"source_checked","source_ids":["kmetashot-2025"],"source_locator":"Methods: CAMI datasets; real dataset; cached text lines 36–42"},{"label":"Splits","value":"Community-specific evaluations compare inferred MAG classifications against reference labels; these are not supervised train/test partitions.","status":"source_checked","source_ids":["kmetashot-2025"],"source_locator":"Methods: CAMI datasets; real dataset; cached text lines 36–42"},{"label":"Metrics","value":"Sensitivity, precision, false-discovery rate, balanced accuracy and F1.","status":"source_checked","source_ids":["kmetashot-2025"],"source_locator":"Methods: CAMI datasets; real dataset; cached text lines 36–42"},{"label":"Baselines","value":"GTDBtk and CAMITAX; two assembly methods are also assessed for the real mock community.","status":"source_checked","source_ids":["kmetashot-2025"],"source_locator":"Methods: CAMI datasets; real dataset; cached text lines 36–42"},{"label":"Leakage controls","value":"HMP genomes are used to tune the assignment threshold, then separate CAMI and mock-community datasets are evaluated. HMP includes both reference-represented and unrepresented taxa, so the tuning comparison is not wholly out of reference; the paper reports these groups separately.","status":"source_checked","source_ids":["kmetashot-2025"],"source_locator":"Methods: HMP testing and mock communities; Results: HMP benchmarking; cached paragraphs 31–42, 54–59"},{"label":"Uncertainty","value":"The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","status":"unreported","source_ids":["kmetashot-2025"],"source_locator":"Methods: CAMI datasets; real dataset; cached text lines 36–42"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["kmetashot-2025"],"source_locator":"Methods: CAMI datasets; real dataset; cached text lines 36–42"},{"label":"Organisms","value":"Microbial communities with known reference genomes.","status":"source_checked","source_ids":["kmetashot-2025"],"source_locator":"Methods: CAMI datasets; real dataset; cached text lines 36–42"},{"label":"Assays","value":"Simulated CAMI data and a measured mock community.","status":"source_checked","source_ids":["kmetashot-2025"],"source_locator":"Methods: CAMI datasets; real dataset; cached text lines 36–42"},{"label":"Allowed inputs","value":"Metagenome-assembled genomes and taxonomic reference resources.","status":"source_checked","source_ids":["kmetashot-2025"],"source_locator":"Methods: CAMI datasets; real dataset; cached text lines 36–42"},{"label":"Adaptation","value":"Genome classification against references; community evaluation is separate from supervised training splits.","status":"source_checked","source_ids":["kmetashot-2025"],"source_locator":"Methods: CAMI datasets; real dataset; cached text lines 36–42"}],"strengths":[],"limitations":[{"text":"Strain-level labels are refined using reference-genome similarity and assessed before/after refinement. Refined labels and original labels are distinct ground-truth conditions.","source_ids":["kmetashot-2025"],"source_locator":"Methods: CAMI datasets; real dataset; cached text lines 36–42"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Metagenome-assembled genomes and taxonomic reference resources.","Evaluation: Genome classification against references; community evaluation is separate from supervised training splits.","Readout: Sensitivity, precision, false-discovery rate, balanced accuracy and F1."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["kmetashot-2025"],"source_locator":"Methods: CAMI datasets; real dataset; cached text lines 36–42"},"coverage":"limited","gaps":["Uncertainty: The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Relevant full-paper computational evaluation sections, tables/captions and cited supplementary task passages were reviewed. Reporting omissions are scoped to the inspected sources. Original numerical results are unchanged."}}}}
{"id":"reported-task-86a628af87ff8f","kind":"benchmark","name":"enhancer recognition","description":"","status":"needs_review","facets":{"areas":["dna-genomes"],"tasks":["enhancer recognition"]},"source_ids":["dnabert2-enhancer-2025"],"links":[{"relation":"dataset","target_id":"reported-dataset-6212e779949708"}],"attributes":{"entity_level":"task","version":null,"task":"enhancer recognition","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Enhancer recognition is evaluated separately from classification of enhancer strength.","summary_source_ids":["dnabert2-enhancer-2025"],"summary_source_locator":"Methods: Benchmark dataset; Performance evaluation metrics; cached text lines 12–13, 47–48","sections":[{"title":"Evaluation methodology","body":"Previously published enhancer/non-enhancer sequences with strong/weak labels for the enhancer subset. Five-fold cross-validation is performed within Liu’s training data; separately listed test subsets distinguish enhancer detection from enhancer-strength classification. Specificity, sensitivity, accuracy, balanced accuracy, MCC and AUC. The source describes CD-HIT filtering of highly similar sequences. The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","source_ids":["dnabert2-enhancer-2025"],"source_locator":"Methods: Benchmark dataset; Performance evaluation metrics; cached text lines 12–13, 47–48"}],"facts":[{"label":"Datasets","value":"Previously published enhancer/non-enhancer sequences with strong/weak labels for the enhancer subset.","status":"source_checked","source_ids":["dnabert2-enhancer-2025"],"source_locator":"Methods: Benchmark dataset; Performance evaluation metrics; cached text lines 12–13, 47–48"},{"label":"Splits","value":"Five-fold cross-validation is performed within Liu’s training data; separately listed test subsets distinguish enhancer detection from enhancer-strength classification.","status":"source_checked","source_ids":["dnabert2-enhancer-2025"],"source_locator":"Methods: Benchmark dataset; Performance evaluation metrics; cached text lines 12–13, 47–48"},{"label":"Metrics","value":"Specificity, sensitivity, accuracy, balanced accuracy, MCC and AUC.","status":"source_checked","source_ids":["dnabert2-enhancer-2025"],"source_locator":"Methods: Benchmark dataset; Performance evaluation metrics; cached text lines 12–13, 47–48"},{"label":"Baselines","value":"Liu cross-validation compares iEnhancer-ECNN, BERT-Enhancer and iEnhancer-BERT. Its independent test additionally includes EnhancerPred, iEnhancer-EL, iEnhancer-XG and Enhancer-MDLF; not every comparator reports both recognition and strength metrics.","status":"source_checked","source_ids":["dnabert2-enhancer-2025"],"source_locator":"Results: Comparison with existing methods; Tables 4–5"},{"label":"Leakage controls","value":"The source describes CD-HIT filtering of highly similar sequences.","status":"source_checked","source_ids":["dnabert2-enhancer-2025"],"source_locator":"Methods: Benchmark dataset; Performance evaluation metrics; cached text lines 12–13, 47–48"},{"label":"Uncertainty","value":"The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","status":"unreported","source_ids":["dnabert2-enhancer-2025"],"source_locator":"Methods: Benchmark dataset; Performance evaluation metrics; cached text lines 12–13, 47–48"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["dnabert2-enhancer-2025"],"source_locator":"Methods: Benchmark dataset; Performance evaluation metrics; cached text lines 12–13, 47–48"},{"label":"Organisms","value":"The dataset section identifies Liu’s reused enhancer collection and eight Basith cell-line collections, but does not state a species or genome assembly for the Liu collection. These identities cannot be inferred from DNABERT2’s multispecies pretraining.","status":"unreported","source_ids":["dnabert2-enhancer-2025"],"source_locator":"Materials and methods: Benchmark dataset; Tables 1–2"},{"label":"Assays","value":"Enhancer/non-enhancer and strong/weak enhancer annotations.","status":"source_checked","source_ids":["dnabert2-enhancer-2025"],"source_locator":"Methods: Benchmark dataset; Performance evaluation metrics; cached text lines 12–13, 47–48"},{"label":"Allowed inputs","value":"DNA sequences.","status":"source_checked","source_ids":["dnabert2-enhancer-2025"],"source_locator":"Methods: Benchmark dataset; Performance evaluation metrics; cached text lines 12–13, 47–48"},{"label":"Adaptation","value":"DNABERT-2 converts sequences to feature matrices; a convolution/pooling module and two-layer perceptron perform supervised enhancer classification. The architecture passage does not specify the encoder freezing boundary.","status":"source_checked","source_ids":["dnabert2-enhancer-2025"],"source_locator":"Methods: DNABERT-2 representation and CNN model, cached paragraphs 36–43"}],"strengths":[],"limitations":[{"text":"Enhancer recognition and strong-versus-weak enhancer classification use different labels and test counts. Pretraining genome coverage does not establish independence of the downstream examples.","source_ids":["dnabert2-enhancer-2025"],"source_locator":"Methods: Benchmark dataset; Performance evaluation metrics; cached text lines 12–13, 47–48"}],"diagram":{"title":"Computational evaluation flow","steps":["Allowed inputs: DNA sequences.","Datasets: Previously published enhancer/non-enhancer sequences with strong/weak labels for the enhancer subset.","Metrics: Specificity, sensitivity, accuracy, balanced accuracy, MCC and AUC."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["dnabert2-enhancer-2025"],"source_locator":"Methods: Benchmark dataset; Performance evaluation metrics; cached text lines 12–13, 47–48; Methods: Benchmark dataset; Performance evaluation metrics; cached text lines 12–13, 47–48; Methods: Benchmark dataset; Performance evaluation metrics; cached text lines 12–13, 47–48"},"coverage":"limited","gaps":["Uncertainty: The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","Organisms: The dataset section identifies Liu’s reused enhancer collection and eight Basith cell-line collections, but does not state a species or genome assembly for the Liu collection. These identities cannot be inferred from DNABERT2’s multispecies pretraining."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Relevant full-paper computational evaluation sections, tables/captions and cited supplementary task passages were reviewed. Reporting omissions are scoped to the inspected sources. Original numerical results are unchanged."}}}}
{"id":"reported-task-92137759a9e7b0","kind":"benchmark","name":"Metagenomic taxonomic classification","description":"","status":"needs_review","facets":{"areas":["microbes-communities"],"tasks":["Metagenomic taxonomic classification"]},"source_ids":["nabas-plus-2025"],"links":[{"relation":"dataset","target_id":"reported-dataset-b462aa24561fba"}],"attributes":{"entity_level":"task","version":null,"task":"Metagenomic taxonomic classification","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Metagenomic classification compares reference-based tools and examines dependence on reference-genome quality.","summary_source_ids":["nabas-plus-2025"],"summary_source_locator":"Methods: reference classifiers; Discussion: CAMI evaluation; cached text lines 13–15, 108–109; task metric definitions and corresponding results table","sections":[{"title":"Evaluation methodology","body":"CAMI2 Toy Human Microbiome Project samples are among the benchmark collections. Precision, recall, F1, Jaccard distance and Bray–Curtis distance distinguish detection from abundance agreement. MetaPhlAn3, Kraken and GOTTCHA, using defaults where applicable. The source discusses reference assembly age and database coverage as factors affecting apparent accuracy. The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","source_ids":["nabas-plus-2025"],"source_locator":"Methods: reference classifiers; Discussion: CAMI evaluation; cached text lines 13–15, 108–109; task metric definitions and corresponding results table"}],"facts":[{"label":"Datasets","value":"CAMI2 Toy Human Microbiome Project samples are among the benchmark collections.","status":"source_checked","source_ids":["nabas-plus-2025"],"source_locator":"Methods: reference classifiers; Discussion: CAMI evaluation; cached text lines 13–15, 108–109; task metric definitions and corresponding results table"},{"label":"Splits","value":"This is evaluation of reference-database classifiers on CAMI2 samples and Zymo mock communities, rather than a supervised train/validation/test split. Reference-database versions are specified separately; the regenerated CAMI sample19 is a distinct condition.","status":"inapplicable","source_ids":["nabas-plus-2025"],"source_locator":"Methods: CAMI2 reference databases, Recreating CAMI sample19, Zymo community standards, Running classifiers"},{"label":"Metrics","value":"Precision, recall, F1, Jaccard distance and Bray–Curtis distance distinguish detection from abundance agreement.","status":"source_checked","source_ids":["nabas-plus-2025"],"source_locator":"Methods: reference classifiers; Discussion: CAMI evaluation; cached text lines 13–15, 108–109; task metric definitions and corresponding results table"},{"label":"Baselines","value":"MetaPhlAn3, Kraken and GOTTCHA, using defaults where applicable.","status":"source_checked","source_ids":["nabas-plus-2025"],"source_locator":"Methods: reference classifiers; Discussion: CAMI evaluation; cached text lines 13–15, 108–109; task metric definitions and corresponding results table"},{"label":"Leakage controls","value":"The source discusses reference assembly age and database coverage as factors affecting apparent accuracy.","status":"source_checked","source_ids":["nabas-plus-2025"],"source_locator":"Methods: reference classifiers; Discussion: CAMI evaluation; cached text lines 13–15, 108–109; task metric definitions and corresponding results table"},{"label":"Uncertainty","value":"The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","status":"unreported","source_ids":["nabas-plus-2025"],"source_locator":"Methods: reference classifiers; Discussion: CAMI evaluation; cached text lines 13–15, 108–109; task metric definitions and corresponding results table"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["nabas-plus-2025"],"source_locator":"Methods: reference classifiers; Discussion: CAMI evaluation; cached text lines 13–15, 108–109; task metric definitions and corresponding results table"},{"label":"Organisms","value":"CAMI2 human-microbiome community taxa.","status":"source_checked","source_ids":["nabas-plus-2025"],"source_locator":"Methods: reference classifiers; Discussion: CAMI evaluation; cached text lines 13–15, 108–109; task metric definitions and corresponding results table"},{"label":"Assays","value":"Metagenomic sequence mixtures with reference taxonomy.","status":"source_checked","source_ids":["nabas-plus-2025"],"source_locator":"Methods: reference classifiers; Discussion: CAMI evaluation; cached text lines 13–15, 108–109; task metric definitions and corresponding results table"},{"label":"Allowed inputs","value":"Sequencing reads and model-specific reference databases.","status":"source_checked","source_ids":["nabas-plus-2025"],"source_locator":"Methods: reference classifiers; Discussion: CAMI evaluation; cached text lines 13–15, 108–109; task metric definitions and corresponding results table"},{"label":"Adaptation","value":"Taxonomic classification with default comparator settings where applicable.","status":"source_checked","source_ids":["nabas-plus-2025"],"source_locator":"Methods: reference classifiers; Discussion: CAMI evaluation; cached text lines 13–15, 108–109; task metric definitions and corresponding results table"}],"strengths":[],"limitations":[{"text":"The regenerated synthetic sample using newer assemblies is a separate evaluation condition. Exact sample identity must remain attached to each reported result.","source_ids":["nabas-plus-2025"],"source_locator":"Methods: reference classifiers; Discussion: CAMI evaluation; cached text lines 13–15, 108–109; task metric definitions and corresponding results table"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Sequencing reads and model-specific reference databases.","Evaluation: Taxonomic classification with default comparator settings where applicable.","Readout: Precision, recall, F1, Jaccard distance and Bray–Curtis distance distinguish detection from abundance agreement."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["nabas-plus-2025"],"source_locator":"Methods: reference classifiers; Discussion: CAMI evaluation; cached text lines 13–15, 108–109; task metric definitions and corresponding results table"},"coverage":"limited","gaps":["Uncertainty: The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Relevant full-paper computational evaluation sections, tables/captions and cited supplementary task passages were reviewed. Reporting omissions are scoped to the inspected sources. Original numerical results are unchanged."}}}}
{"id":"reported-task-94802534b7026d","kind":"benchmark","name":"Protein–ligand binding energy prediction","description":"","status":"needs_review","facets":{"areas":["molecular-interactions"],"tasks":["Protein–ligand binding energy prediction"]},"source_ids":["fingerprint-scoring-2022"],"links":[{"relation":"dataset","target_id":"reported-dataset-16d01b5ef88e84"}],"attributes":{"entity_level":"task","version":null,"task":"Protein–ligand binding energy prediction","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Protein–ligand energy prediction compares alternative fingerprint descriptors and regression methods.","summary_source_ids":["fingerprint-scoring-2022"],"summary_source_locator":"Results and Discussion; Conclusion; cached text lines 44–46, 61","sections":[{"title":"Evaluation methodology","body":"Kd-labelled complexes from PDBbind v2019, with the corresponding CASF-2016 core structures reserved as the external test subset. Pearson correlation and root-mean-square error. LASSO and LightGBM with alternative protein/ligand fingerprint combinations.","source_ids":["fingerprint-scoring-2022"],"source_locator":"Computational Details, cached paragraph 30; Results and Discussion; Conclusion; cached text lines 44–46, 61; Results and Discussion; Conclusion; cached text lines 44–46, 61"}],"facts":[{"label":"Datasets","value":"Kd-labelled complexes from PDBbind v2019, with the corresponding CASF-2016 core structures reserved as the external test subset.","status":"source_checked","source_ids":["fingerprint-scoring-2022"],"source_locator":"Computational Details, cached paragraph 30"},{"label":"Splits","value":"Kd-labelled PDBbind v2019 complexes are divided into training, hyperparameter-selection (“pretest”) and an external CASF-2016 core test subset. Ten random training/pretest allocations are assessed, and the allocation with correlation closest to their average is retained; the external test set is not used for fitting.","status":"source_checked","source_ids":["fingerprint-scoring-2022"],"source_locator":"Computational Details; cached paragraphs 30–35"},{"label":"Metrics","value":"Pearson correlation and root-mean-square error.","status":"source_checked","source_ids":["fingerprint-scoring-2022"],"source_locator":"Results and Discussion; Conclusion; cached text lines 44–46, 61"},{"label":"Baselines","value":"LASSO and LightGBM with alternative protein/ligand fingerprint combinations.","status":"source_checked","source_ids":["fingerprint-scoring-2022"],"source_locator":"Results and Discussion; Conclusion; cached text lines 44–46, 61"},{"label":"Leakage controls","value":"The checked fingerprint-comparison passages do not define scaffold- or protein-target-disjoint evaluation.","status":"unreported","source_ids":["fingerprint-scoring-2022"],"source_locator":"Results and Discussion; Conclusion; cached text lines 44–46, 61"},{"label":"Uncertainty","value":"The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","status":"unreported","source_ids":["fingerprint-scoring-2022"],"source_locator":"Results and Discussion; Conclusion; cached text lines 44–46, 61"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["fingerprint-scoring-2022"],"source_locator":"Results and Discussion; Conclusion; cached text lines 44–46, 61"},{"label":"Organisms","value":"PDBbind v2019 complexes with Kd measurements are selected for binding-energy assessment. Computational Details does not tabulate their source organisms or a species-specific evaluation subset.","status":"unreported","source_ids":["fingerprint-scoring-2022"],"source_locator":"Computational Details, dataset selection and external CASF test"},{"label":"Assays","value":"Protein–ligand binding-energy labels.","status":"source_checked","source_ids":["fingerprint-scoring-2022"],"source_locator":"Results and Discussion; Conclusion; cached text lines 44–46, 61"},{"label":"Allowed inputs","value":"Protein/ligand fingerprint combinations.","status":"source_checked","source_ids":["fingerprint-scoring-2022"],"source_locator":"Results and Discussion; Conclusion; cached text lines 44–46, 61"},{"label":"Adaptation","value":"Supervised regression using LASSO or LightGBM.","status":"source_checked","source_ids":["fingerprint-scoring-2022"],"source_locator":"Results and Discussion; Conclusion; cached text lines 44–46, 61"}],"strengths":[],"limitations":[{"text":"The external CASF subset separates evaluation structures from fitted examples, but the described split is not a protein-family- or ligand-scaffold-disjoint test.","source_ids":["fingerprint-scoring-2022"],"source_locator":"Results and Discussion; Conclusion; cached text lines 44–46, 61"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Protein/ligand fingerprint combinations.","Evaluation: Supervised regression using LASSO or LightGBM.","Readout: Pearson correlation and root-mean-square error."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["fingerprint-scoring-2022"],"source_locator":"Results and Discussion; Conclusion; cached text lines 44–46, 61"},"coverage":"limited","gaps":["Leakage controls: The checked fingerprint-comparison passages do not define scaffold- or protein-target-disjoint evaluation.","Uncertainty: The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","Organisms: PDBbind v2019 complexes with Kd measurements are selected for binding-energy assessment. Computational Details does not tabulate their source organisms or a species-specific evaluation subset."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Relevant full-paper computational evaluation sections, tables/captions and cited supplementary task passages were reviewed. Reporting omissions are scoped to the inspected sources. Original numerical results are unchanged."}}}}
{"id":"reported-task-988ff78f86471e","kind":"benchmark","name":"Human 5mC detection","description":"","status":"needs_review","facets":{"areas":["dna-genomes"],"tasks":["Human 5mC detection"]},"source_ids":["dna-foundation-models-2025"],"links":[{"relation":"dataset","target_id":"reported-dataset-463197d6a98b99"}],"attributes":{"entity_level":"task","version":null,"task":"Human 5mC detection","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Human DNA methylation-site classification compares frozen foundation-model embeddings with supervised classifiers and a compact CNN baseline.","summary_source_ids":["dna-foundation-models-2025"],"summary_source_locator":"Methods: 5mC and 6mA modifications detection; Sequence classification benchmark; cached text lines 86, 110–114","sections":[{"title":"Evaluation methodology","body":"Human 5mC annotation windows from previously published benchmark datasets; positives and negatives derive from experimental profiling. The classification framework retains clearly specified source splits or applies randomized partitions; exact 5mC membership remains unextracted. AUC for binary tasks; pairwise DeLong tests assess AUC differences. Random forest, Naive Bayes and elastic-net classifiers over embeddings; a sequence CNN baseline. Similar sequences are filtered in the source dataset; complete pretraining exclusion is not established by this review. The paper uses significance testing for AUC differences; test-set uncertainty is distinct from variation across retraining.","source_ids":["dna-foundation-models-2025"],"source_locator":"Methods: 5mC and 6mA modifications detection; Sequence classification benchmark; cached text lines 86, 110–114"}],"facts":[{"label":"Datasets","value":"Human 5mC annotation windows from previously published benchmark datasets; positives and negatives derive from experimental profiling.","status":"source_checked","source_ids":["dna-foundation-models-2025"],"source_locator":"Methods: 5mC and 6mA modifications detection; Sequence classification benchmark; cached text lines 86, 110–114"},{"label":"Splits","value":"The classification framework retains clearly specified source splits or applies randomized partitions; exact 5mC membership remains unextracted.","status":"source_checked","source_ids":["dna-foundation-models-2025"],"source_locator":"Methods: 5mC and 6mA modifications detection; Sequence classification benchmark; cached text lines 86, 110–114"},{"label":"Metrics","value":"AUC for binary tasks; pairwise DeLong tests assess AUC differences.","status":"source_checked","source_ids":["dna-foundation-models-2025"],"source_locator":"Methods: 5mC and 6mA modifications detection; Sequence classification benchmark; cached text lines 86, 110–114"},{"label":"Baselines","value":"Random forest, Naive Bayes and elastic-net classifiers over embeddings; a sequence CNN baseline.","status":"source_checked","source_ids":["dna-foundation-models-2025"],"source_locator":"Methods: 5mC and 6mA modifications detection; Sequence classification benchmark; cached text lines 86, 110–114"},{"label":"Leakage controls","value":"Similar sequences are filtered in the source dataset; complete pretraining exclusion is not established by this review.","status":"source_checked","source_ids":["dna-foundation-models-2025"],"source_locator":"Methods: 5mC and 6mA modifications detection; Sequence classification benchmark; cached text lines 86, 110–114"},{"label":"Uncertainty","value":"The paper uses significance testing for AUC differences; test-set uncertainty is distinct from variation across retraining.","status":"source_checked","source_ids":["dna-foundation-models-2025"],"source_locator":"Methods: 5mC and 6mA modifications detection; Sequence classification benchmark; cached text lines 86, 110–114"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["dna-foundation-models-2025"],"source_locator":"Methods: 5mC and 6mA modifications detection; Sequence classification benchmark; cached text lines 86, 110–114"},{"label":"Organisms","value":"Human.","status":"source_checked","source_ids":["dna-foundation-models-2025"],"source_locator":"Methods: 5mC and 6mA modifications detection; Sequence classification benchmark; cached text lines 86, 110–114"},{"label":"Assays","value":"Experimental 5mC site annotations.","status":"source_checked","source_ids":["dna-foundation-models-2025"],"source_locator":"Methods: 5mC and 6mA modifications detection; Sequence classification benchmark; cached text lines 86, 110–114"},{"label":"Allowed inputs","value":"DNA windows represented by foundation-model embeddings.","status":"source_checked","source_ids":["dna-foundation-models-2025"],"source_locator":"Methods: 5mC and 6mA modifications detection; Sequence classification benchmark; cached text lines 86, 110–114"},{"label":"Adaptation","value":"Supervised conventional classifiers over embeddings; the sequence CNN is a separate baseline.","status":"source_checked","source_ids":["dna-foundation-models-2025"],"source_locator":"Methods: 5mC and 6mA modifications detection; Sequence classification benchmark; cached text lines 86, 110–114"}],"strengths":[],"limitations":[{"text":"The chromosome-held-out nested protocol elsewhere in the paper concerns variant-effect quantification and is not asserted for this methylation classification task.","source_ids":["dna-foundation-models-2025"],"source_locator":"Methods: 5mC and 6mA modifications detection; Sequence classification benchmark; cached text lines 86, 110–114"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: DNA windows represented by foundation-model embeddings.","Evaluation: Supervised conventional classifiers over embeddings; the sequence CNN is a separate baseline.","Readout: AUC for binary tasks; pairwise DeLong tests assess AUC differences."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["dna-foundation-models-2025"],"source_locator":"Methods: 5mC and 6mA modifications detection; Sequence classification benchmark; cached text lines 86, 110–114"},"coverage":"limited","gaps":["The chromosome-held-out nested protocol elsewhere in the paper concerns variant-effect quantification and is not asserted for this methylation classification task."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Task-specific computational methodology and field context checked in the cited primary-source artifact. Source-backed fields, inapplicable evaluator dimensions and unresolved details are distinguished. Numerical results were not reproduced."}}}}
{"id":"reported-task-9917a0e69f33e7","kind":"benchmark","name":"DNA-binding residue prediction","description":"","status":"needs_review","facets":{"areas":["proteins-complexes"],"tasks":["DNA-binding residue prediction"]},"source_ids":["megsite-2025"],"links":[{"relation":"dataset","target_id":"reported-dataset-739aee3cf8d6f1"}],"attributes":{"entity_level":"task","version":null,"task":"DNA-binding residue prediction","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"DNA-binding residue prediction uses published protein-sequence training and independent test collections.","summary_source_ids":["megsite-2025"],"summary_source_locator":"Methods: Benchmark datasets; Results; cached text lines 9–10, 63; matching task comparison table/ablation captions","sections":[{"title":"Evaluation methodology","body":"DNA-573_Train, DNA-129_Test and DNA-181_Test from prior binding-site benchmarks. Published training/test collections are maintained and similarity filtering is applied. AUC, average precision and MCC are central comparison metrics; specificity and precision are also reported. ESM2, ProtTrans, ProteinBERT and ESM3 embedding variants evaluated on the same named residue-label test collections. The source describes a 30% sequence-identity clustering cutoff between training and test sequences. The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","source_ids":["megsite-2025"],"source_locator":"Methods: Benchmark datasets; Results; cached text lines 9–10, 63; matching task comparison table/ablation captions"}],"facts":[{"label":"Datasets","value":"DNA-573_Train, DNA-129_Test and DNA-181_Test from prior binding-site benchmarks.","status":"source_checked","source_ids":["megsite-2025"],"source_locator":"Methods: Benchmark datasets; Results; cached text lines 9–10, 63; matching task comparison table/ablation captions"},{"label":"Splits","value":"Published training/test collections are maintained and similarity filtering is applied.","status":"source_checked","source_ids":["megsite-2025"],"source_locator":"Methods: Benchmark datasets; Results; cached text lines 9–10, 63; matching task comparison table/ablation captions"},{"label":"Metrics","value":"AUC, average precision and MCC are central comparison metrics; specificity and precision are also reported.","status":"source_checked","source_ids":["megsite-2025"],"source_locator":"Methods: Benchmark datasets; Results; cached text lines 9–10, 63; matching task comparison table/ablation captions"},{"label":"Baselines","value":"ESM2, ProtTrans, ProteinBERT and ESM3 embedding variants evaluated on the same named residue-label test collections.","status":"source_checked","source_ids":["megsite-2025"],"source_locator":"Methods: Benchmark datasets; Results; cached text lines 9–10, 63; matching task comparison table/ablation captions"},{"label":"Leakage controls","value":"The source describes a 30% sequence-identity clustering cutoff between training and test sequences.","status":"source_checked","source_ids":["megsite-2025"],"source_locator":"Methods: Benchmark datasets; Results; cached text lines 9–10, 63; matching task comparison table/ablation captions"},{"label":"Uncertainty","value":"The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","status":"unreported","source_ids":["megsite-2025"],"source_locator":"Methods: Benchmark datasets; Results; cached text lines 9–10, 63; matching task comparison table/ablation captions"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["megsite-2025"],"source_locator":"Methods: Benchmark datasets; Results; cached text lines 9–10, 63; matching task comparison table/ablation captions"},{"label":"Organisms","value":"The DNA-binding test uses the GraphBind/GraphSite DNA protein collections. The benchmark-data section and Table 1 describe proteins and residue labels without a species inventory; the separate RNA-binding collections do not define this task’s taxa.","status":"unreported","source_ids":["megsite-2025"],"source_locator":"Benchmark datasets and Table 1"},{"label":"Assays","value":"Curated DNA-binding residue labels.","status":"source_checked","source_ids":["megsite-2025"],"source_locator":"Methods: Benchmark datasets; Results; cached text lines 9–10, 63; matching task comparison table/ablation captions"},{"label":"Allowed inputs","value":"Protein representations for residue-level prediction.","status":"source_checked","source_ids":["megsite-2025"],"source_locator":"Methods: Benchmark datasets; Results; cached text lines 9–10, 63; matching task comparison table/ablation captions"},{"label":"Adaptation","value":"Supervised binding-site prediction on the published training set.","status":"source_checked","source_ids":["megsite-2025"],"source_locator":"Methods: Benchmark datasets; Results; cached text lines 9–10, 63; matching task comparison table/ablation captions"}],"strengths":[],"limitations":[{"text":"The paper separately evaluates RNA-binding sites; RNA structural split details must not be imported into this DNA-binding result.","source_ids":["megsite-2025"],"source_locator":"Methods: Benchmark datasets; Results; cached text lines 9–10, 63; matching task comparison table/ablation captions"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Protein representations for residue-level prediction.","Evaluation: Supervised binding-site prediction on the published training set.","Readout: AUC, average precision and MCC are central comparison metrics; specificity and precision are also reported."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["megsite-2025"],"source_locator":"Methods: Benchmark datasets; Results; cached text lines 9–10, 63; matching task comparison table/ablation captions"},"coverage":"limited","gaps":["Uncertainty: The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","Organisms: The DNA-binding test uses the GraphBind/GraphSite DNA protein collections. The benchmark-data section and Table 1 describe proteins and residue labels without a species inventory; the separate RNA-binding collections do not define this task’s taxa."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Relevant full-paper computational evaluation sections, tables/captions and cited supplementary task passages were reviewed. Reporting omissions are scoped to the inspected sources. Original numerical results are unchanged."}}}}
{"id":"reported-task-99afd88cb12895","kind":"benchmark","name":"gene-regulatory signal prediction","description":"","status":"needs_review","facets":{"areas":["cells-tissues"],"tasks":["gene-regulatory signal prediction"]},"source_ids":["single-cell-residual-geometry-2026"],"links":[{"relation":"dataset","target_id":"reported-dataset-d9fdd8dc7a0184"}],"attributes":{"entity_level":"task","version":null,"task":"gene-regulatory signal prediction","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Gene-regulatory signal prediction tests whether combined representations generalize beyond repeated gene identities.","summary_source_ids":["single-cell-residual-geometry-2026"],"summary_source_locator":"Results: nested cross-validation; Grouped cross-validation; cached text lines 48–56; Methods dataset and scGPT passages","sections":[{"title":"Evaluation methodology","body":"Regulatory-edge evaluation across tissue domains; exact dataset releases remain unextracted. Nested outer-split evaluation is complemented by leave-TF-out, leave-target-out and leave-both-out tests. AUROC in the reviewed stacked-model comparison. Combined representation is compared with the individual model branches. Nested validation separates feature construction from testing; entity-grouped tests address gene-identity carryover. Bootstrap confidence intervals support reported comparisons.","source_ids":["single-cell-residual-geometry-2026"],"source_locator":"Results: nested cross-validation; Grouped cross-validation; cached text lines 48–56; Methods dataset and scGPT passages"}],"facts":[{"label":"Datasets","value":"Regulatory-edge evaluation across tissue domains; exact dataset releases remain unextracted.","status":"source_checked","source_ids":["single-cell-residual-geometry-2026"],"source_locator":"Results: nested cross-validation; Grouped cross-validation; cached text lines 48–56; Methods dataset and scGPT passages"},{"label":"Splits","value":"Nested outer-split evaluation is complemented by leave-TF-out, leave-target-out and leave-both-out tests.","status":"source_checked","source_ids":["single-cell-residual-geometry-2026"],"source_locator":"Results: nested cross-validation; Grouped cross-validation; cached text lines 48–56; Methods dataset and scGPT passages"},{"label":"Metrics","value":"AUROC in the reviewed stacked-model comparison.","status":"source_checked","source_ids":["single-cell-residual-geometry-2026"],"source_locator":"Results: nested cross-validation; Grouped cross-validation; cached text lines 48–56; Methods dataset and scGPT passages"},{"label":"Baselines","value":"Combined representation is compared with the individual model branches.","status":"source_checked","source_ids":["single-cell-residual-geometry-2026"],"source_locator":"Results: nested cross-validation; Grouped cross-validation; cached text lines 48–56; Methods dataset and scGPT passages"},{"label":"Leakage controls","value":"Nested validation separates feature construction from testing; entity-grouped tests address gene-identity carryover.","status":"source_checked","source_ids":["single-cell-residual-geometry-2026"],"source_locator":"Results: nested cross-validation; Grouped cross-validation; cached text lines 48–56; Methods dataset and scGPT passages"},{"label":"Uncertainty","value":"Bootstrap confidence intervals support reported comparisons.","status":"source_checked","source_ids":["single-cell-residual-geometry-2026"],"source_locator":"Results: nested cross-validation; Grouped cross-validation; cached text lines 48–56; Methods dataset and scGPT passages"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["single-cell-residual-geometry-2026"],"source_locator":"Results: nested cross-validation; Grouped cross-validation; cached text lines 48–56; Methods dataset and scGPT passages"},{"label":"Organisms","value":"Human tissues from the Tabula Sapiens atlas.","status":"source_checked","source_ids":["single-cell-residual-geometry-2026"],"source_locator":"Results: nested cross-validation; Grouped cross-validation; cached text lines 48–56; Methods dataset and scGPT passages"},{"label":"Assays","value":"Tabula Sapiens single-cell RNA-seq with TRRUST regulatory reference annotations.","status":"source_checked","source_ids":["single-cell-residual-geometry-2026"],"source_locator":"Results: nested cross-validation; Grouped cross-validation; cached text lines 48–56; Methods dataset and scGPT passages"},{"label":"Allowed inputs","value":"Candidate transcription-factor/target pairs and model representations.","status":"source_checked","source_ids":["single-cell-residual-geometry-2026"],"source_locator":"Results: nested cross-validation; Grouped cross-validation; cached text lines 48–56; Methods dataset and scGPT passages"},{"label":"Adaptation","value":"Supervised prediction with nested splits and TF/target holdout stress tests.","status":"source_checked","source_ids":["single-cell-residual-geometry-2026"],"source_locator":"Results: nested cross-validation; Grouped cross-validation; cached text lines 48–56; Methods dataset and scGPT passages"}],"strengths":[{"text":"Separate TF, target and joint holdouts diagnose which entity-sharing shortcuts remain possible.","source_ids":["single-cell-residual-geometry-2026"],"source_locator":"Results: nested cross-validation; Grouped cross-validation; cached text lines 48–56; Methods dataset and scGPT passages"}],"limitations":[{"text":"Random edge splits and entity-held-out splits answer different questions. Exact bootstrap unit and reference-edge uncertainty remain unextracted.","source_ids":["single-cell-residual-geometry-2026"],"source_locator":"Results: nested cross-validation; Grouped cross-validation; cached text lines 48–56; Methods dataset and scGPT passages"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Candidate transcription-factor/target pairs and model representations.","Evaluation: Supervised prediction with nested splits and TF/target holdout stress tests.","Readout: AUROC in the reviewed stacked-model comparison."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["single-cell-residual-geometry-2026"],"source_locator":"Results: nested cross-validation; Grouped cross-validation; cached text lines 48–56; Methods dataset and scGPT passages"},"coverage":"limited","gaps":["Random edge splits and entity-held-out splits answer different questions. Exact bootstrap unit and reference-edge uncertainty remain unextracted."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Task-specific computational methodology and field context checked in the cited primary-source artifact. Source-backed fields, inapplicable evaluator dimensions and unresolved details are distinguished. Numerical results were not reproduced."}}}}
{"id":"reported-task-9f62e739c6371e","kind":"benchmark","name":"human core-promoter classification","description":"","status":"needs_review","facets":{"areas":["dna-genomes"],"tasks":["human core-promoter classification"]},"source_ids":["eden-genomic-classification-2026"],"links":[{"relation":"dataset","target_id":"reported-dataset-8e9488896becd4"}],"attributes":{"entity_level":"task","version":null,"task":"human core-promoter classification","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Human core-promoter classification is evaluated as a constituent task of the GUE sequence-classification benchmark.","summary_source_ids":["eden-genomic-classification-2026"],"summary_source_locator":"Methods: Datasets; cached text lines 119–120; uncertainty/repeat-run/statistical-comparison passages; matching task comparison table/ablation captions","sections":[{"title":"Evaluation methodology","body":"GUE human core-promoter data; the paper evaluates a broader collection of human and mouse classification datasets. Predefined training, validation and test subsets supplied with GUE. Matthews correlation coefficient is the primary benchmark metric. OneHot with the high-frequency branch and K-merFreq with XGBoost are controlled in-house baselines; other methods include literature-reported results. Mean and standard deviation across five independent evaluations are reported for EDEN and the OneHot/HFBranch baseline; imported results keep their original reporting conventions.","source_ids":["eden-genomic-classification-2026"],"source_locator":"Methods: Datasets; cached text lines 119–120; uncertainty/repeat-run/statistical-comparison passages; matching task comparison table/ablation captions"}],"facts":[{"label":"Datasets","value":"GUE human core-promoter data; the paper evaluates a broader collection of human and mouse classification datasets.","status":"source_checked","source_ids":["eden-genomic-classification-2026"],"source_locator":"Methods: Datasets; cached text lines 119–120; uncertainty/repeat-run/statistical-comparison passages; matching task comparison table/ablation captions"},{"label":"Splits","value":"Predefined training, validation and test subsets supplied with GUE.","status":"source_checked","source_ids":["eden-genomic-classification-2026"],"source_locator":"Methods: Datasets; cached text lines 119–120; uncertainty/repeat-run/statistical-comparison passages; matching task comparison table/ablation captions"},{"label":"Metrics","value":"Matthews correlation coefficient is the primary benchmark metric.","status":"source_checked","source_ids":["eden-genomic-classification-2026"],"source_locator":"Methods: Datasets; cached text lines 119–120; uncertainty/repeat-run/statistical-comparison passages; matching task comparison table/ablation captions"},{"label":"Baselines","value":"OneHot with the high-frequency branch and K-merFreq with XGBoost are controlled in-house baselines; other methods include literature-reported results.","status":"source_checked","source_ids":["eden-genomic-classification-2026"],"source_locator":"Methods: Datasets; cached text lines 119–120; uncertainty/repeat-run/statistical-comparison passages; matching task comparison table/ablation captions"},{"label":"Leakage controls","value":"The evaluation uses predefined training, validation and test subsets of the reused genomic datasets. The inspected dataset and implementation sections do not document an additional chromosome-, locus- or sequence-similarity exclusion audit, or an audit of foundation-model pretraining overlap.","status":"unreported","source_ids":["eden-genomic-classification-2026"],"source_locator":"Experimental settings: datasets and implementation; cached paragraphs 116–137"},{"label":"Uncertainty","value":"Mean and standard deviation across five independent evaluations are reported for EDEN and the OneHot/HFBranch baseline; imported results keep their original reporting conventions.","status":"source_checked","source_ids":["eden-genomic-classification-2026"],"source_locator":"Methods: Datasets; cached text lines 119–120; uncertainty/repeat-run/statistical-comparison passages; matching task comparison table/ablation captions"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["eden-genomic-classification-2026"],"source_locator":"Methods: Datasets; cached text lines 119–120; uncertainty/repeat-run/statistical-comparison passages; matching task comparison table/ablation captions"},{"label":"Organisms","value":"Human for the core-promoter task; the broader suite also contains mouse datasets.","status":"source_checked","source_ids":["eden-genomic-classification-2026"],"source_locator":"Methods: Datasets; cached text lines 119–120; uncertainty/repeat-run/statistical-comparison passages; matching task comparison table/ablation captions"},{"label":"Assays","value":"GUE promoter/non-promoter annotations.","status":"source_checked","source_ids":["eden-genomic-classification-2026"],"source_locator":"Methods: Datasets; cached text lines 119–120; uncertainty/repeat-run/statistical-comparison passages; matching task comparison table/ablation captions"},{"label":"Allowed inputs","value":"DNA sequences.","status":"source_checked","source_ids":["eden-genomic-classification-2026"],"source_locator":"Methods: Datasets; cached text lines 119–120; uncertainty/repeat-run/statistical-comparison passages; matching task comparison table/ablation captions"},{"label":"Adaptation","value":"Supervised classification using the predefined GUE train/validation/test data.","status":"source_checked","source_ids":["eden-genomic-classification-2026"],"source_locator":"Methods: Datasets; cached text lines 119–120; uncertainty/repeat-run/statistical-comparison passages; matching task comparison table/ablation captions"}],"strengths":[],"limitations":[{"text":"Task-level scores are tied to the reused dataset partitions. Predefined partitions alone do not establish absence of genomic or pretraining overlap.","source_ids":["eden-genomic-classification-2026"],"source_locator":"Methods: Datasets; cached text lines 119–120; uncertainty/repeat-run/statistical-comparison passages; matching task comparison table/ablation captions"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: DNA sequences.","Evaluation: Supervised classification using the predefined GUE train/validation/test data.","Readout: Matthews correlation coefficient is the primary benchmark metric."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["eden-genomic-classification-2026"],"source_locator":"Methods: Datasets; cached text lines 119–120; uncertainty/repeat-run/statistical-comparison passages; matching task comparison table/ablation captions"},"coverage":"limited","gaps":["Leakage controls: The evaluation uses predefined training, validation and test subsets of the reused genomic datasets. The inspected dataset and implementation sections do not document an additional chromosome-, locus- or sequence-similarity exclusion audit, or an audit of foundation-model pretraining overlap."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Relevant full-paper computational evaluation sections, tables/captions and cited supplementary task passages were reviewed. Reporting omissions are scoped to the inspected sources. Original numerical results are unchanged."}}}}
{"id":"reported-task-9f9ab0090f6522","kind":"benchmark","name":"Natural vs artificial microbial genome sequence","description":"","status":"needs_review","facets":{"areas":["microbes-communities"],"tasks":["Natural vs artificial microbial genome sequence"]},"source_ids":["genomeocean-2025"],"links":[{"relation":"dataset","target_id":"reported-dataset-0c3ac7efe99c37"}],"attributes":{"entity_level":"task","version":null,"task":"Natural vs artificial microbial genome sequence","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"This GenomeOcean task distinguishes natural microbial sequence fragments from model-generated fragments. It measures discrimination under the study’s dataset construction.","summary_source_ids":["genomeocean-2025"],"summary_source_locator":"Methods 4.2.5 Generated Sequence Discrimination; Table 2","sections":[{"title":"What the evaluation establishes","body":"The study uses CAMI2-derived examples for training and validation and GTDB-derived examples for testing. Natural and generated examples form separate classes. This source separation defines the assessment; a discrimination score is not experimental evidence that a generated sequence has biological function.","source_ids":["genomeocean-2025"],"source_locator":"Methods 4.2.5 Generated Sequence Discrimination; Table 2"}],"facts":[{"label":"Entity type","value":"Paper-specific evaluation task; this profile is a descriptive evidence summary.","status":"source_checked","source_ids":["genomeocean-2025"],"source_locator":"Methods 4.2.5 Generated Sequence Discrimination; Table 2"},{"label":"Datasets","value":"Natural and generated sequence-fragment collections derived from CAMI2 and GTDB references.","status":"source_checked","source_ids":["genomeocean-2025"],"source_locator":"Methods 4.2.5 Generated Sequence Discrimination; Table 2"},{"label":"Organisms","value":"Microbial reference sequence collections; organism membership follows the named source datasets.","status":"source_checked","source_ids":["genomeocean-2025"],"source_locator":"Methods 4.2.5 Generated Sequence Discrimination; Table 2"},{"label":"Assays","value":"Computational class labels for sequence origin, without a functional measurement in this particular endpoint.","status":"source_checked","source_ids":["genomeocean-2025"],"source_locator":"Methods 4.2.5 Generated Sequence Discrimination; Table 2"},{"label":"Splits","value":"CAMI2-derived training/validation and GTDB-derived test collections are kept distinct. Exact membership and reference releases remain necessary for reproducibility.","status":"source_checked","source_ids":["genomeocean-2025"],"source_locator":"Methods 4.2.5 Generated Sequence Discrimination; Table 2"},{"label":"Allowed inputs","value":"DNA sequence representations and the natural/generated class label.","status":"source_checked","source_ids":["genomeocean-2025"],"source_locator":"Methods 4.2.5 Generated Sequence Discrimination; Table 2"},{"label":"Adaptation","value":"The retained comparison assesses representations through the paper’s classification task; generation and classification are distinct stages.","status":"source_checked","source_ids":["genomeocean-2025"],"source_locator":"Methods 4.2.5 Generated Sequence Discrimination; Table 2"},{"label":"Metrics","value":"Table 2 reports classification measures including F1; each preserved result keeps its original metric and unit.","status":"source_checked","source_ids":["genomeocean-2025"],"source_locator":"Methods 4.2.5 Generated Sequence Discrimination; Table 2"},{"label":"Baselines","value":"Table 2 includes GenomeOcean and DNABERT-2 representation-based comparators. Their classification results do not by themselves assess sequence function.","status":"source_checked","source_ids":["genomeocean-2025"],"source_locator":"Methods 4.2.5 Generated Sequence Discrimination; Table 2"}],"strengths":[{"text":"Using different reference collections for fitting and testing makes the intended dataset transfer explicit.","source_ids":["genomeocean-2025"],"source_locator":"Methods 4.2.5 Generated Sequence Discrimination; Table 2"}],"limitations":[{"text":"A discriminator can learn properties of collection or generation provenance. Its score does not validate a generated genome, pathway or phenotype.","source_ids":["genomeocean-2025"],"source_locator":"Methods 4.2.5 Generated Sequence Discrimination; Table 2"}],"diagram":{"title":"Conceptual assessment outline","steps":["Identify reference-derived classes","Separate fitting and test collections","Assess sequence discrimination","Report the specified metric"],"caption":"Conceptual overview of the published statistical assessment.","source_ids":["genomeocean-2025"],"source_locator":"Methods 4.2.5 Generated Sequence Discrimination; Table 2"},"coverage":"limited","gaps":["This profile reviews the evaluation question, data type and statistical readout. It does not reconstruct an executable protocol, verify exact split membership or independently reproduce the experiment."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Inspected the cited primary abstract and descriptive computational-evaluation sections. Review covers the descriptive claims shown; no executable protocol was reconstructed. Original numerical records retain their prior transcription review."}}}}
{"id":"reported-task-a1151e386a3d3f","kind":"benchmark","name":"Multi-species prokaryotic promoter detection","description":"","status":"needs_review","facets":{"areas":["microbes-communities"],"tasks":["Multi-species prokaryotic promoter detection"]},"source_ids":["ipromp-2025"],"links":[{"relation":"dataset","target_id":"reported-dataset-e0f34dcaa1ba3b"}],"attributes":{"entity_level":"task","version":null,"task":"Multi-species prokaryotic promoter detection","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Multi-species promoter classification uses curated positives and genomic-background negatives.","summary_source_ids":["ipromp-2025"],"summary_source_locator":"Methods: Data collection and preprocessing; Performance evaluation; cached text lines 54–56, 69–72","sections":[{"title":"Evaluation methodology","body":"Promoter annotations from PPD and species-matched negative sequence collections. Five-fold cross-validation is used for fitting and an independent collection for comparison. Sensitivity, specificity, accuracy, MCC and ROC-AUC. The source describes CD-HIT redundancy filtering for both classes. The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","source_ids":["ipromp-2025"],"source_locator":"Methods: Data collection and preprocessing; Performance evaluation; cached text lines 54–56, 69–72"}],"facts":[{"label":"Datasets","value":"Promoter annotations from PPD and species-matched negative sequence collections.","status":"source_checked","source_ids":["ipromp-2025"],"source_locator":"Methods: Data collection and preprocessing; Performance evaluation; cached text lines 54–56, 69–72"},{"label":"Splits","value":"Five-fold cross-validation is used for fitting and an independent collection for comparison.","status":"source_checked","source_ids":["ipromp-2025"],"source_locator":"Methods: Data collection and preprocessing; Performance evaluation; cached text lines 54–56, 69–72"},{"label":"Metrics","value":"Sensitivity, specificity, accuracy, MCC and ROC-AUC.","status":"source_checked","source_ids":["ipromp-2025"],"source_locator":"Methods: Data collection and preprocessing; Performance evaluation; cached text lines 54–56, 69–72"},{"label":"Baselines","value":"Generic classifiers are random forest, XGBoost, logistic regression and LSTM, trained with five-fold cross-validation. Tool comparisons use Prompt, PromoterLCNN and iPro-WAEL on the same species-specific independent tests; a separate repeat uses their official five-fold training protocols.","status":"source_checked","source_ids":["ipromp-2025"],"source_locator":"Results: Tables 1–2; comparison with baseline classifiers and existing tools; cached paragraphs 29–37"},{"label":"Leakage controls","value":"The source describes CD-HIT redundancy filtering for both classes.","status":"source_checked","source_ids":["ipromp-2025"],"source_locator":"Methods: Data collection and preprocessing; Performance evaluation; cached text lines 54–56, 69–72"},{"label":"Uncertainty","value":"The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","status":"unreported","source_ids":["ipromp-2025"],"source_locator":"Methods: Data collection and preprocessing; Performance evaluation; cached text lines 54–56, 69–72"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["ipromp-2025"],"source_locator":"Methods: Data collection and preprocessing; Performance evaluation; cached text lines 54–56, 69–72"},{"label":"Organisms","value":"Multiple prokaryotic species in PPD.","status":"source_checked","source_ids":["ipromp-2025"],"source_locator":"Methods: Data collection and preprocessing; Performance evaluation; cached text lines 54–56, 69–72"},{"label":"Assays","value":"Promoter annotations and species-matched negatives.","status":"source_checked","source_ids":["ipromp-2025"],"source_locator":"Methods: Data collection and preprocessing; Performance evaluation; cached text lines 54–56, 69–72"},{"label":"Allowed inputs","value":"DNA sequence windows.","status":"source_checked","source_ids":["ipromp-2025"],"source_locator":"Methods: Data collection and preprocessing; Performance evaluation; cached text lines 54–56, 69–72"},{"label":"Adaptation","value":"Supervised promoter classification with cross-validation and an independent collection.","status":"source_checked","source_ids":["ipromp-2025"],"source_locator":"Methods: Data collection and preprocessing; Performance evaluation; cached text lines 54–56, 69–72"}],"strengths":[],"limitations":[{"text":"Within-species independent tests and cross-species transfer matrices are separate evaluation conditions; their aggregate scores must not be interchanged.","source_ids":["ipromp-2025"],"source_locator":"Methods: Data collection and preprocessing; Performance evaluation; cached text lines 54–56, 69–72"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: DNA sequence windows.","Evaluation: Supervised promoter classification with cross-validation and an independent collection.","Readout: Sensitivity, specificity, accuracy, MCC and ROC-AUC."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["ipromp-2025"],"source_locator":"Methods: Data collection and preprocessing; Performance evaluation; cached text lines 54–56, 69–72"},"coverage":"limited","gaps":["Uncertainty: The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Relevant full-paper computational evaluation sections, tables/captions and cited supplementary task passages were reviewed. Reporting omissions are scoped to the inspected sources. Original numerical results are unchanged."}}}}
{"id":"reported-task-a2bf7ddbc71d23","kind":"benchmark","name":"RNA secondary-structure prediction","description":"","status":"needs_review","facets":{"areas":["rna-transcriptomes"],"tasks":["RNA secondary-structure prediction"]},"source_ids":["ernie-rna-2025"],"links":[{"relation":"dataset","target_id":"reported-dataset-abdfba8cce7486"}],"attributes":{"entity_level":"task","version":null,"task":"RNA secondary-structure prediction","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"RNA secondary-structure prediction uses a specified bpRNA benchmark version while distinguishing pretraining from supervised task data.","summary_source_ids":["ernie-rna-2025"],"summary_source_locator":"Methods: Training dataset; RNA secondary structure dataset; cached text lines 71–72, 84–87; task metric definitions and corresponding results table","sections":[{"title":"Evaluation methodology","body":"bpRNA-1m similarity-filtered variants; the main task uses the standard TR0/VL0/TS0 benchmark. Base-pairing F1 is used for secondary-structure model selection and evaluation; contact-map post-processing is shared across methods. The partition is aligned with UFold and RNA-FM comparisons; UNI-RNA uses additional task-training data.","source_ids":["ernie-rna-2025"],"source_locator":"Methods: Training dataset; RNA secondary structure dataset; cached text lines 71–72, 84–87; task metric definitions and corresponding results table; Methods: Training dataset; RNA secondary structure dataset; cached text lines 71–72, 84–87; task metric definitions and corresponding results table; Methods: Training dataset; RNA secondary structure dataset; cached text lines 71–72, 84–87; task metric definitions and corresponding results table"}],"facts":[{"label":"Datasets","value":"bpRNA-1m similarity-filtered variants; the main task uses the standard TR0/VL0/TS0 benchmark.","status":"source_checked","source_ids":["ernie-rna-2025"],"source_locator":"Methods: Training dataset; RNA secondary structure dataset; cached text lines 71–72, 84–87; task metric definitions and corresponding results table"},{"label":"Splits","value":"TR0 is used for task training, VL0 for validation and TS0 for testing.","status":"source_checked","source_ids":["ernie-rna-2025"],"source_locator":"Methods: Training dataset; RNA secondary structure dataset; cached text lines 71–72, 84–87; task metric definitions and corresponding results table"},{"label":"Metrics","value":"Base-pairing F1 is used for secondary-structure model selection and evaluation; contact-map post-processing is shared across methods.","status":"source_checked","source_ids":["ernie-rna-2025"],"source_locator":"Methods: Training dataset; RNA secondary structure dataset; cached text lines 71–72, 84–87; task metric definitions and corresponding results table"},{"label":"Baselines","value":"The partition is aligned with UFold and RNA-FM comparisons; UNI-RNA uses additional task-training data.","status":"source_checked","source_ids":["ernie-rna-2025"],"source_locator":"Methods: Training dataset; RNA secondary structure dataset; cached text lines 71–72, 84–87; task metric definitions and corresponding results table"},{"label":"Leakage controls","value":"The bpRNA benchmark variants apply explicit sequence-similarity filtering, including the 80% variant used for TR0/VL0/TS0. The cited dataset Methods do not give a cross-corpus exclusion audit against RNAcentral pretraining.","status":"source_checked","source_ids":["ernie-rna-2025"],"source_locator":"Methods: Training dataset and RNA secondary structure dataset; cached paragraphs 71–72,84–87"},{"label":"Uncertainty","value":"The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","status":"unreported","source_ids":["ernie-rna-2025"],"source_locator":"Methods: Training dataset; RNA secondary structure dataset; cached text lines 71–72, 84–87; task metric definitions and corresponding results table"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["ernie-rna-2025"],"source_locator":"Methods: Training dataset; RNA secondary structure dataset; cached text lines 71–72, 84–87; task metric definitions and corresponding results table"},{"label":"Organisms","value":"The secondary-structure task is defined by bpRNA sequences and RNA-family collections. Its downstream-dataset Methods describe similarity filtering and partition membership, but do not tabulate the evaluated sequences by organism.","status":"unreported","source_ids":["ernie-rna-2025"],"source_locator":"Methods: RNA secondary structure datasets; bpRNA-1m and ArchiveII"},{"label":"Assays","value":"Reference RNA secondary structures.","status":"source_checked","source_ids":["ernie-rna-2025"],"source_locator":"Methods: Training dataset; RNA secondary structure dataset; cached text lines 71–72, 84–87; task metric definitions and corresponding results table"},{"label":"Allowed inputs","value":"RNA sequence.","status":"source_checked","source_ids":["ernie-rna-2025"],"source_locator":"Methods: Training dataset; RNA secondary structure dataset; cached text lines 71–72, 84–87; task metric definitions and corresponding results table"},{"label":"Adaptation","value":"Supervised structure-task training on TR0, validation on VL0 and testing on TS0.","status":"source_checked","source_ids":["ernie-rna-2025"],"source_locator":"Methods: Training dataset; RNA secondary structure dataset; cached text lines 71–72, 84–87; task metric definitions and corresponding results table"}],"strengths":[],"limitations":[{"text":"The different bpRNA similarity-filtered variants have different sequence memberships. Reporting a dataset family name alone is insufficient to identify the evaluated split.","source_ids":["ernie-rna-2025"],"source_locator":"Methods: Training dataset; RNA secondary structure dataset; cached text lines 71–72, 84–87; task metric definitions and corresponding results table"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: RNA sequence.","Evaluation: Supervised structure-task training on TR0, validation on VL0 and testing on TS0.","Readout: Base-pairing F1 is used for secondary-structure model selection and evaluation; contact-map post-processing is shared across methods."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["ernie-rna-2025"],"source_locator":"Methods: Training dataset; RNA secondary structure dataset; cached text lines 71–72, 84–87; task metric definitions and corresponding results table"},"coverage":"limited","gaps":["Uncertainty: The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","Organisms: The secondary-structure task is defined by bpRNA sequences and RNA-family collections. Its downstream-dataset Methods describe similarity filtering and partition membership, but do not tabulate the evaluated sequences by organism."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Relevant full-paper computational evaluation sections, tables/captions and cited supplementary task passages were reviewed. Reporting omissions are scoped to the inspected sources. Original numerical results are unchanged."}}}}
{"id":"reported-task-a2c37b8c420bc3","kind":"benchmark","name":"CAMI II superkingdom read classification","description":"","status":"needs_review","facets":{"areas":["microbes-communities"],"tasks":["CAMI II superkingdom read classification"]},"source_ids":["ncd-metagenomics-2026"],"links":[{"relation":"dataset","target_id":"reported-dataset-beb4f5da29da0a"}],"attributes":{"entity_level":"task","version":null,"task":"CAMI II superkingdom read classification","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"CAMI II read classification evaluates a computationally limited subsample, with taxonomic-rank-specific interpretation.","summary_source_ids":["ncd-metagenomics-2026"],"summary_source_locator":"Methods: Evaluation protocol; Results: CAMI II; cached text lines 67–68, 113–116","sections":[{"title":"Evaluation methodology","body":"A 10,000-read subset of CAMI II Sample_0 from the human-microbiome collection. Queries are classified against the paper’s separately assembled metagenomic reference/training data. Macro-F1 is the arithmetic mean of classwise F1; only classes present in the evaluated subset enter the macro calculation. NCD-gzip and Kraken2.","source_ids":["ncd-metagenomics-2026"],"source_locator":"Methods: Evaluation protocol; Results: CAMI II; cached text lines 67–68, 113–116"}],"facts":[{"label":"Datasets","value":"A 10,000-read subset of CAMI II Sample_0 from the human-microbiome collection.","status":"source_checked","source_ids":["ncd-metagenomics-2026"],"source_locator":"Methods: Evaluation protocol; Results: CAMI II; cached text lines 67–68, 113–116"},{"label":"Splits","value":"Queries are classified against the paper’s separately assembled metagenomic reference/training data.","status":"source_checked","source_ids":["ncd-metagenomics-2026"],"source_locator":"Methods: Evaluation protocol; Results: CAMI II; cached text lines 67–68, 113–116"},{"label":"Metrics","value":"Macro-F1 is the arithmetic mean of classwise F1; only classes present in the evaluated subset enter the macro calculation.","status":"source_checked","source_ids":["ncd-metagenomics-2026"],"source_locator":"Methods: Evaluation protocol; Results: CAMI II; cached text lines 67–68, 113–116"},{"label":"Baselines","value":"NCD-gzip and Kraken2.","status":"source_checked","source_ids":["ncd-metagenomics-2026"],"source_locator":"Methods: Evaluation protocol; Results: CAMI II; cached text lines 67–68, 113–116"},{"label":"Leakage controls","value":"The CAMI experiment classifies a single Sample_0 read subsample with the separately assembled metagenomic reference collection. The CAMI section does not report removal of CAMI source genomes or close relatives from that reference. Gene-out and taxon-out partitions elsewhere in the paper belong to different evaluations.","status":"unreported","source_ids":["ncd-metagenomics-2026"],"source_locator":"Methods: metagenomic reference data; Results: CAMI dataset and Table 5; cached paragraphs 49–58,112–115"},{"label":"Uncertainty","value":"Table 5 reports point metrics for one 10,000-read CAMI subsample. The CAMI results section supplies no repeated-subsampling variability or confidence intervals; five-run cross-validation in Tables 3–4 concerns the separate human-gene classification experiment.","status":"unreported","source_ids":["ncd-metagenomics-2026"],"source_locator":"Results: Human DNA classification, CAMI dataset; Tables 3–5"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["ncd-metagenomics-2026"],"source_locator":"Methods: Evaluation protocol; Results: CAMI II; cached text lines 67–68, 113–116"},{"label":"Organisms","value":"CAMI II human-microbiome community taxa.","status":"source_checked","source_ids":["ncd-metagenomics-2026"],"source_locator":"Methods: Evaluation protocol; Results: CAMI II; cached text lines 67–68, 113–116"},{"label":"Assays","value":"Read-origin taxonomy at the selected evaluation rank.","status":"source_checked","source_ids":["ncd-metagenomics-2026"],"source_locator":"Methods: Evaluation protocol; Results: CAMI II; cached text lines 67–68, 113–116"},{"label":"Allowed inputs","value":"DNA reads and a separately assembled reference/training collection.","status":"source_checked","source_ids":["ncd-metagenomics-2026"],"source_locator":"Methods: Evaluation protocol; Results: CAMI II; cached text lines 67–68, 113–116"},{"label":"Adaptation","value":"Reference-based classification using NCD-gzip or Kraken2.","status":"source_checked","source_ids":["ncd-metagenomics-2026"],"source_locator":"Methods: Evaluation protocol; Results: CAMI II; cached text lines 67–68, 113–116"}],"strengths":[],"limitations":[{"text":"The subsample contains only Bacteria at superkingdom level, so that result does not establish broad multiclass discrimination. Phylum and superkingdom are separate catalogue outcomes.","source_ids":["ncd-metagenomics-2026"],"source_locator":"Methods: Evaluation protocol; Results: CAMI II; cached text lines 67–68, 113–116"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: DNA reads and a separately assembled reference/training collection.","Evaluation: Reference-based classification using NCD-gzip or Kraken2.","Readout: Macro-F1 is the arithmetic mean of classwise F1; only classes present in the evaluated subset enter the macro calculation."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["ncd-metagenomics-2026"],"source_locator":"Methods: Evaluation protocol; Results: CAMI II; cached text lines 67–68, 113–116"},"coverage":"limited","gaps":["Leakage controls: The CAMI experiment classifies a single Sample_0 read subsample with the separately assembled metagenomic reference collection. The CAMI section does not report removal of CAMI source genomes or close relatives from that reference. Gene-out and taxon-out partitions elsewhere in the paper belong to different evaluations.","Uncertainty: Table 5 reports point metrics for one 10,000-read CAMI subsample. The CAMI results section supplies no repeated-subsampling variability or confidence intervals; five-run cross-validation in Tables 3–4 concerns the separate human-gene classification experiment."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Relevant full-paper computational evaluation sections, tables/captions and cited supplementary task passages were reviewed. Reporting omissions are scoped to the inspected sources. Original numerical results are unchanged."}}}}
{"id":"reported-task-a5141363b0ee45","kind":"benchmark","name":"Zero-shot variant effect prediction","description":"","status":"needs_review","facets":{"areas":["proteins-complexes"],"tasks":["Zero-shot variant effect prediction"]},"source_ids":["pst-2025"],"links":[{"relation":"dataset","target_id":"reported-dataset-bd9255afb783d6"}],"attributes":{"entity_level":"task","version":null,"task":"Zero-shot variant effect prediction","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Zero-shot variant-effect assessment compares structural protein representations with their sequence-model backbones.","summary_source_ids":["pst-2025"],"summary_source_locator":"Results: zero-shot variant prediction; implementation; cached text lines 65, 84","sections":[{"title":"Evaluation methodology","body":"Variant-effect measurements for 38 proteins from the Riesselman/Meier DMS collections, assessed without assay-specific fine-tuning. The supplementary task section points to the original assay inventory. Average Spearman rank correlation. PST and ESM-2.","source_ids":["pst-2025","evidence-task-final-b-pst-supplement"],"source_locator":"Main §4.1.3; Supplementary Appendix C.6; Results: zero-shot variant prediction; implementation; cached text lines 65, 84; Results: zero-shot variant prediction; implementation; cached text lines 65, 84"}],"facts":[{"label":"Datasets","value":"Variant-effect measurements for 38 proteins from the Riesselman/Meier DMS collections, assessed without assay-specific fine-tuning. The supplementary task section points to the original assay inventory.","status":"source_checked","source_ids":["pst-2025","evidence-task-final-b-pst-supplement"],"source_locator":"Main §4.1.3; Supplementary Appendix C.6"},{"label":"Splits","value":"No additional task fine-tuning for the reported zero-shot comparison.","status":"source_checked","source_ids":["pst-2025"],"source_locator":"Results: zero-shot variant prediction; implementation; cached text lines 65, 84"},{"label":"Metrics","value":"Average Spearman rank correlation.","status":"source_checked","source_ids":["pst-2025"],"source_locator":"Results: zero-shot variant prediction; implementation; cached text lines 65, 84"},{"label":"Baselines","value":"PST and ESM-2.","status":"source_checked","source_ids":["pst-2025"],"source_locator":"Results: zero-shot variant prediction; implementation; cached text lines 65, 84"},{"label":"Leakage controls","value":"Variant-effect scoring uses no assay-label fine-tuning. The main variant-effect section and Supplementary Appendix C.6 do not report exclusion of the evaluated proteins or homologues from ESM-2 or structure pretraining; the identity/family exclusions in Appendix C.5 apply to other downstream tasks.","status":"unreported","source_ids":["pst-2025","evidence-task-final-b-pst-supplement"],"source_locator":"Main §4.1.3–4.1.4; Supplementary Appendix C.5–C.6, pp11–12"},{"label":"Uncertainty","value":"The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","status":"unreported","source_ids":["pst-2025"],"source_locator":"Results: zero-shot variant prediction; implementation; cached text lines 65, 84"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["pst-2025"],"source_locator":"Results: zero-shot variant prediction; implementation; cached text lines 65, 84"},{"label":"Organisms","value":"The variant-effect benchmark comprises 38 proteins from the Riesselman/Meier DMS collections. The main task section and supplementary variant-effect section do not enumerate their taxa; Supplementary C.6 instead points to the original DMS assay inventory.","status":"unreported","source_ids":["pst-2025","evidence-task-final-b-pst-supplement"],"source_locator":"Main §4.1.3; Supplementary Appendix C.6, pp11–12"},{"label":"Assays","value":"Protein variant-effect measurements.","status":"source_checked","source_ids":["pst-2025"],"source_locator":"Results: zero-shot variant prediction; implementation; cached text lines 65, 84"},{"label":"Allowed inputs","value":"Protein representations under PST and ESM-2 scoring.","status":"source_checked","source_ids":["pst-2025"],"source_locator":"Results: zero-shot variant prediction; implementation; cached text lines 65, 84"},{"label":"Adaptation","value":"No additional task fine-tuning in the reported zero-shot comparison.","status":"source_checked","source_ids":["pst-2025"],"source_locator":"Results: zero-shot variant prediction; implementation; cached text lines 65, 84"}],"strengths":[],"limitations":[{"text":"Structure-pretraining sources and downstream assay overlap require a separate audit. Structure-based splits described for ProteinShake are not assumed to define these variant-effect tasks.","source_ids":["pst-2025"],"source_locator":"Results: zero-shot variant prediction; implementation; cached text lines 65, 84"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Protein representations under PST and ESM-2 scoring.","Evaluation: No additional task fine-tuning in the reported zero-shot comparison.","Readout: Average Spearman rank correlation."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["pst-2025"],"source_locator":"Results: zero-shot variant prediction; implementation; cached text lines 65, 84"},"coverage":"limited","gaps":["Leakage controls: Variant-effect scoring uses no assay-label fine-tuning. The main variant-effect section and Supplementary Appendix C.6 do not report exclusion of the evaluated proteins or homologues from ESM-2 or structure pretraining; the identity/family exclusions in Appendix C.5 apply to other downstream tasks.","Uncertainty: The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","Organisms: The variant-effect benchmark comprises 38 proteins from the Riesselman/Meier DMS collections. The main task section and supplementary variant-effect section do not enumerate their taxa; Supplementary C.6 instead points to the original DMS assay inventory."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Relevant full-paper computational evaluation sections, tables/captions and cited supplementary task passages were reviewed. Reporting omissions are scoped to the inspected sources. Original numerical results are unchanged."}}}}
{"id":"reported-task-a7803ecf7708cc","kind":"benchmark","name":"Protein–ligand virtual screening","description":"","status":"needs_review","facets":{"areas":["molecular-interactions"],"tasks":["Protein–ligand virtual screening"]},"source_ids":["nmdn-2025"],"links":[{"relation":"dataset","target_id":"reported-dataset-065b9fcc8da573"}],"attributes":{"entity_level":"task","version":null,"task":"Protein–ligand virtual screening","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Virtual screening evaluates enrichment when candidate ligands are ranked using the paper’s docking/scoring workflow.","summary_source_ids":["nmdn-2025"],"summary_source_locator":"Results/Conclusion: virtual screening; cached text lines 97–110, 147","sections":[{"title":"Evaluation methodology","body":"LIT-PCBA is the virtual-screening collection; CASF and MerckFEP are separate scoring evaluations. Top-one-percent enrichment factor, summarized across the virtual-screening evaluation. DiffDock combined with NMDN scoring is the assessed workflow.","source_ids":["nmdn-2025"],"source_locator":"Results/Conclusion: virtual screening; cached text lines 97–110, 147"}],"facts":[{"label":"Datasets","value":"LIT-PCBA is the virtual-screening collection; CASF and MerckFEP are separate scoring evaluations.","status":"source_checked","source_ids":["nmdn-2025"],"source_locator":"Results/Conclusion: virtual screening; cached text lines 97–110, 147"},{"label":"Splits","value":"Models are fitted on the first RTMScore random PDBbind training/validation allocation. LIT-PCBA is a separate external screening collection, evaluated target by target; its compounds are not partitioned into a new supervised screening training set in the described experiment.","status":"source_checked","source_ids":["nmdn-2025"],"source_locator":"§4.1.1 Binder Data Set; LIT-PCBA screening description and Table 4"},{"label":"Metrics","value":"Top-one-percent enrichment factor, summarized across the virtual-screening evaluation.","status":"source_checked","source_ids":["nmdn-2025"],"source_locator":"Results/Conclusion: virtual screening; cached text lines 97–110, 147"},{"label":"Baselines","value":"DiffDock combined with NMDN scoring is the assessed workflow.","status":"source_checked","source_ids":["nmdn-2025"],"source_locator":"Results/Conclusion: virtual screening; cached text lines 97–110, 147"},{"label":"Leakage controls","value":"The binder-training filters remove PDB entries in the test sets. The weak-binder collection is reported to have no overlap with the adopted benchmarks. LIT-PCBA pose generation and scoring do not use known binder information for the evaluated target; the paper does not claim a universal protein-family or scaffold-disjoint test.","status":"source_checked","source_ids":["nmdn-2025"],"source_locator":"§4.1.1–4.1.2 including list items omitted from paragraph-only cache; LIT-PCBA screening discussion before Table 4"},{"label":"Uncertainty","value":"The LIT-PCBA table reports per-target enrichment factors, an across-target average and counts above enrichment thresholds. It does not report confidence intervals or repeated-run dispersion for these screening estimates; sampling variability reported for the PDBbind docking experiment is a different endpoint.","status":"unreported","source_ids":["nmdn-2025"],"source_locator":"LIT-PCBA screening results, Table 4 and its footnotes; contrast Table 1"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["nmdn-2025"],"source_locator":"Results/Conclusion: virtual screening; cached text lines 97–110, 147"},{"label":"Organisms","value":"The LIT-PCBA table identifies molecular targets, with many activity labels derived from cell-based phenotypic assays. The screening description and target table do not give a per-assay organism or cell-line inventory; target identity alone does not establish assay provenance.","status":"unreported","source_ids":["nmdn-2025"],"source_locator":"LIT-PCBA screening description, Table 4 and reference to Table S6"},{"label":"Assays","value":"Active/inactive compound screening labels.","status":"source_checked","source_ids":["nmdn-2025"],"source_locator":"Results/Conclusion: virtual screening; cached text lines 97–110, 147"},{"label":"Allowed inputs","value":"Ligands, target proteins and docking/scoring outputs.","status":"source_checked","source_ids":["nmdn-2025"],"source_locator":"Results/Conclusion: virtual screening; cached text lines 97–110, 147"},{"label":"Adaptation","value":"DiffDock pose generation combined with NMDN scoring; the screen is separate from CASF affinity evaluation.","status":"source_checked","source_ids":["nmdn-2025"],"source_locator":"Results/Conclusion: virtual screening; cached text lines 97–110, 147"}],"strengths":[],"limitations":[{"text":"Blind-docked screening and screening supplied with known binder information are different settings. Many LIT-PCBA actives arise from phenotypic assays and are not validated against the nominal target.","source_ids":["nmdn-2025"],"source_locator":"Results/Conclusion: virtual screening; cached text lines 97–110, 147"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Ligands, target proteins and docking/scoring outputs.","Evaluation: DiffDock pose generation combined with NMDN scoring; the screen is separate from CASF affinity evaluation.","Readout: Top-one-percent enrichment factor, summarized across the virtual-screening evaluation."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["nmdn-2025"],"source_locator":"Results/Conclusion: virtual screening; cached text lines 97–110, 147"},"coverage":"limited","gaps":["Uncertainty: The LIT-PCBA table reports per-target enrichment factors, an across-target average and counts above enrichment thresholds. It does not report confidence intervals or repeated-run dispersion for these screening estimates; sampling variability reported for the PDBbind docking experiment is a different endpoint.","Organisms: The LIT-PCBA table identifies molecular targets, with many activity labels derived from cell-based phenotypic assays. The screening description and target table do not give a per-assay organism or cell-line inventory; target identity alone does not establish assay provenance."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Relevant full-paper computational evaluation sections, tables/captions and cited supplementary task passages were reviewed. Reporting omissions are scoped to the inspected sources. Original numerical results are unchanged."}}}}
{"id":"reported-task-a78312d5df6dad","kind":"benchmark","name":"Protein–ligand binding affinity scoring","description":"","status":"needs_review","facets":{"areas":["molecular-interactions"],"tasks":["Protein–ligand binding affinity scoring"]},"source_ids":["akscore-2020"],"links":[{"relation":"dataset","target_id":"reported-dataset-f18fcc23dfa798"}],"attributes":{"entity_level":"task","version":null,"task":"Protein–ligand binding affinity scoring","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"AK-score is evaluated as a protein–ligand scoring function using PDBbind and CASF tasks that separate scoring, ranking and pose selection.","summary_source_ids":["akscore-2020"],"summary_source_locator":"Methods §§3.1, 3.5; Results §2.3; Table 4; cached text lines 24, 27, 40–41, 59–63","sections":[{"title":"Evaluation methodology","body":"PDBbind-2016 refined complexes with affinity labels; the core set supplies the principal test complexes. The core set is excluded from the refined training set; an additional evaluation uses entries newly added in PDBbind-2018. Pearson correlation for scoring; Spearman, Kendall and predictive index for ranking; top-ranked pose success for docking. Reimplemented KDEEP, AutoDock Vina and X-score are evaluated in the paper. Exact core complexes are removed from training; protein-family or ligand-scaffold independence is not established by that exclusion. The paper describes bootstrap comparison of correlation coefficients.","source_ids":["akscore-2020"],"source_locator":"Methods §§3.1, 3.5; Results §2.3; Table 4; cached text lines 24, 27, 40–41, 59–63"}],"facts":[{"label":"Datasets","value":"PDBbind-2016 refined complexes with affinity labels; the core set supplies the principal test complexes.","status":"source_checked","source_ids":["akscore-2020"],"source_locator":"Methods §§3.1, 3.5; Results §2.3; Table 4; cached text lines 24, 27, 40–41, 59–63"},{"label":"Splits","value":"The core set is excluded from the refined training set; an additional evaluation uses entries newly added in PDBbind-2018.","status":"source_checked","source_ids":["akscore-2020"],"source_locator":"Methods §§3.1, 3.5; Results §2.3; Table 4; cached text lines 24, 27, 40–41, 59–63"},{"label":"Metrics","value":"Pearson correlation for scoring; Spearman, Kendall and predictive index for ranking; top-ranked pose success for docking.","status":"source_checked","source_ids":["akscore-2020"],"source_locator":"Methods §§3.1, 3.5; Results §2.3; Table 4; cached text lines 24, 27, 40–41, 59–63"},{"label":"Baselines","value":"Reimplemented KDEEP, AutoDock Vina and X-score are evaluated in the paper.","status":"source_checked","source_ids":["akscore-2020"],"source_locator":"Methods §§3.1, 3.5; Results §2.3; Table 4; cached text lines 24, 27, 40–41, 59–63"},{"label":"Leakage controls","value":"Exact core complexes are removed from training; protein-family or ligand-scaffold independence is not established by that exclusion.","status":"source_checked","source_ids":["akscore-2020"],"source_locator":"Methods §§3.1, 3.5; Results §2.3; Table 4; cached text lines 24, 27, 40–41, 59–63"},{"label":"Uncertainty","value":"The paper describes bootstrap comparison of correlation coefficients.","status":"source_checked","source_ids":["akscore-2020"],"source_locator":"Methods §§3.1, 3.5; Results §2.3; Table 4; cached text lines 24, 27, 40–41, 59–63"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["akscore-2020"],"source_locator":"Methods §§3.1, 3.5; Results §2.3; Table 4; cached text lines 24, 27, 40–41, 59–63"},{"label":"Organisms","value":"The benchmark selects PDBbind/CASF protein–ligand complexes by structural and affinity criteria. The dataset Methods and CASF evaluation section do not report a species-stratified inventory.","status":"unreported","source_ids":["akscore-2020"],"source_locator":"Methods: Protein–ligand data; CASF-2016 evaluation"},{"label":"Assays","value":"Experimentally annotated protein–ligand affinities and structural poses.","status":"source_checked","source_ids":["akscore-2020"],"source_locator":"Methods §§3.1, 3.5; Results §2.3; Table 4; cached text lines 24, 27, 40–41, 59–63"},{"label":"Allowed inputs","value":"Protein–ligand complex structures.","status":"source_checked","source_ids":["akscore-2020"],"source_locator":"Methods §§3.1, 3.5; Results §2.3; Table 4; cached text lines 24, 27, 40–41, 59–63"},{"label":"Adaptation","value":"Supervised affinity scoring fitted on the refined set after excluding core test complexes.","status":"source_checked","source_ids":["akscore-2020"],"source_locator":"Methods §§3.1, 3.5; Results §2.3; Table 4; cached text lines 24, 27, 40–41, 59–63"}],"strengths":[{"text":"Core test complexes are removed from the affinity-training set, and CASF separates pose sampling from scoring.","source_ids":["akscore-2020"],"source_locator":"Methods §§3.1, 3.5; Results §2.3; Table 4; cached text lines 24, 27, 40–41, 59–63"}],"limitations":[{"text":"CASF scoring evaluates supplied structures and decouples scoring from pose sampling; it is not an end-to-end docking benchmark.","source_ids":["akscore-2020"],"source_locator":"Methods §§3.1, 3.5; Results §2.3; Table 4; cached text lines 24, 27, 40–41, 59–63"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Protein–ligand complex structures.","Evaluation: Supervised affinity scoring fitted on the refined set after excluding core test complexes.","Readout: Pearson correlation for scoring; Spearman, Kendall and predictive index for ranking; top-ranked pose success for docking."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["akscore-2020"],"source_locator":"Methods §§3.1, 3.5; Results §2.3; Table 4; cached text lines 24, 27, 40–41, 59–63"},"coverage":"limited","gaps":["Organisms: The benchmark selects PDBbind/CASF protein–ligand complexes by structural and affinity criteria. The dataset Methods and CASF evaluation section do not report a species-stratified inventory."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Relevant full-paper computational evaluation sections, tables/captions and cited supplementary task passages were reviewed. Reporting omissions are scoped to the inspected sources. Original numerical results are unchanged."}}}}
{"id":"reported-task-ac191e878dff5e","kind":"benchmark","name":"transcription-factor DNA binding-site prediction","description":"","status":"needs_review","facets":{"areas":["molecular-interactions"],"tasks":["transcription-factor DNA binding-site prediction"]},"source_ids":["transbind-2026"],"links":[{"relation":"dataset","target_id":"reported-dataset-034c60a2dabc73"}],"attributes":{"entity_level":"task","version":null,"task":"transcription-factor DNA binding-site prediction","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Transcription-factor binding classification evaluates many TF–cell-type labels on chromosome-held-out genomic sequences.","summary_source_ids":["transbind-2026"],"summary_source_locator":"Methods: DNA data; Training and evaluation; cached text lines 13–18, 49–51; comparative evaluation and ablation passages; uncertainty/repeat-run/statistical-comparison passages","sections":[{"title":"Evaluation methodology","body":"EPBDXDNA/ENCODE ChIP-seq-derived human genomic bins with multilabel binding annotations. Chromosomes 8/9 are held out for testing, chromosome 7 for validation and remaining specified chromosomes for training. Per-label AUROC and AUPR are macro-averaged across TF–cell-type experiments. TBiNet and EPBDXDNABERT-2 are explicit comparators; the latter’s underlying predictions were unavailable for paired statistical analysis. Chromosome-separated partitions reduce direct genomic overlap; validation data drive hyperparameter selection. Paired t-tests against TBiNet use TF–cell-type combinations as the comparison units; unavailable comparator predictions prevent the same paired analysis for every baseline.","source_ids":["transbind-2026"],"source_locator":"Methods: DNA data; Training and evaluation; cached text lines 13–18, 49–51; comparative evaluation and ablation passages; uncertainty/repeat-run/statistical-comparison passages"}],"facts":[{"label":"Datasets","value":"EPBDXDNA/ENCODE ChIP-seq-derived human genomic bins with multilabel binding annotations.","status":"source_checked","source_ids":["transbind-2026"],"source_locator":"Methods: DNA data; Training and evaluation; cached text lines 13–18, 49–51; comparative evaluation and ablation passages; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Splits","value":"Chromosomes 8/9 are held out for testing, chromosome 7 for validation and remaining specified chromosomes for training.","status":"source_checked","source_ids":["transbind-2026"],"source_locator":"Methods: DNA data; Training and evaluation; cached text lines 13–18, 49–51; comparative evaluation and ablation passages; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Metrics","value":"Per-label AUROC and AUPR are macro-averaged across TF–cell-type experiments.","status":"source_checked","source_ids":["transbind-2026"],"source_locator":"Methods: DNA data; Training and evaluation; cached text lines 13–18, 49–51; comparative evaluation and ablation passages; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Baselines","value":"TBiNet and EPBDXDNABERT-2 are explicit comparators; the latter’s underlying predictions were unavailable for paired statistical analysis.","status":"source_checked","source_ids":["transbind-2026"],"source_locator":"Methods: DNA data; Training and evaluation; cached text lines 13–18, 49–51; comparative evaluation and ablation passages; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Leakage controls","value":"Chromosome-separated partitions reduce direct genomic overlap; validation data drive hyperparameter selection.","status":"source_checked","source_ids":["transbind-2026"],"source_locator":"Methods: DNA data; Training and evaluation; cached text lines 13–18, 49–51; comparative evaluation and ablation passages; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Uncertainty","value":"Paired t-tests against TBiNet use TF–cell-type combinations as the comparison units; unavailable comparator predictions prevent the same paired analysis for every baseline.","status":"source_checked","source_ids":["transbind-2026"],"source_locator":"Methods: DNA data; Training and evaluation; cached text lines 13–18, 49–51; comparative evaluation and ablation passages; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["transbind-2026"],"source_locator":"Methods: DNA data; Training and evaluation; cached text lines 13–18, 49–51; comparative evaluation and ablation passages; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Organisms","value":"Human.","status":"source_checked","source_ids":["transbind-2026"],"source_locator":"Methods: DNA data; Training and evaluation; cached text lines 13–18, 49–51; comparative evaluation and ablation passages; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Assays","value":"ENCODE ChIP-seq transcription-factor binding annotations.","status":"source_checked","source_ids":["transbind-2026"],"source_locator":"Methods: DNA data; Training and evaluation; cached text lines 13–18, 49–51; comparative evaluation and ablation passages; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Allowed inputs","value":"Genomic sequence bins with multilabel targets.","status":"source_checked","source_ids":["transbind-2026"],"source_locator":"Methods: DNA data; Training and evaluation; cached text lines 13–18, 49–51; comparative evaluation and ablation passages; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Adaptation","value":"Supervised binding prediction with chromosome-separated training/validation/test data.","status":"source_checked","source_ids":["transbind-2026"],"source_locator":"Methods: DNA data; Training and evaluation; cached text lines 13–18, 49–51; comparative evaluation and ablation passages; uncertainty/repeat-run/statistical-comparison passages"}],"strengths":[],"limitations":[{"text":"Equal label weighting differs from pooling all bins and labels. Exact comparator identities, confidence intervals and long-range/homology overlap remain unextracted.","source_ids":["transbind-2026"],"source_locator":"Methods: DNA data; Training and evaluation; cached text lines 13–18, 49–51; comparative evaluation and ablation passages; uncertainty/repeat-run/statistical-comparison passages"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Genomic sequence bins with multilabel targets.","Evaluation: Supervised binding prediction with chromosome-separated training/validation/test data.","Readout: Per-label AUROC and AUPR are macro-averaged across TF–cell-type experiments."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["transbind-2026"],"source_locator":"Methods: DNA data; Training and evaluation; cached text lines 13–18, 49–51; comparative evaluation and ablation passages; uncertainty/repeat-run/statistical-comparison passages"},"coverage":"limited","gaps":["Equal label weighting differs from pooling all bins and labels. Exact comparator identities, confidence intervals and long-range/homology overlap remain unextracted."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Task-specific computational methodology and field context checked in the cited primary-source artifact. Source-backed fields, inapplicable evaluator dimensions and unresolved details are distinguished. Numerical results were not reproduced."}}}}
{"id":"reported-task-b00a636d1ed8d9","kind":"benchmark","name":"RNA-small-molecule binding-site prediction","description":"","status":"needs_review","facets":{"areas":["rna-transcriptomes"],"tasks":["RNA-small-molecule binding-site prediction"]},"source_ids":["rlsite-rna-binding-2025"],"links":[{"relation":"dataset","target_id":"reported-dataset-1c7f8ebb1968d9"}],"attributes":{"entity_level":"task","version":null,"task":"RNA-small-molecule binding-site prediction","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"RNA small-molecule binding-site prediction uses independent RNA–ligand test collections with explicit redundancy filtering.","summary_source_ids":["rlsite-rna-binding-2025"],"summary_source_locator":"Methods §§2.1, 2.7; cached text lines 13–15, 57–58","sections":[{"title":"Evaluation methodology","body":"Combined RNAglib/RNAsite training data; T18, T3 and newly curated T10 tests. The named test collections are held separate; T10 uses later PDB entries. Precision, recall, MCC and ROC-AUC. Training data align with MultiModRLBP for comparison. T3 removes high-similarity RNAs relative to training; T10 is clustered and filtered against training RNA. The paper reports t-tests comparing metrics; the exact replication unit should be confirmed before interpreting them as independent-sample uncertainty.","source_ids":["rlsite-rna-binding-2025"],"source_locator":"Methods §§2.1, 2.7; cached text lines 13–15, 57–58"}],"facts":[{"label":"Datasets","value":"Combined RNAglib/RNAsite training data; T18, T3 and newly curated T10 tests.","status":"source_checked","source_ids":["rlsite-rna-binding-2025"],"source_locator":"Methods §§2.1, 2.7; cached text lines 13–15, 57–58"},{"label":"Splits","value":"The named test collections are held separate; T10 uses later PDB entries.","status":"source_checked","source_ids":["rlsite-rna-binding-2025"],"source_locator":"Methods §§2.1, 2.7; cached text lines 13–15, 57–58"},{"label":"Metrics","value":"Precision, recall, MCC and ROC-AUC.","status":"source_checked","source_ids":["rlsite-rna-binding-2025"],"source_locator":"Methods §§2.1, 2.7; cached text lines 13–15, 57–58"},{"label":"Baselines","value":"Training data align with MultiModRLBP for comparison.","status":"source_checked","source_ids":["rlsite-rna-binding-2025"],"source_locator":"Methods §§2.1, 2.7; cached text lines 13–15, 57–58"},{"label":"Leakage controls","value":"T3 removes high-similarity RNAs relative to training; T10 is clustered and filtered against training RNA.","status":"source_checked","source_ids":["rlsite-rna-binding-2025"],"source_locator":"Methods §§2.1, 2.7; cached text lines 13–15, 57–58"},{"label":"Uncertainty","value":"The paper reports t-tests comparing metrics; the exact replication unit should be confirmed before interpreting them as independent-sample uncertainty.","status":"source_checked","source_ids":["rlsite-rna-binding-2025"],"source_locator":"Methods §§2.1, 2.7; cached text lines 13–15, 57–58"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["rlsite-rna-binding-2025"],"source_locator":"Methods §§2.1, 2.7; cached text lines 13–15, 57–58"},{"label":"Organisms","value":"RNAglib/RNAsite-derived RNA chains and PDB-derived test sets define this benchmark. Section 2.1 reports chain and ligand membership criteria but not a taxonomic inventory; synthetic aptamer examples cannot be assigned an organism from the surrounding disease discussion.","status":"unreported","source_ids":["rlsite-rna-binding-2025"],"source_locator":"§2.1 Benchmark datasets; RNA–ligand test examples"},{"label":"Assays","value":"RNA–small-molecule binding-site annotations.","status":"source_checked","source_ids":["rlsite-rna-binding-2025"],"source_locator":"Methods §§2.1, 2.7; cached text lines 13–15, 57–58"},{"label":"Allowed inputs","value":"RNA representations for site prediction.","status":"source_checked","source_ids":["rlsite-rna-binding-2025"],"source_locator":"Methods §§2.1, 2.7; cached text lines 13–15, 57–58"},{"label":"Adaptation","value":"Supervised fitting on the combined training data; later PDB entries supply one held-out test.","status":"source_checked","source_ids":["rlsite-rna-binding-2025"],"source_locator":"Methods §§2.1, 2.7; cached text lines 13–15, 57–58"}],"strengths":[],"limitations":[{"text":"Small RNA test sets limit the number of independent examples. Chain/sequence exclusions do not by themselves establish ligand-disjoint evaluation.","source_ids":["rlsite-rna-binding-2025"],"source_locator":"Methods §§2.1, 2.7; cached text lines 13–15, 57–58"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: RNA representations for site prediction.","Evaluation: Supervised fitting on the combined training data; later PDB entries supply one held-out test.","Readout: Precision, recall, MCC and ROC-AUC."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["rlsite-rna-binding-2025"],"source_locator":"Methods §§2.1, 2.7; cached text lines 13–15, 57–58"},"coverage":"limited","gaps":["Organisms: RNAglib/RNAsite-derived RNA chains and PDB-derived test sets define this benchmark. Section 2.1 reports chain and ligand membership criteria but not a taxonomic inventory; synthetic aptamer examples cannot be assigned an organism from the surrounding disease discussion."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Relevant full-paper computational evaluation sections, tables/captions and cited supplementary task passages were reviewed. Reporting omissions are scoped to the inspected sources. Original numerical results are unchanged."}}}}
{"id":"reported-task-b181ed450cdd41","kind":"benchmark","name":"protein-small molecule binding-site prediction","description":"","status":"needs_review","facets":{"areas":["proteins-complexes"],"tasks":["protein-small molecule binding-site prediction"]},"source_ids":["clape-smb-2024"],"links":[{"relation":"dataset","target_id":"reported-dataset-701d910b02d25c"}],"attributes":{"entity_level":"task","version":null,"task":"protein-small molecule binding-site prediction","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Small-molecule binding-site classification evaluates residue predictions on structural and curated protein annotations.","summary_source_ids":["clape-smb-2024"],"summary_source_locator":"Methods: Evaluation metrics; SJC dataset preparation; UniProtSMB dataset preparation; Discussion; cached text lines 24–25, 35–43, 93; comparative evaluation and ablation passages; uncertainty/repeat-run/statistical-comparison passages","sections":[{"title":"Evaluation methodology","body":"SJC combines sc-PDB, JOINED and COACH420; UniProtSMB supplies a separately curated binding-site dataset. UniProtSMB representative proteins are partitioned 80:10:10 for training, validation and testing. Precision, recall, MCC, AUROC and AUPRC. ESM-2 versus ProtBert embeddings and MLP/CNN/Transformer head ablations; some head/model comparisons use the test set. Protein similarity clustering and cluster-aware split comparisons are described; ESM-2 pretraining sequences are not universally excluded. Multiple random seeds are evaluated; the paper reports fold-averaged metrics and standard deviations for its robustness experiment.","source_ids":["clape-smb-2024"],"source_locator":"Methods: Evaluation metrics; SJC dataset preparation; UniProtSMB dataset preparation; Discussion; cached text lines 24–25, 35–43, 93; comparative evaluation and ablation passages; uncertainty/repeat-run/statistical-comparison passages"}],"facts":[{"label":"Datasets","value":"SJC combines sc-PDB, JOINED and COACH420; UniProtSMB supplies a separately curated binding-site dataset.","status":"source_checked","source_ids":["clape-smb-2024"],"source_locator":"Methods: Evaluation metrics; SJC dataset preparation; UniProtSMB dataset preparation; Discussion; cached text lines 24–25, 35–43, 93; comparative evaluation and ablation passages; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Splits","value":"UniProtSMB representative proteins are partitioned 80:10:10 for training, validation and testing.","status":"source_checked","source_ids":["clape-smb-2024"],"source_locator":"Methods: Evaluation metrics; SJC dataset preparation; UniProtSMB dataset preparation; Discussion; cached text lines 24–25, 35–43, 93; comparative evaluation and ablation passages; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Metrics","value":"Precision, recall, MCC, AUROC and AUPRC.","status":"source_checked","source_ids":["clape-smb-2024"],"source_locator":"Methods: Evaluation metrics; SJC dataset preparation; UniProtSMB dataset preparation; Discussion; cached text lines 24–25, 35–43, 93; comparative evaluation and ablation passages; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Baselines","value":"ESM-2 versus ProtBert embeddings and MLP/CNN/Transformer head ablations; some head/model comparisons use the test set.","status":"source_checked","source_ids":["clape-smb-2024"],"source_locator":"Methods: Evaluation metrics; SJC dataset preparation; UniProtSMB dataset preparation; Discussion; cached text lines 24–25, 35–43, 93; comparative evaluation and ablation passages; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Leakage controls","value":"Protein similarity clustering and cluster-aware split comparisons are described; ESM-2 pretraining sequences are not universally excluded.","status":"source_checked","source_ids":["clape-smb-2024"],"source_locator":"Methods: Evaluation metrics; SJC dataset preparation; UniProtSMB dataset preparation; Discussion; cached text lines 24–25, 35–43, 93; comparative evaluation and ablation passages; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Uncertainty","value":"Multiple random seeds are evaluated; the paper reports fold-averaged metrics and standard deviations for its robustness experiment.","status":"source_checked","source_ids":["clape-smb-2024"],"source_locator":"Methods: Evaluation metrics; SJC dataset preparation; UniProtSMB dataset preparation; Discussion; cached text lines 24–25, 35–43, 93; comparative evaluation and ablation passages; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["clape-smb-2024"],"source_locator":"Methods: Evaluation metrics; SJC dataset preparation; UniProtSMB dataset preparation; Discussion; cached text lines 24–25, 35–43, 93; comparative evaluation and ablation passages; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Organisms","value":"The SJC structural collections and UniProtSMB annotations are selected by binding-site evidence and protein redundancy. Their preparation sections and dataset tables do not summarize organism composition or define a species-specific test.","status":"unreported","source_ids":["clape-smb-2024"],"source_locator":"SJC dataset preparation; UniProtSMB dataset preparation; Table 1"},{"label":"Assays","value":"Curated protein–small-molecule binding-site annotations.","status":"source_checked","source_ids":["clape-smb-2024"],"source_locator":"Methods: Evaluation metrics; SJC dataset preparation; UniProtSMB dataset preparation; Discussion; cached text lines 24–25, 35–43, 93; comparative evaluation and ablation passages; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Allowed inputs","value":"Protein amino-acid sequences.","status":"source_checked","source_ids":["clape-smb-2024"],"source_locator":"Methods: Evaluation metrics; SJC dataset preparation; UniProtSMB dataset preparation; Discussion; cached text lines 24–25, 35–43, 93; comparative evaluation and ablation passages; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Adaptation","value":"Supervised residue-level predictor fitted on the training proteins.","status":"source_checked","source_ids":["clape-smb-2024"],"source_locator":"Methods: Evaluation metrics; SJC dataset preparation; UniProtSMB dataset preparation; Discussion; cached text lines 24–25, 35–43, 93; comparative evaluation and ablation passages; uncertainty/repeat-run/statistical-comparison passages"}],"strengths":[],"limitations":[{"text":"Binding residues are rare and incomplete annotations are discussed in the source. SJC, COACH420 and UniProtSMB results are distinct evaluations, not one shared test set.","source_ids":["clape-smb-2024"],"source_locator":"Methods: Evaluation metrics; SJC dataset preparation; UniProtSMB dataset preparation; Discussion; cached text lines 24–25, 35–43, 93; comparative evaluation and ablation passages; uncertainty/repeat-run/statistical-comparison passages"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Protein amino-acid sequences.","Evaluation: Supervised residue-level predictor fitted on the training proteins.","Readout: Precision, recall, MCC, AUROC and AUPRC."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["clape-smb-2024"],"source_locator":"Methods: Evaluation metrics; SJC dataset preparation; UniProtSMB dataset preparation; Discussion; cached text lines 24–25, 35–43, 93; comparative evaluation and ablation passages; uncertainty/repeat-run/statistical-comparison passages"},"coverage":"limited","gaps":["Organisms: The SJC structural collections and UniProtSMB annotations are selected by binding-site evidence and protein redundancy. Their preparation sections and dataset tables do not summarize organism composition or define a species-specific test."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Relevant full-paper computational evaluation sections, tables/captions and cited supplementary task passages were reviewed. Reporting omissions are scoped to the inspected sources. Original numerical results are unchanged."}}}}
{"id":"reported-task-b46b7b839bff93","kind":"benchmark","name":"PBMC cell-type classification","description":"","status":"needs_review","facets":{"areas":["cells-tissues"],"tasks":["PBMC cell-type classification"]},"source_ids":["scalr-2025"],"links":[{"relation":"dataset","target_id":"reported-dataset-33a41fe5fc66cf"}],"attributes":{"entity_level":"task","version":null,"task":"PBMC cell-type classification","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"PBMC cell classification assesses predictive performance and resource use under a common computing environment.","summary_source_ids":["scalr-2025"],"summary_source_locator":"Results: Performance evaluation using all genes; robustness evaluations; cached text lines 13–14, 22–24","sections":[{"title":"Evaluation methodology","body":"PBMC bacterial-sepsis data are used for the principal all-gene comparison. Accuracy and, in broader experiments, precision, recall and F1, alongside memory/runtime. scVI/scANVI, SingleCellNet, CellTypist, ACTINN and devCellPy.","source_ids":["scalr-2025"],"source_locator":"Results: Performance evaluation using all genes; robustness evaluations; cached text lines 13–14, 22–24; Results: Performance evaluation using all genes; robustness evaluations; cached text lines 13–14, 22–24; Results: Performance evaluation using all genes; robustness evaluations; cached text lines 13–14, 22–24"}],"facts":[{"label":"Datasets","value":"PBMC bacterial-sepsis data are used for the principal all-gene comparison.","status":"source_checked","source_ids":["scalr-2025"],"source_locator":"Results: Performance evaluation using all genes; robustness evaluations; cached text lines 13–14, 22–24"},{"label":"Splits","value":"Input cells are divided into training, validation and testing subsets, with dataset sizes in Table 1. The Methods do not specify donor-disjoint grouping for the PBMC comparison.","status":"source_checked","source_ids":["scalr-2025"],"source_locator":"Table 1; Methods: Data processing and Training"},{"label":"Metrics","value":"Accuracy and, in broader experiments, precision, recall and F1, alongside memory/runtime.","status":"source_checked","source_ids":["scalr-2025"],"source_locator":"Results: Performance evaluation using all genes; robustness evaluations; cached text lines 13–14, 22–24"},{"label":"Baselines","value":"scVI/scANVI, SingleCellNet, CellTypist, ACTINN and devCellPy.","status":"source_checked","source_ids":["scalr-2025"],"source_locator":"Results: Performance evaluation using all genes; robustness evaluations; cached text lines 13–14, 22–24"},{"label":"Leakage controls","value":"Feature-selection models use training data and validation data; the final classifier is selected with validation performance. The inspected PBMC comparison and data-processing sections do not specify donor-disjoint or batch-disjoint partitions, so a cell split cannot be treated as a held-out-donor experiment.","status":"unreported","source_ids":["scalr-2025"],"source_locator":"Methods: Data processing, Feature extraction, Training and tool comparisons; cached paragraphs 63–82"},{"label":"Uncertainty","value":"The PBMC cell-type comparison reports classification metrics, but its results, figure captions and comparison methods do not define repeated-seed dispersion or a donor/sample-level confidence-interval procedure for those metrics.","status":"unreported","source_ids":["scalr-2025"],"source_locator":"PBMC comparison results; associated figure captions; Methods: comparison of scaLR with other pipelines"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["scalr-2025"],"source_locator":"Results: Performance evaluation using all genes; robustness evaluations; cached text lines 13–14, 22–24"},{"label":"Organisms","value":"Human PBMCs.","status":"source_checked","source_ids":["scalr-2025"],"source_locator":"Results: Performance evaluation using all genes; robustness evaluations; cached text lines 13–14, 22–24"},{"label":"Assays","value":"Single-cell bacterial-sepsis expression with cell-type annotations.","status":"source_checked","source_ids":["scalr-2025"],"source_locator":"Results: Performance evaluation using all genes; robustness evaluations; cached text lines 13–14, 22–24"},{"label":"Allowed inputs","value":"Single-cell gene-expression profiles.","status":"source_checked","source_ids":["scalr-2025"],"source_locator":"Results: Performance evaluation using all genes; robustness evaluations; cached text lines 13–14, 22–24"},{"label":"Adaptation","value":"Supervised neural-network cell-type classification; validation data select the training checkpoint.","status":"source_checked","source_ids":["scalr-2025"],"source_locator":"Methods: Feature extraction and Training"}],"strengths":[],"limitations":[{"text":"The paper contains separate cell-state and cell-type outcomes and larger cross-cohort evaluations. These should not be merged into one PBMC protocol.","source_ids":["scalr-2025"],"source_locator":"Results: Performance evaluation using all genes; robustness evaluations; cached text lines 13–14, 22–24"}],"diagram":{"title":"Computational evaluation flow","steps":["Allowed inputs: Single-cell gene-expression profiles.","Datasets: PBMC bacterial-sepsis data are used for the principal all-gene comparison.","Metrics: Accuracy and, in broader experiments, precision, recall and F1, alongside memory/runtime."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["scalr-2025"],"source_locator":"Results: Performance evaluation using all genes; robustness evaluations; cached text lines 13–14, 22–24; Results: Performance evaluation using all genes; robustness evaluations; cached text lines 13–14, 22–24; Results: Performance evaluation using all genes; robustness evaluations; cached text lines 13–14, 22–24"},"coverage":"limited","gaps":["Leakage controls: Feature-selection models use training data and validation data; the final classifier is selected with validation performance. The inspected PBMC comparison and data-processing sections do not specify donor-disjoint or batch-disjoint partitions, so a cell split cannot be treated as a held-out-donor experiment.","Uncertainty: The PBMC cell-type comparison reports classification metrics, but its results, figure captions and comparison methods do not define repeated-seed dispersion or a donor/sample-level confidence-interval procedure for those metrics."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Relevant full-paper computational evaluation sections, tables/captions and cited supplementary task passages were reviewed. Reporting omissions are scoped to the inspected sources. Original numerical results are unchanged."}}}}
{"id":"reported-task-b9199a30a0bcb2","kind":"benchmark","name":"regulatory-variant scoring","description":"","status":"needs_review","facets":{"areas":["dna-genomes"],"tasks":["regulatory-variant scoring"]},"source_ids":["arsenal-regulatory-dna-2026"],"links":[{"relation":"dataset","target_id":"reported-dataset-158b121281b650"}],"attributes":{"entity_level":"task","version":null,"task":"regulatory-variant scoring","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Regulatory-variant scoring is assessed against DART-Eval quantitative-trait-locus effect annotations.","summary_source_ids":["arsenal-regulatory-dna-2026"],"summary_source_locator":"Methods: Zero-Shot Variant Effect Prediction; Pretraining Data; cached text lines 50–52, 85–89","sections":[{"title":"Evaluation methodology","body":"DNase-sensitivity QTLs and chromatin-accessibility QTLs used in DART-Eval. The paper describes chromosome-separated pretraining partitions; evaluation is zero-shot variant scoring. Predicted allele scores are compared with observed effects; correlation analysis is restricted to variants with significant observed effects.","source_ids":["arsenal-regulatory-dna-2026"],"source_locator":"Methods: Zero-Shot Variant Effect Prediction; Pretraining Data; cached text lines 50–52, 85–89"}],"facts":[{"label":"Datasets","value":"DNase-sensitivity QTLs and chromatin-accessibility QTLs used in DART-Eval.","status":"source_checked","source_ids":["arsenal-regulatory-dna-2026"],"source_locator":"Methods: Zero-Shot Variant Effect Prediction; Pretraining Data; cached text lines 50–52, 85–89"},{"label":"Splits","value":"The paper describes chromosome-separated pretraining partitions; evaluation is zero-shot variant scoring.","status":"source_checked","source_ids":["arsenal-regulatory-dna-2026"],"source_locator":"Methods: Zero-Shot Variant Effect Prediction; Pretraining Data; cached text lines 50–52, 85–89"},{"label":"Metrics","value":"Predicted allele scores are compared with observed effects; correlation analysis is restricted to variants with significant observed effects.","status":"source_checked","source_ids":["arsenal-regulatory-dna-2026"],"source_locator":"Methods: Zero-Shot Variant Effect Prediction; Pretraining Data; cached text lines 50–52, 85–89"},{"label":"Baselines","value":"Figure 4 compares with DNA-language-model scores taken from DART-EVAL and includes an ARSENAL No Prior ablation. These are imported zero-shot likelihood-scoring comparisons, separate from the paper’s supervised ChromBPNet experiments.","status":"source_checked","source_ids":["arsenal-regulatory-dna-2026"],"source_locator":"Results: zero-shot regulatory-QTL scoring, Fig.4 caption; Methods: Zero-shot Variant Effect Prediction"},{"label":"Leakage controls","value":"The variant experiment uses no variant-annotation supervision and evaluates significant-effect QTLs. The zero-shot methods do not document removal of these QTL loci from sequence pretraining; chromosome splits described for pretraining and supervised models are not an explicit QTL-overlap audit.","status":"unreported","source_ids":["arsenal-regulatory-dna-2026"],"source_locator":"Zero-shot Variant Effect Prediction; pretraining partitions; Supervised Model Training; Fig.4"},{"label":"Uncertainty","value":"The zero-shot results and Figure 4 caption do not specify a QTL resampling unit, replicate count or confidence-interval procedure. Variability reported in the supervised ChromBPNet comparison does not establish uncertainty for these zero-shot correlations.","status":"unreported","source_ids":["arsenal-regulatory-dna-2026"],"source_locator":"Zero-shot regulatory-QTL results and Fig.4 caption; Methods: Zero-shot Variant Effect Prediction"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["arsenal-regulatory-dna-2026"],"source_locator":"Methods: Zero-Shot Variant Effect Prediction; Pretraining Data; cached text lines 50–52, 85–89"},{"label":"Organisms","value":"Human.","status":"source_checked","source_ids":["arsenal-regulatory-dna-2026"],"source_locator":"Methods: Zero-Shot Variant Effect Prediction; Pretraining Data; cached text lines 50–52, 85–89"},{"label":"Assays","value":"DNase-sensitivity and chromatin-accessibility QTL annotations.","status":"source_checked","source_ids":["arsenal-regulatory-dna-2026"],"source_locator":"Methods: Zero-Shot Variant Effect Prediction; Pretraining Data; cached text lines 50–52, 85–89"},{"label":"Allowed inputs","value":"Reference/alternate DNA sequences at the evaluated variants.","status":"source_checked","source_ids":["arsenal-regulatory-dna-2026"],"source_locator":"Methods: Zero-Shot Variant Effect Prediction; Pretraining Data; cached text lines 50–52, 85–89"},{"label":"Adaptation","value":"Zero-shot variant scoring; pretraining chromosome partitions are separate from the QTL evaluation.","status":"source_checked","source_ids":["arsenal-regulatory-dna-2026"],"source_locator":"Methods: Zero-Shot Variant Effect Prediction; Pretraining Data; cached text lines 50–52, 85–89"}],"strengths":[],"limitations":[{"text":"Correlation is calculated only on significant-effect variants, changing the evaluated population. Zero-shot label use does not establish that their genomic contexts were absent from pretraining.","source_ids":["arsenal-regulatory-dna-2026"],"source_locator":"Methods: Zero-Shot Variant Effect Prediction; Pretraining Data; cached text lines 50–52, 85–89"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Reference/alternate DNA sequences at the evaluated variants.","Evaluation: Zero-shot variant scoring; pretraining chromosome partitions are separate from the QTL evaluation.","Readout: Predicted allele scores are compared with observed effects; correlation analysis is restricted to variants with significant observed effects."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["arsenal-regulatory-dna-2026"],"source_locator":"Methods: Zero-Shot Variant Effect Prediction; Pretraining Data; cached text lines 50–52, 85–89"},"coverage":"limited","gaps":["Leakage controls: The variant experiment uses no variant-annotation supervision and evaluates significant-effect QTLs. The zero-shot methods do not document removal of these QTL loci from sequence pretraining; chromosome splits described for pretraining and supervised models are not an explicit QTL-overlap audit.","Uncertainty: The zero-shot results and Figure 4 caption do not specify a QTL resampling unit, replicate count or confidence-interval procedure. Variability reported in the supervised ChromBPNet comparison does not establish uncertainty for these zero-shot correlations."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Relevant full-paper computational evaluation sections, tables/captions and cited supplementary task passages were reviewed. Reporting omissions are scoped to the inspected sources. Original numerical results are unchanged."}}}}
{"id":"reported-task-bf513ed6db92c5","kind":"benchmark","name":"Protein–ligand pose prediction","description":"","status":"needs_review","facets":{"areas":["molecular-interactions"],"tasks":["Protein–ligand pose prediction"]},"source_ids":["boltz-stereochemistry-2025"],"links":[{"relation":"dataset","target_id":"reported-dataset-5afaefb87c8a94"}],"attributes":{"entity_level":"task","version":null,"task":"Protein–ligand pose prediction","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Protein–ligand pose assessment distinguishes geometric accuracy from stereochemical validity and examines temporal distribution shift.","summary_source_ids":["boltz-stereochemistry-2025"],"summary_source_locator":"Methods: Benchmark Dataset; Evaluation Metrics; Results: Table 1; cached text lines 8–14, 41–44","sections":[{"title":"Evaluation methodology","body":"PLINDER and PDBbind2020 complexes; PLINDER subsets use ligand-similarity filtering. Before/After subsets are separated by the structure-prediction training cutoff; stricter subsets additionally filter protein and ligand similarity. Protein and ligand RMSD, chirality reproduction, bond-length RMSD and bond-angle RMSD; aggregate definitions vary by metric. AlphaFold3, Boltz-1 and restraint variants; AutoDock Vina and DiffDock for applicable ligand metrics. Temporal and similarity-based subsets are examined separately; Before-set scores include structures potentially seen in training. The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","source_ids":["boltz-stereochemistry-2025"],"source_locator":"Methods: Benchmark Dataset; Evaluation Metrics; Results: Table 1; cached text lines 8–14, 41–44"}],"facts":[{"label":"Datasets","value":"PLINDER and PDBbind2020 complexes; PLINDER subsets use ligand-similarity filtering.","status":"source_checked","source_ids":["boltz-stereochemistry-2025"],"source_locator":"Methods: Benchmark Dataset; Evaluation Metrics; Results: Table 1; cached text lines 8–14, 41–44"},{"label":"Splits","value":"Before/After subsets are separated by the structure-prediction training cutoff; stricter subsets additionally filter protein and ligand similarity.","status":"source_checked","source_ids":["boltz-stereochemistry-2025"],"source_locator":"Methods: Benchmark Dataset; Evaluation Metrics; Results: Table 1; cached text lines 8–14, 41–44"},{"label":"Metrics","value":"Protein and ligand RMSD, chirality reproduction, bond-length RMSD and bond-angle RMSD; aggregate definitions vary by metric.","status":"source_checked","source_ids":["boltz-stereochemistry-2025"],"source_locator":"Methods: Benchmark Dataset; Evaluation Metrics; Results: Table 1; cached text lines 8–14, 41–44"},{"label":"Baselines","value":"AlphaFold3, Boltz-1 and restraint variants; AutoDock Vina and DiffDock for applicable ligand metrics.","status":"source_checked","source_ids":["boltz-stereochemistry-2025"],"source_locator":"Methods: Benchmark Dataset; Evaluation Metrics; Results: Table 1; cached text lines 8–14, 41–44"},{"label":"Leakage controls","value":"Temporal and similarity-based subsets are examined separately; Before-set scores include structures potentially seen in training.","status":"source_checked","source_ids":["boltz-stereochemistry-2025"],"source_locator":"Methods: Benchmark Dataset; Evaluation Metrics; Results: Table 1; cached text lines 8–14, 41–44"},{"label":"Uncertainty","value":"The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","status":"unreported","source_ids":["boltz-stereochemistry-2025"],"source_locator":"Methods: Benchmark Dataset; Evaluation Metrics; Results: Table 1; cached text lines 8–14, 41–44"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["boltz-stereochemistry-2025"],"source_locator":"Methods: Benchmark Dataset; Evaluation Metrics; Results: Table 1; cached text lines 8–14, 41–44"},{"label":"Organisms","value":"PLINDER and PDBbind subsets are filtered by structure, time and similarity for the broad benchmark. Their curation section does not tabulate species coverage. The separately described human MDM2 case is not the organism scope of the full benchmark.","status":"unreported","source_ids":["boltz-stereochemistry-2025"],"source_locator":"Dataset selection; separate MDM2 case-study setup"},{"label":"Assays","value":"Protein–ligand structures with stereochemical and pose references.","status":"source_checked","source_ids":["boltz-stereochemistry-2025"],"source_locator":"Methods: Benchmark Dataset; Evaluation Metrics; Results: Table 1; cached text lines 8–14, 41–44"},{"label":"Allowed inputs","value":"Complex prediction inputs; restraint-conditioned variants are explicitly separate comparisons.","status":"source_checked","source_ids":["boltz-stereochemistry-2025"],"source_locator":"Methods: Benchmark Dataset; Evaluation Metrics; Results: Table 1; cached text lines 8–14, 41–44"},{"label":"Adaptation","value":"Pretrained structure predictors evaluated across cutoff-defined subsets and restraint variants.","status":"source_checked","source_ids":["boltz-stereochemistry-2025"],"source_locator":"Methods: Benchmark Dataset; Evaluation Metrics; Results: Table 1; cached text lines 8–14, 41–44"}],"strengths":[{"text":"Pose agreement and stereochemical validity are assessed as different criteria.","source_ids":["boltz-stereochemistry-2025"],"source_locator":"Methods: Benchmark Dataset; Evaluation Metrics; Results: Table 1; cached text lines 8–14, 41–44"}],"limitations":[{"text":"Rigid-receptor docking comparators cannot meaningfully supply every protein-structure metric.","source_ids":["boltz-stereochemistry-2025"],"source_locator":"Methods: Benchmark Dataset; Evaluation Metrics; Results: Table 1; cached text lines 8–14, 41–44"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Complex prediction inputs; restraint-conditioned variants are explicitly separate comparisons.","Evaluation: Pretrained structure predictors evaluated across cutoff-defined subsets and restraint variants.","Readout: Protein and ligand RMSD, chirality reproduction, bond-length RMSD and bond-angle RMSD; aggregate definitions vary by metric."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["boltz-stereochemistry-2025"],"source_locator":"Methods: Benchmark Dataset; Evaluation Metrics; Results: Table 1; cached text lines 8–14, 41–44"},"coverage":"limited","gaps":["Uncertainty: The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","Organisms: PLINDER and PDBbind subsets are filtered by structure, time and similarity for the broad benchmark. Their curation section does not tabulate species coverage. The separately described human MDM2 case is not the organism scope of the full benchmark."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Relevant full-paper computational evaluation sections, tables/captions and cited supplementary task passages were reviewed. Reporting omissions are scoped to the inspected sources. Original numerical results are unchanged."}}}}
{"id":"reported-task-c04bb5ee6ecea6","kind":"benchmark","name":"Protein–ligand pose prediction","description":"","status":"needs_review","facets":{"areas":["molecular-interactions"],"tasks":["Protein–ligand pose prediction"]},"source_ids":["boltz1-2025"],"links":[{"relation":"dataset","target_id":"reported-dataset-e45a5a140888ee"}],"attributes":{"entity_level":"task","version":null,"task":"Protein–ligand pose prediction","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Boltz-1 evaluates complex and interface predictions with both confidence-selected and best-of-sample reporting.","summary_source_ids":["boltz1-2025"],"summary_source_locator":"Methods: Validation and test sets curation; Evaluation criteria; cached text lines 24–26, 158–160; comparative evaluation and ablation passages","sections":[{"title":"Evaluation methodology","body":"PDB-derived complexes curated for validation and testing. OpenStructure 2.8.0 computes LDDT-PLI, DockQ and ligand RMSD success. Interface values are averaged within complexes and then across complexes. AlphaFold3 and Chai-1; Boltz-1x is assessed separately for physical-validity constraints.","source_ids":["boltz1-2025"],"source_locator":"Methods: Validation and test sets curation; Evaluation criteria; cached text lines 24–26, 158–160; comparative evaluation and ablation passages; Methods: Validation and test sets curation; Evaluation criteria; cached text lines 24–26, 158–160; comparative evaluation and ablation passages; Methods: Validation and test sets curation; Evaluation criteria; cached text lines 24–26, 158–160; comparative evaluation and ablation passages"}],"facts":[{"label":"Datasets","value":"PDB-derived complexes curated for validation and testing.","status":"source_checked","source_ids":["boltz1-2025"],"source_locator":"Methods: Validation and test sets curation; Evaluation criteria; cached text lines 24–26, 158–160; comparative evaluation and ablation passages"},{"label":"Splits","value":"Validation and test structures are selected by release date and protein/ligand similarity exclusions. Test curation uses a 2023-01-13 cutoff relative to earlier structures; reported comparisons additionally restrict to structures all tested tools can process. CASP15 is evaluated as a separate collection.","status":"source_checked","source_ids":["boltz1-2025"],"source_locator":"Data pipeline: Validation and test sets curation; Evaluation filtering"},{"label":"Metrics","value":"OpenStructure 2.8.0 computes LDDT-PLI, DockQ and ligand RMSD success. Interface values are averaged within complexes and then across complexes.","status":"source_checked","source_ids":["boltz1-2025"],"source_locator":"Methods: Validation and test sets curation; Evaluation criteria; cached text lines 24–26, 158–160; comparative evaluation and ablation passages"},{"label":"Baselines","value":"AlphaFold3 and Chai-1; Boltz-1x is assessed separately for physical-validity constraints.","status":"source_checked","source_ids":["boltz1-2025"],"source_locator":"Methods: Validation and test sets curation; Evaluation criteria; cached text lines 24–26, 158–160; comparative evaluation and ablation passages"},{"label":"Leakage controls","value":"Protein chains are excluded using sequence clusters formed at 40% identity. Ligand filtering requires at least one ligand sufficiently dissimilar to earlier ligands, rather than requiring every ligand to be novel. Test exclusion compares against all structures preceding the test cutoff.","status":"source_checked","source_ids":["boltz1-2025"],"source_locator":"Data pipeline §2.2 Validation and test sets curation, full-XML filter list"},{"label":"Uncertainty","value":"Five samples per method are evaluated as both confidence-selected top-1 and oracle-best predictions; these are selection regimes, not confidence intervals.","status":"source_checked","source_ids":["boltz1-2025"],"source_locator":"Methods: Validation and test sets curation; Evaluation criteria; cached text lines 24–26, 158–160; comparative evaluation and ablation passages"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["boltz1-2025"],"source_locator":"Methods: Validation and test sets curation; Evaluation criteria; cached text lines 24–26, 158–160; comparative evaluation and ablation passages"},{"label":"Organisms","value":"Evaluation spans PDB complexes and eligible CASP15 structures, grouped by molecular/interface type. Data curation and evaluation sections do not provide organism-stratified membership; taxonomy used to pair MSA rows is a separate input-processing property.","status":"unreported","source_ids":["boltz1-2025"],"source_locator":"Data pipeline: structural data and test curation; Evaluation"},{"label":"Assays","value":"Protein–ligand complex structures.","status":"source_checked","source_ids":["boltz1-2025"],"source_locator":"Methods: Validation and test sets curation; Evaluation criteria; cached text lines 24–26, 158–160; comparative evaluation and ablation passages"},{"label":"Allowed inputs","value":"Protein/ligand descriptions for complex-structure prediction.","status":"source_checked","source_ids":["boltz1-2025"],"source_locator":"Methods: Validation and test sets curation; Evaluation criteria; cached text lines 24–26, 158–160; comparative evaluation and ablation passages"},{"label":"Adaptation","value":"Released structure prediction model; task-specific retraining is not established by this evaluation excerpt.","status":"source_checked","source_ids":["boltz1-2025"],"source_locator":"Methods: Validation and test sets curation; Evaluation criteria; cached text lines 24–26, 158–160; comparative evaluation and ablation passages"}],"strengths":[],"limitations":[{"text":"Best-of-five is not equivalent to a deployable top-1 result. The summary does not infer a checkpoint or merge distinct evaluation subsets.","source_ids":["boltz1-2025"],"source_locator":"Methods: Validation and test sets curation; Evaluation criteria; cached text lines 24–26, 158–160; comparative evaluation and ablation passages"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Protein/ligand descriptions for complex-structure prediction.","Evaluation: Released structure prediction model; task-specific retraining is not established by this evaluation excerpt.","Readout: OpenStructure 2.8.0 computes LDDT-PLI, DockQ and ligand RMSD success. Interface values are averaged within complexes and then across complexes."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["boltz1-2025"],"source_locator":"Methods: Validation and test sets curation; Evaluation criteria; cached text lines 24–26, 158–160; comparative evaluation and ablation passages"},"coverage":"limited","gaps":["Organisms: Evaluation spans PDB complexes and eligible CASP15 structures, grouped by molecular/interface type. Data curation and evaluation sections do not provide organism-stratified membership; taxonomy used to pair MSA rows is a separate input-processing property."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Relevant full-paper computational evaluation sections, tables/captions and cited supplementary task passages were reviewed. Reporting omissions are scoped to the inspected sources. Original numerical results are unchanged."}}}}
{"id":"reported-task-c40dac20d9af66","kind":"benchmark","name":"extremely long RNA species classification","description":"","status":"needs_review","facets":{"areas":["rna-transcriptomes"],"tasks":["extremely long RNA species classification"]},"source_ids":["birna-bert-2025"],"links":[{"relation":"dataset","target_id":"reported-dataset-ebc3f5fda43972"}],"attributes":{"entity_level":"task","version":null,"task":"extremely long RNA species classification","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Long-RNA species classification evaluates representation quality on an RNAcentral-derived seven-species dataset.","summary_source_ids":["birna-bert-2025"],"summary_source_locator":"Results: BiRNA-BERT significantly outperforms in extremely long sequence task; Table 2; cached text lines 37–40","sections":[{"title":"Evaluation methodology","body":"Long noncoding RNA sequences from RNAcentral with species labels. The long-sequence species-classification Results paragraph specifies the dataset and F1 comparisons but does not give a train/validation/test assignment rule. F1 score; the reviewed task description does not establish the averaging convention. RNA-FM and RiNALMo are compared under their sequence-length constraints. The species-classification paragraph does not specify RNAcentral overlap exclusion between its benchmark sequences and model pretraining. Structure-task deduplication elsewhere is not this task.","source_ids":["birna-bert-2025"],"source_locator":"Results: BiRNA-BERT significantly outperforms in extremely long sequence task; Table 2; cached text lines 37–40"}],"facts":[{"label":"Datasets","value":"Long noncoding RNA sequences from RNAcentral with species labels.","status":"source_checked","source_ids":["birna-bert-2025"],"source_locator":"Results: BiRNA-BERT significantly outperforms in extremely long sequence task; Table 2; cached text lines 37–40"},{"label":"Splits","value":"The long-sequence species-classification Results paragraph specifies the dataset and F1 comparisons but does not give a train/validation/test assignment rule.","status":"unreported","source_ids":["birna-bert-2025"],"source_locator":"Results: BiRNA-BERT significantly outperforms in extremely long sequence task; Table 2; cached text lines 37–40"},{"label":"Metrics","value":"F1 score; the reviewed task description does not establish the averaging convention.","status":"source_checked","source_ids":["birna-bert-2025"],"source_locator":"Results: BiRNA-BERT significantly outperforms in extremely long sequence task; Table 2; cached text lines 37–40"},{"label":"Baselines","value":"RNA-FM and RiNALMo are compared under their sequence-length constraints.","status":"source_checked","source_ids":["birna-bert-2025"],"source_locator":"Results: BiRNA-BERT significantly outperforms in extremely long sequence task; Table 2; cached text lines 37–40"},{"label":"Leakage controls","value":"The species-classification paragraph does not specify RNAcentral overlap exclusion between its benchmark sequences and model pretraining. Structure-task deduplication elsewhere is not this task.","status":"unreported","source_ids":["birna-bert-2025"],"source_locator":"Results: BiRNA-BERT significantly outperforms in extremely long sequence task; Table 2; cached text lines 37–40"},{"label":"Uncertainty","value":"Table 2 reports one F1 score per model for long-sequence species classification. The corresponding main-text section and Supplementary Information do not define repeated runs, confidence intervals or a statistical comparison for this particular task.","status":"unreported","source_ids":["birna-bert-2025","evidence-task-final-b-birna-supplement"],"source_locator":"Results: extremely long sequence task and Table 2; Supplementary Information §§1–3"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["birna-bert-2025"],"source_locator":"Results: BiRNA-BERT significantly outperforms in extremely long sequence task; Table 2; cached text lines 37–40"},{"label":"Organisms","value":"Bos taurus, Gallus gallus, Gorilla gorilla, Homo sapiens, Mus musculus, Pan troglodytes and Rattus norvegicus.","status":"source_checked","source_ids":["birna-bert-2025"],"source_locator":"Results: BiRNA-BERT significantly outperforms in extremely long sequence task; Table 2; cached text lines 37–40"},{"label":"Assays","value":"RNAcentral sequence/species annotations.","status":"source_checked","source_ids":["birna-bert-2025"],"source_locator":"Results: BiRNA-BERT significantly outperforms in extremely long sequence task; Table 2; cached text lines 37–40"},{"label":"Allowed inputs","value":"Long RNA sequences.","status":"source_checked","source_ids":["birna-bert-2025"],"source_locator":"Results: BiRNA-BERT significantly outperforms in extremely long sequence task; Table 2; cached text lines 37–40"},{"label":"Adaptation","value":"The long-sequence benchmark compares BiRNA-BERT, RiNALMo and RNA-FM, with comparator truncation stated. Its main-text description and Supplementary Information do not specify the classification head or whether each encoder is frozen or fine-tuned for this species task.","status":"unreported","source_ids":["birna-bert-2025","evidence-task-final-b-birna-supplement"],"source_locator":"Results: extremely long sequence task, Table 2; Supplementary Information §§1–3"}],"strengths":[],"limitations":[{"text":"Comparator truncation is part of the published species-classification setting. Its task section and supplement do not establish split membership, related-sequence exclusion or task-specific fitting details.","source_ids":["birna-bert-2025","evidence-task-final-b-birna-supplement"],"source_locator":"Results: BiRNA-BERT significantly outperforms in extremely long sequence task; Table 2; cached text lines 37–40 ; Supplementary Information task sections"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Long RNA sequences.","Task: Long-RNA species classification evaluates representation quality on an RNAcentral-derived seven-species dataset."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["birna-bert-2025"],"source_locator":"Results: BiRNA-BERT significantly outperforms in extremely long sequence task; Table 2; cached text lines 37–40"},"coverage":"limited","gaps":["Splits: The long-sequence species-classification Results paragraph specifies the dataset and F1 comparisons but does not give a train/validation/test assignment rule.","Leakage controls: The species-classification paragraph does not specify RNAcentral overlap exclusion between its benchmark sequences and model pretraining. Structure-task deduplication elsewhere is not this task.","Uncertainty: Table 2 reports one F1 score per model for long-sequence species classification. The corresponding main-text section and Supplementary Information do not define repeated runs, confidence intervals or a statistical comparison for this particular task.","Adaptation: The long-sequence benchmark compares BiRNA-BERT, RiNALMo and RNA-FM, with comparator truncation stated. Its main-text description and Supplementary Information do not specify the classification head or whether each encoder is frozen or fine-tuned for this species task."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Relevant full-paper computational evaluation sections, tables/captions and cited supplementary task passages were reviewed. Reporting omissions are scoped to the inspected sources. Original numerical results are unchanged."}}}}
{"id":"reported-task-c4a578065f44b2","kind":"benchmark","name":"protein function annotation","description":"","status":"needs_review","facets":{"areas":["proteins-complexes"],"tasks":["protein function annotation"]},"source_ids":["spin-protein-function-2026"],"links":[{"relation":"dataset","target_id":"reported-dataset-dba1707164d296"}],"attributes":{"entity_level":"task","version":null,"task":"protein function annotation","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Protein-function annotation is evaluated within selected domain-containing protein families.","summary_source_ids":["spin-protein-function-2026"],"summary_source_locator":"Methods §§2.6, 2.9; cached text lines 38–40, 48–50; comparative evaluation and ablation passages; Results embedding comparison and supervised-classifier comparison; uncertainty/repeat-run/statistical-comparison passages","sections":[{"title":"Evaluation methodology","body":"CPF, SH3 and TRX family collections from UniProtKB/TrEMBL, with additional fungal SH3 sequences from NCBI. The source defines a class-weighted macro F1 variant and weighted accuracy; these should not be assumed identical to ordinary macro F1. Classical SVM/gradient-boosting/random-forest/KNN and neural CNN/LSTM/Transformer references; protein-model backbone comparisons form a separate TRX ablation. Sequences are clustered to reduce high similarity before subsequent filtering. Independent training seeds quantify initialization variation; the domain-span bootstrap is a separate uncertainty analysis.","source_ids":["spin-protein-function-2026"],"source_locator":"Methods §§2.6, 2.9; cached text lines 38–40, 48–50; comparative evaluation and ablation passages; Results embedding comparison and supervised-classifier comparison; uncertainty/repeat-run/statistical-comparison passages"}],"facts":[{"label":"Datasets","value":"CPF, SH3 and TRX family collections from UniProtKB/TrEMBL, with additional fungal SH3 sequences from NCBI.","status":"source_checked","source_ids":["spin-protein-function-2026"],"source_locator":"Methods §§2.6, 2.9; cached text lines 38–40, 48–50; comparative evaluation and ablation passages; Results embedding comparison and supervised-classifier comparison; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Splits","value":"Each protein-family dataset has 80% training, 10% validation and 10% test data in Table 2. Three random initializations repeat training; the TRX training-size experiment keeps the validation and test sets fixed. These percentages do not establish a homology-disjoint split.","status":"source_checked","source_ids":["spin-protein-function-2026"],"source_locator":"Table 2 header/footnote; §2.3 Family-specific model tuning; §3.3 and Table 3"},{"label":"Metrics","value":"The source defines a class-weighted macro F1 variant and weighted accuracy; these should not be assumed identical to ordinary macro F1.","status":"source_checked","source_ids":["spin-protein-function-2026"],"source_locator":"Methods §§2.6, 2.9; cached text lines 38–40, 48–50; comparative evaluation and ablation passages; Results embedding comparison and supervised-classifier comparison; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Baselines","value":"Classical SVM/gradient-boosting/random-forest/KNN and neural CNN/LSTM/Transformer references; protein-model backbone comparisons form a separate TRX ablation.","status":"source_checked","source_ids":["spin-protein-function-2026"],"source_locator":"Methods §§2.6, 2.9; cached text lines 38–40, 48–50; comparative evaluation and ablation passages; Results embedding comparison and supervised-classifier comparison; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Leakage controls","value":"Sequences are clustered to reduce high similarity before subsequent filtering.","status":"source_checked","source_ids":["spin-protein-function-2026"],"source_locator":"Methods §§2.6, 2.9; cached text lines 38–40, 48–50; comparative evaluation and ablation passages; Results embedding comparison and supervised-classifier comparison; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Uncertainty","value":"Independent training seeds quantify initialization variation; the domain-span bootstrap is a separate uncertainty analysis.","status":"source_checked","source_ids":["spin-protein-function-2026"],"source_locator":"Methods §§2.6, 2.9; cached text lines 38–40, 48–50; comparative evaluation and ablation passages; Results embedding comparison and supervised-classifier comparison; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["spin-protein-function-2026"],"source_locator":"Methods §§2.6, 2.9; cached text lines 38–40, 48–50; comparative evaluation and ablation passages; Results embedding comparison and supervised-classifier comparison; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Organisms","value":"Protein families CPF, SH3 and TRX; fungal SH3 sequences form an additional collection.","status":"source_checked","source_ids":["spin-protein-function-2026"],"source_locator":"Methods §§2.6, 2.9; cached text lines 38–40, 48–50; comparative evaluation and ablation passages; Results embedding comparison and supervised-classifier comparison; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Assays","value":"UniProtKB/TrEMBL family/function annotations.","status":"source_checked","source_ids":["spin-protein-function-2026"],"source_locator":"Methods §§2.6, 2.9; cached text lines 38–40, 48–50; comparative evaluation and ablation passages; Results embedding comparison and supervised-classifier comparison; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Allowed inputs","value":"Protein sequence representations.","status":"source_checked","source_ids":["spin-protein-function-2026"],"source_locator":"Methods §§2.6, 2.9; cached text lines 38–40, 48–50; comparative evaluation and ablation passages; Results embedding comparison and supervised-classifier comparison; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Adaptation","value":"ESM2-35M fine-tuning with the Domain Span prediction module; the frozen-backbone configuration is a separate ablation.","status":"source_checked","source_ids":["spin-protein-function-2026"],"source_locator":"Methods §§2.6, 2.9; cached text lines 38–40, 48–50; comparative evaluation and ablation passages; Results embedding comparison and supervised-classifier comparison; uncertainty/repeat-run/statistical-comparison passages"}],"strengths":[],"limitations":[{"text":"The 80/10/10 within-family split and repeated random initializations do not establish sequence-homology exclusion. Domain membership alone does not define the functional classes.","source_ids":["spin-protein-function-2026"],"source_locator":"Methods §§2.6, 2.9; cached text lines 38–40, 48–50; comparative evaluation and ablation passages; Results embedding comparison and supervised-classifier comparison; uncertainty/repeat-run/statistical-comparison passages"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Protein sequence representations.","Evaluation: ESM2-35M fine-tuning with the Domain Span prediction module; the frozen-backbone configuration is a separate ablation.","Readout: The source defines a class-weighted macro F1 variant and weighted accuracy; these should not be assumed identical to ordinary macro F1."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["spin-protein-function-2026"],"source_locator":"Methods §§2.6, 2.9; cached text lines 38–40, 48–50; comparative evaluation and ablation passages; Results embedding comparison and supervised-classifier comparison; uncertainty/repeat-run/statistical-comparison passages"},"coverage":"limited","gaps":["The 80/10/10 within-family split and repeated random initializations do not establish sequence-homology exclusion. Domain membership alone does not define the functional classes."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Relevant full-paper computational evaluation sections, tables/captions and cited supplementary task passages were reviewed. Reporting omissions are scoped to the inspected sources. Original numerical results are unchanged."}}}}
{"id":"reported-task-c7a8a372f77886","kind":"benchmark","name":"protein variant fitness prediction","description":"","status":"needs_review","facets":{"areas":["proteins-complexes"],"tasks":["protein variant fitness prediction"]},"source_ids":["esm2-ofs-fitness-2025"],"links":[{"relation":"dataset","target_id":"reported-dataset-9c186c8f4ed3f4"}],"attributes":{"entity_level":"task","version":null,"task":"protein variant fitness prediction","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"ProteinGym substitution assays evaluate agreement between a fixed protein-model scoring method and measured variant effects.","summary_source_ids":["esm2-ofs-fitness-2025"],"summary_source_locator":"ProteinGym evaluation section and Table I; PDF text lines 346–369 and Table I caption","sections":[{"title":"Evaluation methodology","body":"Experimental DMS variant-effect measurements across activity, binding, expression, organismal fitness and stability categories. The downstream assay evaluation is described as zero-shot; a separate approximation-model training partition must not be treated as the assay test split. Spearman rank correlation compares measured variant effects with model scores; Table I groups substitution assays by functional category. CARP, ESM-1v, Progen2 M, RITA L and Tranception L without retrieval use baseline data sourced from ProteinGym; ESM2 scoring variants are also compared. The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","source_ids":["esm2-ofs-fitness-2025"],"source_locator":"ProteinGym evaluation section and Table I; PDF text lines 346–369 and Table I caption"}],"facts":[{"label":"Datasets","value":"Experimental DMS variant-effect measurements across activity, binding, expression, organismal fitness and stability categories.","status":"source_checked","source_ids":["esm2-ofs-fitness-2025"],"source_locator":"ProteinGym evaluation section and Table I; PDF text lines 346–369 and Table I caption"},{"label":"Splits","value":"The downstream assay evaluation is described as zero-shot; a separate approximation-model training partition must not be treated as the assay test split.","status":"source_checked","source_ids":["esm2-ofs-fitness-2025"],"source_locator":"ProteinGym evaluation section and Table I; PDF text lines 346–369 and Table I caption"},{"label":"Metrics","value":"Spearman rank correlation compares measured variant effects with model scores; Table I groups substitution assays by functional category.","status":"source_checked","source_ids":["esm2-ofs-fitness-2025"],"source_locator":"ProteinGym evaluation section and Table I; PDF text lines 346–369 and Table I caption"},{"label":"Baselines","value":"CARP, ESM-1v, Progen2 M, RITA L and Tranception L without retrieval use baseline data sourced from ProteinGym; ESM2 scoring variants are also compared.","status":"source_checked","source_ids":["esm2-ofs-fitness-2025"],"source_locator":"ProteinGym evaluation section and Table I; PDF text lines 346–369 and Table I caption"},{"label":"Leakage controls","value":"Training of the OFS approximation model withholds sequences sharing more than 50% identity with ProteinGym reference sequences. This is an explicit control on the approximation-training corpus; it is not a claim that the original ESM-2 pretraining corpus excludes those proteins.","status":"source_checked","source_ids":["esm2-ofs-fitness-2025"],"source_locator":"PRX Life paper, Methods: OFS model training; extracted PDF lines 918–934"},{"label":"Uncertainty","value":"The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","status":"unreported","source_ids":["esm2-ofs-fitness-2025"],"source_locator":"ProteinGym evaluation section and Table I; PDF text lines 346–369 and Table I caption"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["esm2-ofs-fitness-2025"],"source_locator":"ProteinGym evaluation section and Table I; PDF text lines 346–369 and Table I caption"},{"label":"Organisms","value":"ProteinGym assay-specific taxa.","status":"source_checked","source_ids":["esm2-ofs-fitness-2025"],"source_locator":"ProteinGym evaluation section and Table I; PDF text lines 346–369 and Table I caption"},{"label":"Assays","value":"DMS substitution measurements across functional categories.","status":"source_checked","source_ids":["esm2-ofs-fitness-2025"],"source_locator":"ProteinGym evaluation section and Table I; PDF text lines 346–369 and Table I caption"},{"label":"Allowed inputs","value":"Variant protein sequences for model scoring.","status":"source_checked","source_ids":["esm2-ofs-fitness-2025"],"source_locator":"ProteinGym evaluation section and Table I; PDF text lines 346–369 and Table I caption"},{"label":"Adaptation","value":"Zero-shot downstream assay evaluation; OFS approximation-model fitting is separate.","status":"source_checked","source_ids":["esm2-ofs-fitness-2025"],"source_locator":"ProteinGym evaluation section and Table I; PDF text lines 346–369 and Table I caption"}],"strengths":[],"limitations":[{"text":"This record concerns substitution assays, separate from the indel and other demonstrations. The OFS-training identity filter does not certify that the ESM-2 pretraining corpus is independent of the benchmark.","source_ids":["esm2-ofs-fitness-2025"],"source_locator":"ProteinGym evaluation section and Table I; PDF text lines 346–369 and Table I caption"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Variant protein sequences for model scoring.","Evaluation: Zero-shot downstream assay evaluation; OFS approximation-model fitting is separate.","Readout: Spearman rank correlation compares measured variant effects with model scores; Table I groups substitution assays by functional category."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["esm2-ofs-fitness-2025"],"source_locator":"ProteinGym evaluation section and Table I; PDF text lines 346–369 and Table I caption"},"coverage":"limited","gaps":["Uncertainty: The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Relevant full-paper computational evaluation sections, tables/captions and cited supplementary task passages were reviewed. Reporting omissions are scoped to the inspected sources. Original numerical results are unchanged."}}}}
{"id":"reported-task-c7ce06b753b8b6","kind":"benchmark","name":"non-coding RNA pairwise interaction prediction","description":"","status":"needs_review","facets":{"areas":["rna-transcriptomes"],"tasks":["non-coding RNA pairwise interaction prediction"]},"source_ids":["cupid-rna-interactions-2026"],"links":[{"relation":"dataset","target_id":"reported-dataset-32ccef507a1dd7"}],"attributes":{"entity_level":"task","version":null,"task":"non-coding RNA pairwise interaction prediction","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Noncoding RNA-pair classification evaluates a strongly imbalanced interaction dataset with validation-tuned decision thresholds.","summary_source_ids":["cupid-rna-interactions-2026"],"summary_source_locator":"Methods: Dataset; Experimental evaluation: Data preparation and splitting; Evaluation metrics; cached text lines 15–17, 66–75; comparative evaluation and ablation passages","sections":[{"title":"Evaluation methodology","body":"RNA-KG-derived multispecies interaction pairs; sampled negatives preserve pair-type frequencies with a 20:1 negative-to-positive ratio. Stratified 90:10 train/test partition; a further validation subset is drawn from training data. Accuracy, balanced accuracy, precision, recall, F1, AUROC and AUPRC, both overall and by interacting pair type. Random classifier, IntaRNA and CUPID pooling/data-augmentation ablations. The threshold is selected by validation MCC; sequence-disjoint separation of interacting entities remains unextracted. The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","source_ids":["cupid-rna-interactions-2026"],"source_locator":"Methods: Dataset; Experimental evaluation: Data preparation and splitting; Evaluation metrics; cached text lines 15–17, 66–75; comparative evaluation and ablation passages"}],"facts":[{"label":"Datasets","value":"RNA-KG-derived multispecies interaction pairs; sampled negatives preserve pair-type frequencies with a 20:1 negative-to-positive ratio.","status":"source_checked","source_ids":["cupid-rna-interactions-2026"],"source_locator":"Methods: Dataset; Experimental evaluation: Data preparation and splitting; Evaluation metrics; cached text lines 15–17, 66–75; comparative evaluation and ablation passages"},{"label":"Splits","value":"Stratified 90:10 train/test partition; a further validation subset is drawn from training data.","status":"source_checked","source_ids":["cupid-rna-interactions-2026"],"source_locator":"Methods: Dataset; Experimental evaluation: Data preparation and splitting; Evaluation metrics; cached text lines 15–17, 66–75; comparative evaluation and ablation passages"},{"label":"Metrics","value":"Accuracy, balanced accuracy, precision, recall, F1, AUROC and AUPRC, both overall and by interacting pair type.","status":"source_checked","source_ids":["cupid-rna-interactions-2026"],"source_locator":"Methods: Dataset; Experimental evaluation: Data preparation and splitting; Evaluation metrics; cached text lines 15–17, 66–75; comparative evaluation and ablation passages"},{"label":"Baselines","value":"Random classifier, IntaRNA and CUPID pooling/data-augmentation ablations.","status":"source_checked","source_ids":["cupid-rna-interactions-2026"],"source_locator":"Methods: Dataset; Experimental evaluation: Data preparation and splitting; Evaluation metrics; cached text lines 15–17, 66–75; comparative evaluation and ablation passages"},{"label":"Leakage controls","value":"The threshold is selected by validation MCC; sequence-disjoint separation of interacting entities remains unextracted.","status":"source_checked","source_ids":["cupid-rna-interactions-2026"],"source_locator":"Methods: Dataset; Experimental evaluation: Data preparation and splitting; Evaluation metrics; cached text lines 15–17, 66–75; comparative evaluation and ablation passages"},{"label":"Uncertainty","value":"The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","status":"unreported","source_ids":["cupid-rna-interactions-2026"],"source_locator":"Methods: Dataset; Experimental evaluation: Data preparation and splitting; Evaluation metrics; cached text lines 15–17, 66–75; comparative evaluation and ablation passages"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["cupid-rna-interactions-2026"],"source_locator":"Methods: Dataset; Experimental evaluation: Data preparation and splitting; Evaluation metrics; cached text lines 15–17, 66–75; comparative evaluation and ablation passages"},{"label":"Organisms","value":"Multiple species represented in RNA-KG.","status":"source_checked","source_ids":["cupid-rna-interactions-2026"],"source_locator":"Methods: Dataset; Experimental evaluation: Data preparation and splitting; Evaluation metrics; cached text lines 15–17, 66–75; comparative evaluation and ablation passages"},{"label":"Assays","value":"RNA–RNA interaction annotations with sampled negative pairs.","status":"source_checked","source_ids":["cupid-rna-interactions-2026"],"source_locator":"Methods: Dataset; Experimental evaluation: Data preparation and splitting; Evaluation metrics; cached text lines 15–17, 66–75; comparative evaluation and ablation passages"},{"label":"Allowed inputs","value":"Pairs of noncoding RNA sequences.","status":"source_checked","source_ids":["cupid-rna-interactions-2026"],"source_locator":"Methods: Dataset; Experimental evaluation: Data preparation and splitting; Evaluation metrics; cached text lines 15–17, 66–75; comparative evaluation and ablation passages"},{"label":"Adaptation","value":"Supervised pair classification using a stratified train/test partition and training-derived validation subset.","status":"source_checked","source_ids":["cupid-rna-interactions-2026"],"source_locator":"Methods: Dataset; Experimental evaluation: Data preparation and splitting; Evaluation metrics; cached text lines 15–17, 66–75; comparative evaluation and ablation passages"}],"strengths":[{"text":"A random classifier and explicit pooling/augmentation ablations isolate different sources of performance.","source_ids":["cupid-rna-interactions-2026"],"source_locator":"Methods: Dataset; Experimental evaluation: Data preparation and splitting; Evaluation metrics; cached text lines 15–17, 66–75; comparative evaluation and ablation passages"}],"limitations":[{"text":"A pair-level split does not by itself establish transfer to unseen RNAs. AUPRC depends on the sampled class prevalence.","source_ids":["cupid-rna-interactions-2026"],"source_locator":"Methods: Dataset; Experimental evaluation: Data preparation and splitting; Evaluation metrics; cached text lines 15–17, 66–75; comparative evaluation and ablation passages"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Pairs of noncoding RNA sequences.","Evaluation: Supervised pair classification using a stratified train/test partition and training-derived validation subset.","Readout: Accuracy, balanced accuracy, precision, recall, F1, AUROC and AUPRC, both overall and by interacting pair type."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["cupid-rna-interactions-2026"],"source_locator":"Methods: Dataset; Experimental evaluation: Data preparation and splitting; Evaluation metrics; cached text lines 15–17, 66–75; comparative evaluation and ablation passages"},"coverage":"limited","gaps":["A pair-level split does not by itself establish transfer to unseen RNAs. AUPRC depends on the sampled class prevalence."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Task-specific computational methodology and field context checked in the cited primary-source artifact. Source-backed fields, inapplicable evaluator dimensions and unresolved details are distinguished. Numerical results were not reproduced."}}}}
{"id":"reported-task-c98e91ffc7247d","kind":"benchmark","name":"CATH superfamily annotation","description":"","status":"needs_review","facets":{"areas":["proteins-complexes"],"tasks":["CATH superfamily annotation"]},"source_ids":["cathe2-2025"],"links":[{"relation":"dataset","target_id":"reported-dataset-6e0c28dfde7337"}],"attributes":{"entity_level":"task","version":null,"task":"CATH superfamily annotation","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"CATH superfamily annotation evaluates protein-domain classification with stringent sequence-identity separation.","summary_source_ids":["cathe2-2025"],"summary_source_locator":"Methods: Data and datasets; Results; cached text lines 11–14, 30; comparative evaluation and ablation passages","sections":[{"title":"Evaluation methodology","body":"CATHe-derived large and small superfamily datasets based on CATH v4.3 and CATH-Gene3D. Training, validation and test domain sets are retained from CATHe where structural-token availability permits. F1 is emphasized for class imbalance; balanced accuracy and MCC are also computed. CATHe and alternative embeddings from ProtT5, ProstT5, ESM2, Ankh and TM-Vec; domain availability and input dimensionality differ. The source reports less than 20% sequence identity within and between the original partitions. Bootstrap resampling of the test set supplies interval estimates.","source_ids":["cathe2-2025"],"source_locator":"Methods: Data and datasets; Results; cached text lines 11–14, 30; comparative evaluation and ablation passages"}],"facts":[{"label":"Datasets","value":"CATHe-derived large and small superfamily datasets based on CATH v4.3 and CATH-Gene3D.","status":"source_checked","source_ids":["cathe2-2025"],"source_locator":"Methods: Data and datasets; Results; cached text lines 11–14, 30; comparative evaluation and ablation passages"},{"label":"Splits","value":"Training, validation and test domain sets are retained from CATHe where structural-token availability permits.","status":"source_checked","source_ids":["cathe2-2025"],"source_locator":"Methods: Data and datasets; Results; cached text lines 11–14, 30; comparative evaluation and ablation passages"},{"label":"Metrics","value":"F1 is emphasized for class imbalance; balanced accuracy and MCC are also computed.","status":"source_checked","source_ids":["cathe2-2025"],"source_locator":"Methods: Data and datasets; Results; cached text lines 11–14, 30; comparative evaluation and ablation passages"},{"label":"Baselines","value":"CATHe and alternative embeddings from ProtT5, ProstT5, ESM2, Ankh and TM-Vec; domain availability and input dimensionality differ.","status":"source_checked","source_ids":["cathe2-2025"],"source_locator":"Methods: Data and datasets; Results; cached text lines 11–14, 30; comparative evaluation and ablation passages"},{"label":"Leakage controls","value":"The source reports less than 20% sequence identity within and between the original partitions.","status":"source_checked","source_ids":["cathe2-2025"],"source_locator":"Methods: Data and datasets; Results; cached text lines 11–14, 30; comparative evaluation and ablation passages"},{"label":"Uncertainty","value":"Bootstrap resampling of the test set supplies interval estimates.","status":"source_checked","source_ids":["cathe2-2025"],"source_locator":"Methods: Data and datasets; Results; cached text lines 11–14, 30; comparative evaluation and ablation passages"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["cathe2-2025"],"source_locator":"Methods: Data and datasets; Results; cached text lines 11–14, 30; comparative evaluation and ablation passages"},{"label":"Organisms","value":"The benchmark spans CATH superfamilies using PDB and AlphaFold-database domains. Data and datasets specifies structural-domain and superfamily coverage, but does not provide a species inventory; superfamily classes are not organism labels.","status":"unreported","source_ids":["cathe2-2025"],"source_locator":"Data and datasets; 3Di processing"},{"label":"Assays","value":"CATH domain superfamily annotations and structural-alphabet representations.","status":"source_checked","source_ids":["cathe2-2025"],"source_locator":"Methods: Data and datasets; Results; cached text lines 11–14, 30; comparative evaluation and ablation passages"},{"label":"Allowed inputs","value":"Amino-acid embeddings, 3Di embeddings, or their concatenation.","status":"source_checked","source_ids":["cathe2-2025"],"source_locator":"Methods: Data and datasets; Results; cached text lines 11–14, 30; comparative evaluation and ablation passages"},{"label":"Adaptation","value":"Frozen protein-model embeddings feed a supervised feed-forward classifier; classifier/data-filter hyperparameters are selected in the study.","status":"source_checked","source_ids":["cathe2-2025"],"source_locator":"Methods: Data and datasets; Results; cached text lines 11–14, 30; comparative evaluation and ablation passages"}],"strengths":[{"text":"The source documents training-domain losses caused by structural-token availability, exposing a key comparison confound.","source_ids":["cathe2-2025"],"source_locator":"Methods: Data and datasets; Results; cached text lines 11–14, 30; comparative evaluation and ablation passages"}],"limitations":[{"text":"Unavailable structural inputs remove domains and classes from the larger dataset; comparison with the original CATHe corpus must preserve that change in evaluation coverage.","source_ids":["cathe2-2025"],"source_locator":"Methods: Data and datasets; Results; cached text lines 11–14, 30; comparative evaluation and ablation passages"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Amino-acid embeddings, 3Di embeddings, or their concatenation.","Evaluation: Frozen protein-model embeddings feed a supervised feed-forward classifier; classifier/data-filter hyperparameters are selected in the study.","Readout: F1 is emphasized for class imbalance; balanced accuracy and MCC are also computed."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["cathe2-2025"],"source_locator":"Methods: Data and datasets; Results; cached text lines 11–14, 30; comparative evaluation and ablation passages"},"coverage":"limited","gaps":["Organisms: The benchmark spans CATH superfamilies using PDB and AlphaFold-database domains. Data and datasets specifies structural-domain and superfamily coverage, but does not provide a species inventory; superfamily classes are not organism labels."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Relevant full-paper computational evaluation sections, tables/captions and cited supplementary task passages were reviewed. Reporting omissions are scoped to the inspected sources. Original numerical results are unchanged."}}}}
{"id":"reported-task-c9d2a6435979e9","kind":"benchmark","name":"G-quadruplex classification","description":"","status":"needs_review","facets":{"areas":["dna-genomes"],"tasks":["G-quadruplex classification"]},"source_ids":["quadruplex-llm-benchmark-2025"],"links":[{"relation":"dataset","target_id":"reported-dataset-9e9d18bc5bfb8b"}],"attributes":{"entity_level":"task","version":null,"task":"G-quadruplex classification","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"G-quadruplex classification evaluates balanced positives and sampled genomic-background negatives from several annotation assays.","summary_source_ids":["quadruplex-llm-benchmark-2025"],"summary_source_locator":"Methods: Data preparation; Metrics of evaluation; cached text lines 13–16, 22–23; comparative evaluation and ablation passages","sections":[{"title":"Evaluation methodology","body":"KEx, G4 ChIP-seq, G4-seq and G4 CUT&Tag annotation collections. Accuracy, ROC-AUC, F1 and MCC. DNABERT, DNABERT-2, GENA-LM, HyenaDNA and Caduceus; tested context lengths differ.","source_ids":["quadruplex-llm-benchmark-2025"],"source_locator":"Methods: Data preparation; Metrics of evaluation; cached text lines 13–16, 22–23; comparative evaluation and ablation passages; Methods: Data preparation; Metrics of evaluation; cached text lines 13–16, 22–23; comparative evaluation and ablation passages; Methods: Data preparation; Metrics of evaluation; cached text lines 13–16, 22–23; comparative evaluation and ablation passages"}],"facts":[{"label":"Datasets","value":"KEx, G4 ChIP-seq, G4-seq and G4 CUT&Tag annotation collections.","status":"source_checked","source_ids":["quadruplex-llm-benchmark-2025"],"source_locator":"Methods: Data preparation; Metrics of evaluation; cached text lines 13–16, 22–23; comparative evaluation and ablation passages"},{"label":"Splits","value":"Five-fold cross-validation.","status":"source_checked","source_ids":["quadruplex-llm-benchmark-2025"],"source_locator":"Methods: Data preparation; Metrics of evaluation; cached text lines 13–16, 22–23; comparative evaluation and ablation passages"},{"label":"Metrics","value":"Accuracy, ROC-AUC, F1 and MCC.","status":"source_checked","source_ids":["quadruplex-llm-benchmark-2025"],"source_locator":"Methods: Data preparation; Metrics of evaluation; cached text lines 13–16, 22–23; comparative evaluation and ablation passages"},{"label":"Baselines","value":"DNABERT, DNABERT-2, GENA-LM, HyenaDNA and Caduceus; tested context lengths differ.","status":"source_checked","source_ids":["quadruplex-llm-benchmark-2025"],"source_locator":"Methods: Data preparation; Metrics of evaluation; cached text lines 13–16, 22–23; comparative evaluation and ablation passages"},{"label":"Leakage controls","value":"Sampled negative regions exclude annotated G-quadruplex positives. Data preparation does not specify a chromosome-disjoint or homology-disjoint partition, so non-overlapping positive/negative labels alone do not establish genomic independence.","status":"source_checked","source_ids":["quadruplex-llm-benchmark-2025"],"source_locator":"Materials and methods: Data preparation and Metrics of evaluation"},{"label":"Uncertainty","value":"Mean and standard deviation across the five folds.","status":"source_checked","source_ids":["quadruplex-llm-benchmark-2025"],"source_locator":"Methods: Data preparation; Metrics of evaluation; cached text lines 13–16, 22–23; comparative evaluation and ablation passages"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["quadruplex-llm-benchmark-2025"],"source_locator":"Methods: Data preparation; Metrics of evaluation; cached text lines 13–16, 22–23; comparative evaluation and ablation passages"},{"label":"Organisms","value":"The paper reuses KEx, G4 ChIP-seq, G4-seq and G4 CUT&Tag collections. Data preparation lists assays and sample counts without dataset-by-dataset organism/assembly identifiers. Its references include both mammalian and multispecies studies, so a single genome cannot be assigned to all four from this description.","status":"unreported","source_ids":["quadruplex-llm-benchmark-2025"],"source_locator":"Materials and methods: Data preparation, Table 1; primary dataset references 14–17,24"},{"label":"Assays","value":"G4 ChIP-seq, G4-seq, CUT&Tag and KEx annotations.","status":"source_checked","source_ids":["quadruplex-llm-benchmark-2025"],"source_locator":"Methods: Data preparation; Metrics of evaluation; cached text lines 13–16, 22–23; comparative evaluation and ablation passages"},{"label":"Allowed inputs","value":"DNA sequences around candidate G-quadruplex regions.","status":"source_checked","source_ids":["quadruplex-llm-benchmark-2025"],"source_locator":"Methods: Data preparation; Metrics of evaluation; cached text lines 13–16, 22–23; comparative evaluation and ablation passages"},{"label":"Adaptation","value":"Supervised classification with five-fold cross-validation.","status":"source_checked","source_ids":["quadruplex-llm-benchmark-2025"],"source_locator":"Methods: Data preparation; Metrics of evaluation; cached text lines 13–16, 22–23; comparative evaluation and ablation passages"}],"strengths":[],"limitations":[{"text":"Balanced sampled negatives do not reproduce genome-wide prevalence. Fold variability is not the same as an independent cohort confidence interval.","source_ids":["quadruplex-llm-benchmark-2025"],"source_locator":"Methods: Data preparation; Metrics of evaluation; cached text lines 13–16, 22–23; comparative evaluation and ablation passages"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: DNA sequences around candidate G-quadruplex regions.","Evaluation: Supervised classification with five-fold cross-validation.","Readout: Accuracy, ROC-AUC, F1 and MCC."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["quadruplex-llm-benchmark-2025"],"source_locator":"Methods: Data preparation; Metrics of evaluation; cached text lines 13–16, 22–23; comparative evaluation and ablation passages"},"coverage":"limited","gaps":["Organisms: The paper reuses KEx, G4 ChIP-seq, G4-seq and G4 CUT&Tag collections. Data preparation lists assays and sample counts without dataset-by-dataset organism/assembly identifiers. Its references include both mammalian and multispecies studies, so a single genome cannot be assigned to all four from this description."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Relevant full-paper computational evaluation sections, tables/captions and cited supplementary task passages were reviewed. Reporting omissions are scoped to the inspected sources. Original numerical results are unchanged."}}}}
{"id":"reported-task-cd127e56fb1f04","kind":"benchmark","name":"regulatory sequence classification","description":"","status":"needs_review","facets":{"areas":["dna-genomes"],"tasks":["regulatory sequence classification"]},"source_ids":["genomic-tokenizer-selection-2025"],"links":[{"relation":"dataset","target_id":"reported-dataset-0bba1c9a7ae410"}],"attributes":{"entity_level":"task","version":null,"task":"regulatory sequence classification","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Regulatory-sequence classification compares genomic models and tokenizers across established benchmark collections.","summary_source_ids":["genomic-tokenizer-selection-2025"],"summary_source_locator":"Methods §2.2 Benchmarks; Results §3.1; cached text lines 20–21, 43–44","sections":[{"title":"Evaluation methodology","body":"Nucleotide Transformer tasks, Genomic Benchmarks and GUE. Mean accuracy and Matthews correlation coefficient are reported per fine-tuning task; multiclass tasks use macro averaging. Results remain task-specific across the three benchmark collections. Attention-based and state-space genomic language models with different tokenizer choices.","source_ids":["genomic-tokenizer-selection-2025"],"source_locator":"Methods §2.2 Benchmarks; Results §3.1; cached text lines 20–21, 43–44; §2.2–2.3 Benchmarks and Metrics; Methods §2.2 Benchmarks; Results §3.1; cached text lines 20–21, 43–44"}],"facts":[{"label":"Datasets","value":"Nucleotide Transformer tasks, Genomic Benchmarks and GUE.","status":"source_checked","source_ids":["genomic-tokenizer-selection-2025"],"source_locator":"Methods §2.2 Benchmarks; Results §3.1; cached text lines 20–21, 43–44"},{"label":"Splits","value":"The experiment reuses three published fine-tuning benchmark collections and repeats every task at least ten times. The broad catalogue label spans multiple task partitions; no single regulatory-element split is defined by that label.","status":"source_checked","source_ids":["genomic-tokenizer-selection-2025"],"source_locator":"§2.2 Benchmarks"},{"label":"Metrics","value":"Mean accuracy and Matthews correlation coefficient are reported per fine-tuning task; multiclass tasks use macro averaging. Results remain task-specific across the three benchmark collections.","status":"source_checked","source_ids":["genomic-tokenizer-selection-2025"],"source_locator":"§2.2–2.3 Benchmarks and Metrics"},{"label":"Baselines","value":"Attention-based and state-space genomic language models with different tokenizer choices.","status":"source_checked","source_ids":["genomic-tokenizer-selection-2025"],"source_locator":"Methods §2.2 Benchmarks; Results §3.1; cached text lines 20–21, 43–44"},{"label":"Leakage controls","value":"The benchmark reuses GB, NTTv2 and GUE task datasets and compares models pretrained on different corpora, including the human reference genome. Its main Methods do not report a unified overlap-removal audit between all task examples and every model’s pretraining corpus.","status":"unreported","source_ids":["genomic-tokenizer-selection-2025"],"source_locator":"§2 Materials and methods, model/pretraining and task-dataset descriptions; Table 1"},{"label":"Uncertainty","value":"Each fine-tuning task is replicated at least ten times; hyperparameter search differs by model family.","status":"source_checked","source_ids":["genomic-tokenizer-selection-2025"],"source_locator":"Methods §2.2 Benchmarks; Results §3.1; cached text lines 20–21, 43–44"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["genomic-tokenizer-selection-2025"],"source_locator":"Methods §2.2 Benchmarks; Results §3.1; cached text lines 20–21, 43–44"},{"label":"Organisms","value":"Dataset-specific organisms in Nucleotide Transformer, Genomic Benchmarks and GUE.","status":"source_checked","source_ids":["genomic-tokenizer-selection-2025"],"source_locator":"Methods §2.2 Benchmarks; Results §3.1; cached text lines 20–21, 43–44"},{"label":"Assays","value":"Regulatory and other genomic classification labels.","status":"source_checked","source_ids":["genomic-tokenizer-selection-2025"],"source_locator":"Methods §2.2 Benchmarks; Results §3.1; cached text lines 20–21, 43–44"},{"label":"Allowed inputs","value":"Tokenized DNA sequence.","status":"source_checked","source_ids":["genomic-tokenizer-selection-2025"],"source_locator":"Methods §2.2 Benchmarks; Results §3.1; cached text lines 20–21, 43–44"},{"label":"Adaptation","value":"Repeated task fine-tuning compares tokenizer/model configurations.","status":"source_checked","source_ids":["genomic-tokenizer-selection-2025"],"source_locator":"Methods §2.2 Benchmarks; Results §3.1; cached text lines 20–21, 43–44"}],"strengths":[],"limitations":[{"text":"Different hyperparameter-tuning procedures can affect the comparison. The broad catalogue label does not uniquely identify one dataset or executable protocol.","source_ids":["genomic-tokenizer-selection-2025"],"source_locator":"Methods §2.2 Benchmarks; Results §3.1; cached text lines 20–21, 43–44"}],"diagram":{"title":"Computational evaluation flow","steps":["Allowed inputs: Tokenized DNA sequence.","Datasets: Nucleotide Transformer tasks, Genomic Benchmarks and GUE.","Metrics: Mean accuracy and Matthews correlation coefficient are reported per fine-tuning task; multiclass tasks use macro averaging. Results remain task-specific across the three benchmark collections."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["genomic-tokenizer-selection-2025"],"source_locator":"Methods §2.2 Benchmarks; Results §3.1; cached text lines 20–21, 43–44; Methods §2.2 Benchmarks; Results §3.1; cached text lines 20–21, 43–44; §2.2–2.3 Benchmarks and Metrics"},"coverage":"limited","gaps":["Leakage controls: The benchmark reuses GB, NTTv2 and GUE task datasets and compares models pretrained on different corpora, including the human reference genome. Its main Methods do not report a unified overlap-removal audit between all task examples and every model’s pretraining corpus."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Relevant full-paper computational evaluation sections, tables/captions and cited supplementary task passages were reviewed. Reporting omissions are scoped to the inspected sources. Original numerical results are unchanged."}}}}
{"id":"reported-task-cdbee1c9285568","kind":"benchmark","name":"regulatory element identification","description":"","status":"needs_review","facets":{"areas":["dna-genomes"],"tasks":["regulatory element identification"]},"source_ids":["dart-eval-regulatory-2024"],"links":[{"relation":"dataset","target_id":"reported-dataset-b6ce37ba678d39"}],"attributes":{"entity_level":"task","version":null,"task":"regulatory element identification","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Regulatory-element identification tests whether DNA models distinguish ENCODE regulatory sequences from composition-matched controls.","summary_source_ids":["dart-eval-regulatory-2024"],"summary_source_locator":"Paper §3, §4.1, Table 3; Appendices D.2–D.3 and E.1","sections":[{"title":"Evaluation methodology","body":"ENCODE candidate cis-regulatory elements paired with synthetic dinucleotide-shuffled controls. Test chromosomes are 5, 10, 14, 18, 20 and 22; validation uses 6 and 21; other chromosomes form training data. Supervised checkpoints are selected by validation loss. Zero-shot accuracy measures which member of a matched pair receives higher sequence likelihood. Supervised absolute classification accuracy and paired ranking accuracy are separate quantities. Six DNA language-model families are compared in zero-shot, frozen-embedding probing and LoRA fine-tuning settings; a supervised CNN supplies an ab-initio reference. Chromosome holdout separates supervised fitting from testing. Synthetic negatives preserve dinucleotide composition; pretraining sequence overlap is not resolved by chromosome splitting. The appendix reports a one-sided Wilcoxon rank-sum test for regulatory versus control likelihoods.","source_ids":["dart-eval-regulatory-2024"],"source_locator":"Paper §3, §4.1, Table 3; Appendices D.2–D.3 and E.1"}],"facts":[{"label":"Datasets","value":"ENCODE candidate cis-regulatory elements paired with synthetic dinucleotide-shuffled controls.","status":"source_checked","source_ids":["dart-eval-regulatory-2024"],"source_locator":"Paper §3, §4.1, Table 3; Appendices D.2–D.3 and E.1"},{"label":"Splits","value":"Test chromosomes are 5, 10, 14, 18, 20 and 22; validation uses 6 and 21; other chromosomes form training data. Supervised checkpoints are selected by validation loss.","status":"source_checked","source_ids":["dart-eval-regulatory-2024"],"source_locator":"Paper §3, §4.1, Table 3; Appendices D.2–D.3 and E.1"},{"label":"Metrics","value":"Zero-shot accuracy measures which member of a matched pair receives higher sequence likelihood. Supervised absolute classification accuracy and paired ranking accuracy are separate quantities.","status":"source_checked","source_ids":["dart-eval-regulatory-2024"],"source_locator":"Paper §3, §4.1, Table 3; Appendices D.2–D.3 and E.1"},{"label":"Baselines","value":"Six DNA language-model families are compared in zero-shot, frozen-embedding probing and LoRA fine-tuning settings; a supervised CNN supplies an ab-initio reference.","status":"source_checked","source_ids":["dart-eval-regulatory-2024"],"source_locator":"Paper §3, §4.1, Table 3; Appendices D.2–D.3 and E.1"},{"label":"Leakage controls","value":"Chromosome holdout separates supervised fitting from testing. Synthetic negatives preserve dinucleotide composition; pretraining sequence overlap is not resolved by chromosome splitting.","status":"source_checked","source_ids":["dart-eval-regulatory-2024"],"source_locator":"Paper §3, §4.1, Table 3; Appendices D.2–D.3 and E.1"},{"label":"Uncertainty","value":"The appendix reports a one-sided Wilcoxon rank-sum test for regulatory versus control likelihoods.","status":"source_checked","source_ids":["dart-eval-regulatory-2024"],"source_locator":"Paper §3, §4.1, Table 3; Appendices D.2–D.3 and E.1"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["dart-eval-regulatory-2024"],"source_locator":"Paper §3, §4.1, Table 3; Appendices D.2–D.3 and E.1"},{"label":"Organisms","value":"Human.","status":"source_checked","source_ids":["dart-eval-regulatory-2024"],"source_locator":"Paper §3, §4.1, Table 3; Appendices D.2–D.3 and E.1"},{"label":"Assays","value":"ENCODE regulatory-element annotations with synthetic sequence controls.","status":"source_checked","source_ids":["dart-eval-regulatory-2024"],"source_locator":"Paper §3, §4.1, Table 3; Appendices D.2–D.3 and E.1"},{"label":"Allowed inputs","value":"DNA sequence, with matched controls preserving dinucleotide composition.","status":"source_checked","source_ids":["dart-eval-regulatory-2024"],"source_locator":"Paper §3, §4.1, Table 3; Appendices D.2–D.3 and E.1"},{"label":"Adaptation","value":"Zero-shot likelihood comparison, frozen-model probing, LoRA fine-tuning and supervised CNN training are distinct regimes.","status":"source_checked","source_ids":["dart-eval-regulatory-2024"],"source_locator":"Paper §3, §4.1, Table 3; Appendices D.2–D.3 and E.1"}],"strengths":[{"text":"Matched composition controls and chromosome holdout address distinct confounders.","source_ids":["dart-eval-regulatory-2024"],"source_locator":"Paper §3, §4.1, Table 3; Appendices D.2–D.3 and E.1"}],"limitations":[{"text":"This record concerns Table 3 regulatory-element identification; other DART tasks have different targets and metrics. No prediction of clinical variant effects follows from this task alone.","source_ids":["dart-eval-regulatory-2024"],"source_locator":"Paper §3, §4.1, Table 3; Appendices D.2–D.3 and E.1"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: DNA sequence, with matched controls preserving dinucleotide composition.","Evaluation: Zero-shot likelihood comparison, frozen-model probing, LoRA fine-tuning and supervised CNN training are distinct regimes.","Readout: Zero-shot accuracy measures which member of a matched pair receives higher sequence likelihood. Supervised absolute classification accuracy and paired ranking accuracy are separate quantities."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["dart-eval-regulatory-2024"],"source_locator":"Paper §3, §4.1, Table 3; Appendices D.2–D.3 and E.1"},"coverage":"limited","gaps":["This record concerns Table 3 regulatory-element identification; other DART tasks have different targets and metrics. No prediction of clinical variant effects follows from this task alone."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Task-specific computational methodology and field context checked in the cited primary-source artifact. Source-backed fields, inapplicable evaluator dimensions and unresolved details are distinguished. Numerical results were not reproduced."}}}}
{"id":"reported-task-d1c46526c39983","kind":"benchmark","name":"Physically valid protein–ligand pose selection","description":"","status":"needs_review","facets":{"areas":["molecular-interactions"],"tasks":["Physically valid protein–ligand pose selection"]},"source_ids":["molas-2026"],"links":[{"relation":"dataset","target_id":"reported-dataset-4b6c13924d4256"}],"attributes":{"entity_level":"task","version":null,"task":"Physically valid protein–ligand pose selection","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Docking algorithm selection is evaluated as a workflow-dependent choice among a fixed solver portfolio.","summary_source_ids":["molas-2026"],"summary_source_locator":"Methods: Datasets and preprocessing; Results framing; cached text lines 5, 9, 18–21; task metric definitions and corresponding results table","sections":[{"title":"Evaluation methodology","body":"A curated BindingMOAD-derived set and additional docking benchmarks. Curated subset construction excludes overlap using rules independent of docking scores and solver rankings. PoseBusters-validity-gated pose agreement forms the workflow-specific success score; solver ranking depends on this combined criterion. A portfolio of eight docking algorithms is evaluated; it is explicitly not claimed to be exhaustive. The data-curation procedure is designed to reduce overlap with common PDBbind training corpora.","source_ids":["molas-2026"],"source_locator":"Methods: Datasets and preprocessing; Results framing; cached text lines 5, 9, 18–21; task metric definitions and corresponding results table"}],"facts":[{"label":"Datasets","value":"A curated BindingMOAD-derived set and additional docking benchmarks.","status":"source_checked","source_ids":["molas-2026"],"source_locator":"Methods: Datasets and preprocessing; Results framing; cached text lines 5, 9, 18–21; task metric definitions and corresponding results table"},{"label":"Splits","value":"Curated subset construction excludes overlap using rules independent of docking scores and solver rankings.","status":"source_checked","source_ids":["molas-2026"],"source_locator":"Methods: Datasets and preprocessing; Results framing; cached text lines 5, 9, 18–21; task metric definitions and corresponding results table"},{"label":"Metrics","value":"PoseBusters-validity-gated pose agreement forms the workflow-specific success score; solver ranking depends on this combined criterion.","status":"source_checked","source_ids":["molas-2026"],"source_locator":"Methods: Datasets and preprocessing; Results framing; cached text lines 5, 9, 18–21; task metric definitions and corresponding results table"},{"label":"Baselines","value":"A portfolio of eight docking algorithms is evaluated; it is explicitly not claimed to be exhaustive.","status":"source_checked","source_ids":["molas-2026"],"source_locator":"Methods: Datasets and preprocessing; Results framing; cached text lines 5, 9, 18–21; task metric definitions and corresponding results table"},{"label":"Leakage controls","value":"The data-curation procedure is designed to reduce overlap with common PDBbind training corpora.","status":"source_checked","source_ids":["molas-2026"],"source_locator":"Methods: Datasets and preprocessing; Results framing; cached text lines 5, 9, 18–21; task metric definitions and corresponding results table"},{"label":"Uncertainty","value":"In-domain metrics are means over five folds. Table 3 marks paired significance tests against the single-best solver, while the cross-benchmark discussion explicitly distinguishes marginal, nonsignificant gains. The score-margin diagnostics are benchmark-dependent confidence proxies, not calibrated success probabilities.","status":"source_checked","source_ids":["molas-2026"],"source_locator":"Evaluation regimes; Table 3 and footnote; In-domain learning; Fig.4 and Cross-benchmark generalisation"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["molas-2026"],"source_locator":"Methods: Datasets and preprocessing; Results framing; cached text lines 5, 9, 18–21; task metric definitions and corresponding results table"},{"label":"Organisms","value":"MOAD-curated is selected from BindingMOAD by complex quality and overlap exclusions, alongside separate docking benchmarks. The dataset-construction section does not enumerate organisms or specify a species-conditioned evaluation.","status":"unreported","source_ids":["molas-2026"],"source_locator":"Dataset construction: MOAD-curated, PoseX and PoseBusters"},{"label":"Assays","value":"Protein–ligand structures with physical-validity and reference-pose checks.","status":"source_checked","source_ids":["molas-2026"],"source_locator":"Methods: Datasets and preprocessing; Results framing; cached text lines 5, 9, 18–21; task metric definitions and corresponding results table"},{"label":"Allowed inputs","value":"Candidate docking poses and protein–ligand complexes.","status":"source_checked","source_ids":["molas-2026"],"source_locator":"Methods: Datasets and preprocessing; Results framing; cached text lines 5, 9, 18–21; task metric definitions and corresponding results table"},{"label":"Adaptation","value":"Selection/ranking among poses from a docking-method portfolio; subset construction is independent of scores.","status":"source_checked","source_ids":["molas-2026"],"source_locator":"Methods: Datasets and preprocessing; Results framing; cached text lines 5, 9, 18–21; task metric definitions and corresponding results table"}],"strengths":[],"limitations":[{"text":"Solver rankings and oracle labels depend on benchmark and post-processing. Confidence margins are not uniformly reliable indicators of improvement over the single-best solver.","source_ids":["molas-2026"],"source_locator":"Methods: Datasets and preprocessing; Results framing; cached text lines 5, 9, 18–21; task metric definitions and corresponding results table"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Candidate docking poses and protein–ligand complexes.","Evaluation: Selection/ranking among poses from a docking-method portfolio; subset construction is independent of scores.","Readout: PoseBusters-validity-gated pose agreement forms the workflow-specific success score; solver ranking depends on this combined criterion."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["molas-2026"],"source_locator":"Methods: Datasets and preprocessing; Results framing; cached text lines 5, 9, 18–21; task metric definitions and corresponding results table"},"coverage":"limited","gaps":["Organisms: MOAD-curated is selected from BindingMOAD by complex quality and overlap exclusions, alongside separate docking benchmarks. The dataset-construction section does not enumerate organisms or specify a species-conditioned evaluation."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Relevant full-paper computational evaluation sections, tables/captions and cited supplementary task passages were reviewed. Reporting omissions are scoped to the inspected sources. Original numerical results are unchanged."}}}}
{"id":"reported-task-d3fd502fdc2b38","kind":"benchmark","name":"pathogen detection","description":"","status":"needs_review","facets":{"areas":["microbes-communities"],"tasks":["pathogen detection"]},"source_ids":["metagenomic-pathogens-2025"],"links":[{"relation":"dataset","target_id":"reported-dataset-1c4c71078ffe01"}],"attributes":{"entity_level":"task","version":null,"task":"pathogen detection","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"The source reports a separate metagenomic taxon-detection comparison, but its executable evaluation protocol remains incomplete.","summary_source_ids":["metagenomic-pathogens-2025"],"summary_source_locator":"Methods §§4.1–4.2; Results §4.3, Table 3 and preceding paragraph; cached text lines 118–133","sections":[{"title":"Evaluation methodology","body":"MetaHIT and iHMP shotgun-sequencing cohorts; the source describes expert annotations or high-confidence reference calls as ground truth without resolving their construction here. The genomic experiment in §4.3 does not specify a sample/split manifest. The five-fold procedure in §4.2 concerns other datasets and cannot supply this missing genomic protocol. Precision, recall, F1 and ROC-AUC for the genomic comparison. Kraken2 and MetaPhlAn3. The genomic experiment in §4.3 does not establish independence of the expert/reference-derived labels from the compared prediction systems.","source_ids":["metagenomic-pathogens-2025"],"source_locator":"Methods §§4.1–4.2; Results §4.3, Table 3 and preceding paragraph; cached text lines 118–133"}],"facts":[{"label":"Datasets","value":"MetaHIT and iHMP shotgun-sequencing cohorts; the source describes expert annotations or high-confidence reference calls as ground truth without resolving their construction here.","status":"source_checked","source_ids":["metagenomic-pathogens-2025"],"source_locator":"Methods §§4.1–4.2; Results §4.3, Table 3 and preceding paragraph; cached text lines 118–133"},{"label":"Splits","value":"The genomic experiment in §4.3 does not specify a sample/split manifest. The five-fold procedure in §4.2 concerns other datasets and cannot supply this missing genomic protocol.","status":"unreported","source_ids":["metagenomic-pathogens-2025"],"source_locator":"Methods §§4.1–4.2; Results §4.3, Table 3 and preceding paragraph; cached text lines 118–133"},{"label":"Metrics","value":"Precision, recall, F1 and ROC-AUC for the genomic comparison.","status":"source_checked","source_ids":["metagenomic-pathogens-2025"],"source_locator":"Methods §§4.1–4.2; Results §4.3, Table 3 and preceding paragraph; cached text lines 118–133"},{"label":"Baselines","value":"Kraken2 and MetaPhlAn3.","status":"source_checked","source_ids":["metagenomic-pathogens-2025"],"source_locator":"Methods §§4.1–4.2; Results §4.3, Table 3 and preceding paragraph; cached text lines 118–133"},{"label":"Leakage controls","value":"The genomic experiment in §4.3 does not establish independence of the expert/reference-derived labels from the compared prediction systems.","status":"unreported","source_ids":["metagenomic-pathogens-2025"],"source_locator":"Methods §§4.1–4.2; Results §4.3, Table 3 and preceding paragraph; cached text lines 118–133"},{"label":"Uncertainty","value":"MetaHIT/iHMP Table 3 gives point precision, recall, F1 and AUC values without a replicate count or confidence-interval procedure. The ± values and five-fold evaluation elsewhere concern other datasets and cannot supply uncertainty for this sequencing comparison.","status":"unreported","source_ids":["metagenomic-pathogens-2025"],"source_locator":"§4.3, MetaHIT/iHMP experiment paragraph and Table 3; contrast §4.2 and Table 2"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["metagenomic-pathogens-2025"],"source_locator":"Methods §§4.1–4.2; Results §4.3, Table 3 and preceding paragraph; cached text lines 118–133"},{"label":"Organisms","value":"Human-cohort microbial communities in MetaHIT and iHMP.","status":"source_checked","source_ids":["metagenomic-pathogens-2025"],"source_locator":"Methods §§4.1–4.2; Results §4.3, Table 3 and preceding paragraph; cached text lines 118–133"},{"label":"Assays","value":"Shotgun sequencing; expert/reference-derived labels whose construction remains insufficiently specified.","status":"source_checked","source_ids":["metagenomic-pathogens-2025"],"source_locator":"Methods §§4.1–4.2; Results §4.3, Table 3 and preceding paragraph; cached text lines 118–133"},{"label":"Allowed inputs","value":"Metagenomic sequencing data and model-specific representations.","status":"source_checked","source_ids":["metagenomic-pathogens-2025"],"source_locator":"Methods §§4.1–4.2; Results §4.3, Table 3 and preceding paragraph; cached text lines 118–133"},{"label":"Adaptation","value":"Section 4.3 says the model is applied directly to preprocessed sequencing reads but does not specify its genomic fitting partition, checkpoint or whether MetaHIT/iHMP labels were available during fitting. The five-fold setup in §4.2 is attached to other datasets.","status":"unreported","source_ids":["metagenomic-pathogens-2025"],"source_locator":"§4.2–4.3, genomic sequencing experiment immediately preceding Table 3"}],"strengths":[],"limitations":[{"text":"Section 4.3 reports a sequencing experiment, but does not fix sample accessions, split membership, genomic fitting or reference-label construction. The unrelated five-fold setup in §4.2 does not resolve these protocol gaps.","source_ids":["metagenomic-pathogens-2025"],"source_locator":"Methods §§4.1–4.2; Results §4.3, Table 3 and preceding paragraph; cached text lines 118–133"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Metagenomic sequencing data and model-specific representations.","Task: The source reports a separate metagenomic taxon-detection comparison, but its executable evaluation protocol remains incomplete."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["metagenomic-pathogens-2025"],"source_locator":"Methods §§4.1–4.2; Results §4.3, Table 3 and preceding paragraph; cached text lines 118–133"},"coverage":"limited","gaps":["Splits: The genomic experiment in §4.3 does not specify a sample/split manifest. The five-fold procedure in §4.2 concerns other datasets and cannot supply this missing genomic protocol.","Leakage controls: The genomic experiment in §4.3 does not establish independence of the expert/reference-derived labels from the compared prediction systems.","Uncertainty: MetaHIT/iHMP Table 3 gives point precision, recall, F1 and AUC values without a replicate count or confidence-interval procedure. The ± values and five-fold evaluation elsewhere concern other datasets and cannot supply uncertainty for this sequencing comparison.","Adaptation: Section 4.3 says the model is applied directly to preprocessed sequencing reads but does not specify its genomic fitting partition, checkpoint or whether MetaHIT/iHMP labels were available during fitting. The five-fold setup in §4.2 is attached to other datasets."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Relevant full-paper computational evaluation sections, tables/captions and cited supplementary task passages were reviewed. Reporting omissions are scoped to the inspected sources. Original numerical results are unchanged."}}}}
{"id":"reported-task-d5f897ab0f6f67","kind":"benchmark","name":"Ligand potency prediction using generated poses","description":"","status":"needs_review","facets":{"areas":["molecular-interactions"],"tasks":["Ligand potency prediction using generated poses"]},"source_ids":["mpro-pose-affinity-2025"],"links":[{"relation":"dataset","target_id":"reported-dataset-235520c84b737f"}],"attributes":{"entity_level":"task","version":null,"task":"Ligand potency prediction using generated poses","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Antiviral ligand-potency prediction uses a challenge dataset distinct from the paper’s pose-prediction challenge.","summary_source_ids":["mpro-pose-affinity-2025"],"summary_source_locator":"Methods: Antiviral Potency Prediction Challenge data preparation; cached text lines 41–43; task metric definitions and corresponding results table","sections":[{"title":"Evaluation methodology","body":"ASAP Discovery Antiviral Potency Prediction Challenge 2025 data hosted on Polaris. The challenge supplies training and test compounds for two protein targets. MAE and RMSE for predicted compound potency; these regression errors are separate from pose agreement. Training/validation bootstrap replicates assess model stability. They do not by themselves establish uncertainty on the untouched challenge test set.","source_ids":["mpro-pose-affinity-2025"],"source_locator":"Methods: Antiviral Potency Prediction Challenge data preparation; cached text lines 41–43; task metric definitions and corresponding results table"}],"facts":[{"label":"Datasets","value":"ASAP Discovery Antiviral Potency Prediction Challenge 2025 data hosted on Polaris.","status":"source_checked","source_ids":["mpro-pose-affinity-2025"],"source_locator":"Methods: Antiviral Potency Prediction Challenge data preparation; cached text lines 41–43; task metric definitions and corresponding results table"},{"label":"Splits","value":"The challenge supplies training and test compounds for two protein targets.","status":"source_checked","source_ids":["mpro-pose-affinity-2025"],"source_locator":"Methods: Antiviral Potency Prediction Challenge data preparation; cached text lines 41–43; task metric definitions and corresponding results table"},{"label":"Metrics","value":"MAE and RMSE for predicted compound potency; these regression errors are separate from pose agreement.","status":"source_checked","source_ids":["mpro-pose-affinity-2025"],"source_locator":"Methods: Antiviral Potency Prediction Challenge data preparation; cached text lines 41–43; task metric definitions and corresponding results table"},{"label":"Baselines","value":"The potency study compares a common LRIP-SF predictor supplied with poses from Glide, AutoDock Vina, FlexS, AlphaFold3, Boltz-2, DiffDock and Gnina variants. Thus the comparison varies pose provenance as well as evaluating the downstream potency regressor; some models were added after the challenge.","status":"source_checked","source_ids":["mpro-pose-affinity-2025"],"source_locator":"Introduction: challenge approaches; Methods: pose generation and LRIP-SF training; potency comparison results"},{"label":"Leakage controls","value":"The regression models use repeated random training/validation splits of the challenge training compounds. The inspected training and comparison methods do not specify scaffold-disjoint folds or an audit of challenge compounds/targets against each pose model’s pretraining data.","status":"unreported","source_ids":["mpro-pose-affinity-2025"],"source_locator":"Methods: LRIP-SF model training; challenge train/test description; Discussion of postchallenge comparisons"},{"label":"Uncertainty","value":"Training/validation bootstrap replicates assess model stability. They do not by themselves establish uncertainty on the untouched challenge test set.","status":"source_checked","source_ids":["mpro-pose-affinity-2025"],"source_locator":"Methods: Antiviral Potency Prediction Challenge data preparation; cached text lines 41–43; task metric definitions and corresponding results table"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["mpro-pose-affinity-2025"],"source_locator":"Methods: Antiviral Potency Prediction Challenge data preparation; cached text lines 41–43; task metric definitions and corresponding results table"},{"label":"Organisms","value":"Viral protease targets in the antiviral potency challenge.","status":"source_checked","source_ids":["mpro-pose-affinity-2025"],"source_locator":"Methods: Antiviral Potency Prediction Challenge data preparation; cached text lines 41–43; task metric definitions and corresponding results table"},{"label":"Assays","value":"Measured compound potency.","status":"source_checked","source_ids":["mpro-pose-affinity-2025"],"source_locator":"Methods: Antiviral Potency Prediction Challenge data preparation; cached text lines 41–43; task metric definitions and corresponding results table"},{"label":"Allowed inputs","value":"Ligand representations and generated protein–ligand poses.","status":"source_checked","source_ids":["mpro-pose-affinity-2025"],"source_locator":"Methods: Antiviral Potency Prediction Challenge data preparation; cached text lines 41–43; task metric definitions and corresponding results table"},{"label":"Adaptation","value":"Supervised potency prediction using challenge training compounds and a separate test collection.","status":"source_checked","source_ids":["mpro-pose-affinity-2025"],"source_locator":"Methods: Antiviral Potency Prediction Challenge data preparation; cached text lines 41–43; task metric definitions and corresponding results table"}],"strengths":[],"limitations":[{"text":"Pose generation and downstream potency fitting jointly affect the comparison. Random training/validation resampling does not demonstrate scaffold-disjoint generalization.","source_ids":["mpro-pose-affinity-2025"],"source_locator":"Methods: Antiviral Potency Prediction Challenge data preparation; cached text lines 41–43; task metric definitions and corresponding results table"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Ligand representations and generated protein–ligand poses.","Evaluation: Supervised potency prediction using challenge training compounds and a separate test collection.","Readout: MAE and RMSE for predicted compound potency; these regression errors are separate from pose agreement."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["mpro-pose-affinity-2025"],"source_locator":"Methods: Antiviral Potency Prediction Challenge data preparation; cached text lines 41–43; task metric definitions and corresponding results table"},"coverage":"limited","gaps":["Leakage controls: The regression models use repeated random training/validation splits of the challenge training compounds. The inspected training and comparison methods do not specify scaffold-disjoint folds or an audit of challenge compounds/targets against each pose model’s pretraining data."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Relevant full-paper computational evaluation sections, tables/captions and cited supplementary task passages were reviewed. Reporting omissions are scoped to the inspected sources. Original numerical results are unchanged."}}}}
{"id":"reported-task-d6018ca598e525","kind":"benchmark","name":"Donor-aware age-class prediction","description":"","status":"needs_review","facets":{"areas":["cells-tissues"],"tasks":["Donor-aware age-class prediction"]},"source_ids":["single-cell-aging-probes-2026"],"links":[{"relation":"dataset","target_id":"reported-dataset-571ce000cd74b9"}],"attributes":{"entity_level":"task","version":null,"task":"Donor-aware age-class prediction","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Age-class prediction probes frozen single-cell representations using donor-separated evaluation.","summary_source_ids":["single-cell-aging-probes-2026"],"summary_source_locator":"Methods: Datasets; Models, baselines and null calibrations; Age probe; cached text lines 13–22","sections":[{"title":"Evaluation methodology","body":"Five age-annotated human PBMC cohorts; AIDA cohorts are central and other cohorts provide stress tests. All cells from a donor belong wholly to training or testing; targets are age-quartile classes. Balanced-accuracy comparisons against null expectations are described. Frozen scGPT and Geneformer probes, PCA-on-expression, within-stratum permutations and chance baselines. Donor grouping prevents direct memorization of donor identities across partitions. Permutation nulls calibrate the probe; exact test confidence intervals remain unextracted.","source_ids":["single-cell-aging-probes-2026"],"source_locator":"Methods: Datasets; Models, baselines and null calibrations; Age probe; cached text lines 13–22"}],"facts":[{"label":"Datasets","value":"Five age-annotated human PBMC cohorts; AIDA cohorts are central and other cohorts provide stress tests.","status":"source_checked","source_ids":["single-cell-aging-probes-2026"],"source_locator":"Methods: Datasets; Models, baselines and null calibrations; Age probe; cached text lines 13–22"},{"label":"Splits","value":"All cells from a donor belong wholly to training or testing; targets are age-quartile classes.","status":"source_checked","source_ids":["single-cell-aging-probes-2026"],"source_locator":"Methods: Datasets; Models, baselines and null calibrations; Age probe; cached text lines 13–22"},{"label":"Metrics","value":"Balanced-accuracy comparisons against null expectations are described.","status":"source_checked","source_ids":["single-cell-aging-probes-2026"],"source_locator":"Methods: Datasets; Models, baselines and null calibrations; Age probe; cached text lines 13–22"},{"label":"Baselines","value":"Frozen scGPT and Geneformer probes, PCA-on-expression, within-stratum permutations and chance baselines.","status":"source_checked","source_ids":["single-cell-aging-probes-2026"],"source_locator":"Methods: Datasets; Models, baselines and null calibrations; Age probe; cached text lines 13–22"},{"label":"Leakage controls","value":"Donor grouping prevents direct memorization of donor identities across partitions.","status":"source_checked","source_ids":["single-cell-aging-probes-2026"],"source_locator":"Methods: Datasets; Models, baselines and null calibrations; Age probe; cached text lines 13–22"},{"label":"Uncertainty","value":"Permutation nulls calibrate the probe; exact test confidence intervals remain unextracted.","status":"source_checked","source_ids":["single-cell-aging-probes-2026"],"source_locator":"Methods: Datasets; Models, baselines and null calibrations; Age probe; cached text lines 13–22"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["single-cell-aging-probes-2026"],"source_locator":"Methods: Datasets; Models, baselines and null calibrations; Age probe; cached text lines 13–22"},{"label":"Organisms","value":"Human PBMC donors.","status":"source_checked","source_ids":["single-cell-aging-probes-2026"],"source_locator":"Methods: Datasets; Models, baselines and null calibrations; Age probe; cached text lines 13–22"},{"label":"Assays","value":"Single-cell expression with donor age and cell-type metadata.","status":"source_checked","source_ids":["single-cell-aging-probes-2026"],"source_locator":"Methods: Datasets; Models, baselines and null calibrations; Age probe; cached text lines 13–22"},{"label":"Allowed inputs","value":"Frozen scGPT/Geneformer embeddings or expression-derived PCA representations.","status":"source_checked","source_ids":["single-cell-aging-probes-2026"],"source_locator":"Methods: Datasets; Models, baselines and null calibrations; Age probe; cached text lines 13–22"},{"label":"Adaptation","value":"Supervised probes; all cells of a donor stay within one arm.","status":"source_checked","source_ids":["single-cell-aging-probes-2026"],"source_locator":"Methods: Datasets; Models, baselines and null calibrations; Age probe; cached text lines 13–22"}],"strengths":[{"text":"Whole-donor partitioning prevents cells from one donor appearing in both fitting and testing.","source_ids":["single-cell-aging-probes-2026"],"source_locator":"Methods: Datasets; Models, baselines and null calibrations; Age probe; cached text lines 13–22"}],"limitations":[{"text":"A fitted probe is distinct from fine-tuning the foundation model. External cohorts are not treated as interchangeable training pools.","source_ids":["single-cell-aging-probes-2026"],"source_locator":"Methods: Datasets; Models, baselines and null calibrations; Age probe; cached text lines 13–22"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Frozen scGPT/Geneformer embeddings or expression-derived PCA representations.","Evaluation: Supervised probes; all cells of a donor stay within one arm.","Readout: Balanced-accuracy comparisons against null expectations are described."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["single-cell-aging-probes-2026"],"source_locator":"Methods: Datasets; Models, baselines and null calibrations; Age probe; cached text lines 13–22"},"coverage":"limited","gaps":["A fitted probe is distinct from fine-tuning the foundation model. External cohorts are not treated as interchangeable training pools."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Task-specific computational methodology and field context checked in the cited primary-source artifact. Source-backed fields, inapplicable evaluator dimensions and unresolved details are distinguished. Numerical results were not reproduced."}}}}
{"id":"reported-task-d635fc6c281a27","kind":"benchmark","name":"A-to-I RNA editing site prediction","description":"","status":"needs_review","facets":{"areas":["rna-transcriptomes"],"tasks":["A-to-I RNA editing site prediction"]},"source_ids":["adar-gpt-editing-2026"],"links":[{"relation":"dataset","target_id":"reported-dataset-236eaa4e55147f"}],"attributes":{"entity_level":"task","version":null,"task":"A-to-I RNA editing site prediction","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"An RNA editing-site classifier is compared with sequence-model baselines on held-out human liver annotations.","summary_source_ids":["adar-gpt-editing-2026"],"summary_source_locator":"Methods: Dataset Construction and Annotation; Dataset Design and Labeling; Evaluation and Reproducibility; Table 2; cached text lines 91–110","sections":[{"title":"Evaluation methodology","body":"GTEx liver RNA-seq annotations in Alu-associated regions; labels distinguish editing levels. Random 80:20 splits within disjoint site groups; final comparisons use the held-out highest-threshold validation group. Accuracy, precision, recall, specificity and F1; probability-based AUROC and AUPRC. Static fine-tuning, pretrained and fine-tuned EditPredict, and fine-tuned RNA-FM. Editing sites are assigned to nonoverlapping groups. Independence of overlapping sequence windows or donors was not established in this review. Five inference repeats quantify prediction variability; they do not measure variability across retraining or independent cohorts.","source_ids":["adar-gpt-editing-2026"],"source_locator":"Methods: Dataset Construction and Annotation; Dataset Design and Labeling; Evaluation and Reproducibility; Table 2; cached text lines 91–110"}],"facts":[{"label":"Datasets","value":"GTEx liver RNA-seq annotations in Alu-associated regions; labels distinguish editing levels.","status":"source_checked","source_ids":["adar-gpt-editing-2026"],"source_locator":"Methods: Dataset Construction and Annotation; Dataset Design and Labeling; Evaluation and Reproducibility; Table 2; cached text lines 91–110"},{"label":"Splits","value":"Random 80:20 splits within disjoint site groups; final comparisons use the held-out highest-threshold validation group.","status":"source_checked","source_ids":["adar-gpt-editing-2026"],"source_locator":"Methods: Dataset Construction and Annotation; Dataset Design and Labeling; Evaluation and Reproducibility; Table 2; cached text lines 91–110"},{"label":"Metrics","value":"Accuracy, precision, recall, specificity and F1; probability-based AUROC and AUPRC.","status":"source_checked","source_ids":["adar-gpt-editing-2026"],"source_locator":"Methods: Dataset Construction and Annotation; Dataset Design and Labeling; Evaluation and Reproducibility; Table 2; cached text lines 91–110"},{"label":"Baselines","value":"Static fine-tuning, pretrained and fine-tuned EditPredict, and fine-tuned RNA-FM.","status":"source_checked","source_ids":["adar-gpt-editing-2026"],"source_locator":"Methods: Dataset Construction and Annotation; Dataset Design and Labeling; Evaluation and Reproducibility; Table 2; cached text lines 91–110"},{"label":"Leakage controls","value":"Editing sites are assigned to nonoverlapping groups. Independence of overlapping sequence windows or donors was not established in this review.","status":"source_checked","source_ids":["adar-gpt-editing-2026"],"source_locator":"Methods: Dataset Construction and Annotation; Dataset Design and Labeling; Evaluation and Reproducibility; Table 2; cached text lines 91–110"},{"label":"Uncertainty","value":"Five inference repeats quantify prediction variability; they do not measure variability across retraining or independent cohorts.","status":"source_checked","source_ids":["adar-gpt-editing-2026"],"source_locator":"Methods: Dataset Construction and Annotation; Dataset Design and Labeling; Evaluation and Reproducibility; Table 2; cached text lines 91–110"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["adar-gpt-editing-2026"],"source_locator":"Methods: Dataset Construction and Annotation; Dataset Design and Labeling; Evaluation and Reproducibility; Table 2; cached text lines 91–110"},{"label":"Organisms","value":"Human liver.","status":"source_checked","source_ids":["adar-gpt-editing-2026"],"source_locator":"Methods: Dataset Construction and Annotation; Dataset Design and Labeling; Evaluation and Reproducibility; Table 2; cached text lines 91–110"},{"label":"Assays","value":"RNA-seq editing annotations in Alu-associated regions.","status":"source_checked","source_ids":["adar-gpt-editing-2026"],"source_locator":"Methods: Dataset Construction and Annotation; Dataset Design and Labeling; Evaluation and Reproducibility; Table 2; cached text lines 91–110"},{"label":"Allowed inputs","value":"RNA sequence around candidate editing sites.","status":"source_checked","source_ids":["adar-gpt-editing-2026"],"source_locator":"Methods: Dataset Construction and Annotation; Dataset Design and Labeling; Evaluation and Reproducibility; Table 2; cached text lines 91–110"},{"label":"Adaptation","value":"Task fine-tuning is compared with pretrained and fine-tuned RNA baselines.","status":"source_checked","source_ids":["adar-gpt-editing-2026"],"source_locator":"Methods: Dataset Construction and Annotation; Dataset Design and Labeling; Evaluation and Reproducibility; Table 2; cached text lines 91–110"}],"strengths":[],"limitations":[{"text":"The evaluation partition is called validation in the source and is reused during training monitoring; it should not be relabelled an untouched external test set.","source_ids":["adar-gpt-editing-2026"],"source_locator":"Methods: Dataset Construction and Annotation; Dataset Design and Labeling; Evaluation and Reproducibility; Table 2; cached text lines 91–110"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: RNA sequence around candidate editing sites.","Evaluation: Task fine-tuning is compared with pretrained and fine-tuned RNA baselines.","Readout: Accuracy, precision, recall, specificity and F1; probability-based AUROC and AUPRC."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["adar-gpt-editing-2026"],"source_locator":"Methods: Dataset Construction and Annotation; Dataset Design and Labeling; Evaluation and Reproducibility; Table 2; cached text lines 91–110"},"coverage":"limited","gaps":["The evaluation partition is called validation in the source and is reused during training monitoring; it should not be relabelled an untouched external test set."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Task-specific computational methodology and field context checked in the cited primary-source artifact. Source-backed fields, inapplicable evaluator dimensions and unresolved details are distinguished. Numerical results were not reproduced."}}}}
{"id":"reported-task-d7e6274011946e","kind":"benchmark","name":"mRNA-protein interaction prediction","description":"","status":"needs_review","facets":{"areas":["rna-transcriptomes"],"tasks":["mRNA-protein interaction prediction"]},"source_ids":["mrna-protein-diversity-2026"],"links":[{"relation":"dataset","target_id":"reported-dataset-2e87449871ca47"}],"attributes":{"entity_level":"task","version":null,"task":"mRNA-protein interaction prediction","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"mRNA–protein interaction prediction explicitly contrasts familiar-protein pair prediction with transfer to unseen RNA-binding proteins.","summary_source_ids":["mrna-protein-diversity-2026"],"summary_source_locator":"Datasets and methods; ground-truth dataset construction; cached text lines 8–13; task metric definitions and corresponding results table; matching task comparison table/ablation captions","sections":[{"title":"Evaluation methodology","body":"CLIPdb-derived human RNA-binding annotations with sampled, deduplicated negative pairs. A random-pair test permits protein overlap; an RBP-aware test holds out all test-protein identities. AUROC, AUPRC, F1, precision, recall and specificity, reported separately for each partitioning regime. One-hot attention model, ProteinBERT embedding model and structure-aware encoding; split definitions are compared separately. The protein-held-out test is constructed specifically to expose identity leakage hidden by random pair sampling. The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","source_ids":["mrna-protein-diversity-2026"],"source_locator":"Datasets and methods; ground-truth dataset construction; cached text lines 8–13; task metric definitions and corresponding results table; matching task comparison table/ablation captions"}],"facts":[{"label":"Datasets","value":"CLIPdb-derived human RNA-binding annotations with sampled, deduplicated negative pairs.","status":"source_checked","source_ids":["mrna-protein-diversity-2026"],"source_locator":"Datasets and methods; ground-truth dataset construction; cached text lines 8–13; task metric definitions and corresponding results table; matching task comparison table/ablation captions"},{"label":"Splits","value":"A random-pair test permits protein overlap; an RBP-aware test holds out all test-protein identities.","status":"source_checked","source_ids":["mrna-protein-diversity-2026"],"source_locator":"Datasets and methods; ground-truth dataset construction; cached text lines 8–13; task metric definitions and corresponding results table; matching task comparison table/ablation captions"},{"label":"Metrics","value":"AUROC, AUPRC, F1, precision, recall and specificity, reported separately for each partitioning regime.","status":"source_checked","source_ids":["mrna-protein-diversity-2026"],"source_locator":"Datasets and methods; ground-truth dataset construction; cached text lines 8–13; task metric definitions and corresponding results table; matching task comparison table/ablation captions"},{"label":"Baselines","value":"One-hot attention model, ProteinBERT embedding model and structure-aware encoding; split definitions are compared separately.","status":"source_checked","source_ids":["mrna-protein-diversity-2026"],"source_locator":"Datasets and methods; ground-truth dataset construction; cached text lines 8–13; task metric definitions and corresponding results table; matching task comparison table/ablation captions"},{"label":"Leakage controls","value":"The protein-held-out test is constructed specifically to expose identity leakage hidden by random pair sampling.","status":"source_checked","source_ids":["mrna-protein-diversity-2026"],"source_locator":"Datasets and methods; ground-truth dataset construction; cached text lines 8–13; task metric definitions and corresponding results table; matching task comparison table/ablation captions"},{"label":"Uncertainty","value":"The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","status":"unreported","source_ids":["mrna-protein-diversity-2026"],"source_locator":"Datasets and methods; ground-truth dataset construction; cached text lines 8–13; task metric definitions and corresponding results table; matching task comparison table/ablation captions"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["mrna-protein-diversity-2026"],"source_locator":"Datasets and methods; ground-truth dataset construction; cached text lines 8–13; task metric definitions and corresponding results table; matching task comparison table/ablation captions"},{"label":"Organisms","value":"Human.","status":"source_checked","source_ids":["mrna-protein-diversity-2026"],"source_locator":"Datasets and methods; ground-truth dataset construction; cached text lines 8–13; task metric definitions and corresponding results table; matching task comparison table/ablation captions"},{"label":"Assays","value":"CLIPdb RNA-binding annotations.","status":"source_checked","source_ids":["mrna-protein-diversity-2026"],"source_locator":"Datasets and methods; ground-truth dataset construction; cached text lines 8–13; task metric definitions and corresponding results table; matching task comparison table/ablation captions"},{"label":"Allowed inputs","value":"RNA–protein pairs.","status":"source_checked","source_ids":["mrna-protein-diversity-2026"],"source_locator":"Datasets and methods; ground-truth dataset construction; cached text lines 8–13; task metric definitions and corresponding results table; matching task comparison table/ablation captions"},{"label":"Adaptation","value":"Supervised interaction prediction; random-pair and held-out-protein tests measure different transfer settings.","status":"source_checked","source_ids":["mrna-protein-diversity-2026"],"source_locator":"Datasets and methods; ground-truth dataset construction; cached text lines 8–13; task metric definitions and corresponding results table; matching task comparison table/ablation captions"}],"strengths":[{"text":"RBP-aware evaluation separates unseen-protein transfer from easier random-pair testing.","source_ids":["mrna-protein-diversity-2026"],"source_locator":"Datasets and methods; ground-truth dataset construction; cached text lines 8–13; task metric definitions and corresponding results table; matching task comparison table/ablation captions"}],"limitations":[{"text":"Unobserved interactions are used to construct negatives and may include undiscovered positives. Classifier configuration must remain attached to the partitioning regime.","source_ids":["mrna-protein-diversity-2026"],"source_locator":"Datasets and methods; ground-truth dataset construction; cached text lines 8–13; task metric definitions and corresponding results table; matching task comparison table/ablation captions"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: RNA–protein pairs.","Evaluation: Supervised interaction prediction; random-pair and held-out-protein tests measure different transfer settings.","Readout: AUROC, AUPRC, F1, precision, recall and specificity, reported separately for each partitioning regime."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["mrna-protein-diversity-2026"],"source_locator":"Datasets and methods; ground-truth dataset construction; cached text lines 8–13; task metric definitions and corresponding results table; matching task comparison table/ablation captions"},"coverage":"limited","gaps":["Unobserved interactions are used to construct negatives and may include undiscovered positives. Classifier configuration must remain attached to the partitioning regime."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Task-specific computational methodology and field context checked in the cited primary-source artifact. Source-backed fields, inapplicable evaluator dimensions and unresolved details are distinguished. Numerical results were not reproduced."}}}}
{"id":"reported-task-d81be76396e644","kind":"benchmark","name":"Protein–ligand binding affinity prediction","description":"","status":"needs_review","facets":{"areas":["molecular-interactions"],"tasks":["Protein–ligand binding affinity prediction"]},"source_ids":["deelig-2021"],"links":[{"relation":"dataset","target_id":"reported-dataset-327cfcdae0c937"}],"attributes":{"entity_level":"task","version":null,"task":"Protein–ligand binding affinity prediction","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Protein–ligand affinity regression uses curated structural complexes and a held-out fraction of the assembled dataset.","summary_source_ids":["deelig-2021"],"summary_source_locator":"Methods: Novel data set; Data set refinement; model input construction; cached text lines 10–17, 41; task metric definitions and corresponding results table; matching task comparison table/ablation captions","sections":[{"title":"Evaluation methodology","body":"RCSB PDB-derived chain–ligand pairs with binding-affinity labels, refined to pocket–ligand examples. The feature dataset is divided into training, validation and test partitions in an 80:10:10 ratio. Mean absolute error and root mean squared error for affinity prediction. Atomic versus composite feature models; the PDBbind-core comparison also lists AutoDock Vina, RF::VinaElem, TOPBP and AGL Score. The checked dataset section gives partition proportions but no scaffold- or protein-target-disjoint assignment rule. The reported SD describes dispersion associated with real/predicted values. It is not identified as a confidence interval or independent retraining uncertainty.","source_ids":["deelig-2021"],"source_locator":"Methods: Novel data set; Data set refinement; model input construction; cached text lines 10–17, 41; task metric definitions and corresponding results table; matching task comparison table/ablation captions"}],"facts":[{"label":"Datasets","value":"RCSB PDB-derived chain–ligand pairs with binding-affinity labels, refined to pocket–ligand examples.","status":"source_checked","source_ids":["deelig-2021"],"source_locator":"Methods: Novel data set; Data set refinement; model input construction; cached text lines 10–17, 41; task metric definitions and corresponding results table; matching task comparison table/ablation captions"},{"label":"Splits","value":"The feature dataset is divided into training, validation and test partitions in an 80:10:10 ratio.","status":"source_checked","source_ids":["deelig-2021"],"source_locator":"Methods: Novel data set; Data set refinement; model input construction; cached text lines 10–17, 41; task metric definitions and corresponding results table; matching task comparison table/ablation captions"},{"label":"Metrics","value":"Mean absolute error and root mean squared error for affinity prediction.","status":"source_checked","source_ids":["deelig-2021"],"source_locator":"Methods: Novel data set; Data set refinement; model input construction; cached text lines 10–17, 41; task metric definitions and corresponding results table; matching task comparison table/ablation captions"},{"label":"Baselines","value":"Atomic versus composite feature models; the PDBbind-core comparison also lists AutoDock Vina, RF::VinaElem, TOPBP and AGL Score.","status":"source_checked","source_ids":["deelig-2021"],"source_locator":"Methods: Novel data set; Data set refinement; model input construction; cached text lines 10–17, 41; task metric definitions and corresponding results table; matching task comparison table/ablation captions"},{"label":"Leakage controls","value":"The checked dataset section gives partition proportions but no scaffold- or protein-target-disjoint assignment rule.","status":"unreported","source_ids":["deelig-2021"],"source_locator":"Methods: Novel data set; Data set refinement; model input construction; cached text lines 10–17, 41; task metric definitions and corresponding results table; matching task comparison table/ablation captions"},{"label":"Uncertainty","value":"The reported SD describes dispersion associated with real/predicted values. It is not identified as a confidence interval or independent retraining uncertainty.","status":"source_checked","source_ids":["deelig-2021"],"source_locator":"Methods: Novel data set; Data set refinement; model input construction; cached text lines 10–17, 41; task metric definitions and corresponding results table; matching task comparison table/ablation captions"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["deelig-2021"],"source_locator":"Methods: Novel data set; Data set refinement; model input construction; cached text lines 10–17, 41; task metric definitions and corresponding results table; matching task comparison table/ablation captions"},{"label":"Organisms","value":"The PDB query selects protein–ligand structures with measured affinity and crystallographic criteria. The raw-data and refinement sections specify no taxonomic selection or organism-count table for the retained chain–ligand pairs.","status":"unreported","source_ids":["deelig-2021"],"source_locator":"Novel data set: raw data; Data set refinement"},{"label":"Assays","value":"Protein–ligand structural and binding-affinity labels.","status":"source_checked","source_ids":["deelig-2021"],"source_locator":"Methods: Novel data set; Data set refinement; model input construction; cached text lines 10–17, 41; task metric definitions and corresponding results table; matching task comparison table/ablation captions"},{"label":"Allowed inputs","value":"Pocket–ligand structural features.","status":"source_checked","source_ids":["deelig-2021"],"source_locator":"Methods: Novel data set; Data set refinement; model input construction; cached text lines 10–17, 41; task metric definitions and corresponding results table; matching task comparison table/ablation captions"},{"label":"Adaptation","value":"Supervised affinity regression on the training partition.","status":"source_checked","source_ids":["deelig-2021"],"source_locator":"Methods: Novel data set; Data set refinement; model input construction; cached text lines 10–17, 41; task metric definitions and corresponding results table; matching task comparison table/ablation captions"}],"strengths":[],"limitations":[{"text":"Random pair splitting does not establish scaffold- or target-disjoint generalization; chain-level examples may share protein or ligand information.","source_ids":["deelig-2021"],"source_locator":"Methods: Novel data set; Data set refinement; model input construction; cached text lines 10–17, 41; task metric definitions and corresponding results table; matching task comparison table/ablation captions"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Pocket–ligand structural features.","Evaluation: Supervised affinity regression on the training partition.","Readout: Mean absolute error and root mean squared error for affinity prediction."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["deelig-2021"],"source_locator":"Methods: Novel data set; Data set refinement; model input construction; cached text lines 10–17, 41; task metric definitions and corresponding results table; matching task comparison table/ablation captions"},"coverage":"limited","gaps":["Leakage controls: The checked dataset section gives partition proportions but no scaffold- or protein-target-disjoint assignment rule.","Organisms: The PDB query selects protein–ligand structures with measured affinity and crystallographic criteria. The raw-data and refinement sections specify no taxonomic selection or organism-count table for the retained chain–ligand pairs."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Relevant full-paper computational evaluation sections, tables/captions and cited supplementary task passages were reviewed. Reporting omissions are scoped to the inspected sources. Original numerical results are unchanged."}}}}
{"id":"reported-task-d82b6284f3f431","kind":"benchmark","name":"Cross-platform scATAC cell-type annotation","description":"","status":"needs_review","facets":{"areas":["cells-tissues"],"tasks":["Cross-platform scATAC cell-type annotation"]},"source_ids":["scatac-llmda-2026"],"links":[{"relation":"dataset","target_id":"reported-dataset-7fc59ce4c0ceaa"}],"attributes":{"entity_level":"task","version":null,"task":"Cross-platform scATAC cell-type annotation","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Cross-platform cell annotation transfers labels between scATAC-seq reference and query datasets.","summary_source_ids":["scatac-llmda-2026"],"summary_source_locator":"Methods: Benchmark datasets; Benchmark methods; Problem definition; cached text lines 9–17; task metric definitions and corresponding results table","sections":[{"title":"Evaluation methodology","body":"Mouse brain accessibility datasets from multiple platforms and genome-reference versions. Source/reference and target/query domains are evaluated under platform and tissue distribution shifts. Accuracy and F1 for cell-type annotation, stratified by source-to-target transfer task. scNym, scJoint, Cellcano, SANGO, annATAC, AtacAnnoR and MINGLE. The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","source_ids":["scatac-llmda-2026"],"source_locator":"Methods: Benchmark datasets; Benchmark methods; Problem definition; cached text lines 9–17; task metric definitions and corresponding results table"}],"facts":[{"label":"Datasets","value":"Mouse brain accessibility datasets from multiple platforms and genome-reference versions.","status":"source_checked","source_ids":["scatac-llmda-2026"],"source_locator":"Methods: Benchmark datasets; Benchmark methods; Problem definition; cached text lines 9–17; task metric definitions and corresponding results table"},{"label":"Splits","value":"Source/reference and target/query domains are evaluated under platform and tissue distribution shifts.","status":"source_checked","source_ids":["scatac-llmda-2026"],"source_locator":"Methods: Benchmark datasets; Benchmark methods; Problem definition; cached text lines 9–17; task metric definitions and corresponding results table"},{"label":"Metrics","value":"Accuracy and F1 for cell-type annotation, stratified by source-to-target transfer task.","status":"source_checked","source_ids":["scatac-llmda-2026"],"source_locator":"Methods: Benchmark datasets; Benchmark methods; Problem definition; cached text lines 9–17; task metric definitions and corresponding results table"},{"label":"Baselines","value":"scNym, scJoint, Cellcano, SANGO, annATAC, AtacAnnoR and MINGLE.","status":"source_checked","source_ids":["scatac-llmda-2026"],"source_locator":"Methods: Benchmark datasets; Benchmark methods; Problem definition; cached text lines 9–17; task metric definitions and corresponding results table"},{"label":"Leakage controls","value":"This is transductive domain adaptation: both source and target cells enter the graph-training stage, while the stated classification loss uses labelled source nodes. Target-domain access is part of the protocol and should not be represented as an untouched-query inductive test.","status":"source_checked","source_ids":["scatac-llmda-2026"],"source_locator":"Methods: overall loss, source classification loss, domain-adversarial loss and Parameter settings; cached paragraphs 60–69"},{"label":"Uncertainty","value":"The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","status":"unreported","source_ids":["scatac-llmda-2026"],"source_locator":"Methods: Benchmark datasets; Benchmark methods; Problem definition; cached text lines 9–17; task metric definitions and corresponding results table"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["scatac-llmda-2026"],"source_locator":"Methods: Benchmark datasets; Benchmark methods; Problem definition; cached text lines 9–17; task metric definitions and corresponding results table"},{"label":"Organisms","value":"Mouse brain.","status":"source_checked","source_ids":["scatac-llmda-2026"],"source_locator":"Methods: Benchmark datasets; Benchmark methods; Problem definition; cached text lines 9–17; task metric definitions and corresponding results table"},{"label":"Assays","value":"Single-cell ATAC-seq across platforms and genome-reference versions.","status":"source_checked","source_ids":["scatac-llmda-2026"],"source_locator":"Methods: Benchmark datasets; Benchmark methods; Problem definition; cached text lines 9–17; task metric definitions and corresponding results table"},{"label":"Allowed inputs","value":"Chromatin-accessibility representations in source/reference and target/query domains.","status":"source_checked","source_ids":["scatac-llmda-2026"],"source_locator":"Methods: Benchmark datasets; Benchmark methods; Problem definition; cached text lines 9–17; task metric definitions and corresponding results table"},{"label":"Adaptation","value":"Cross-platform annotation transfer compared with specialized annotation methods.","status":"source_checked","source_ids":["scatac-llmda-2026"],"source_locator":"Methods: Benchmark datasets; Benchmark methods; Problem definition; cached text lines 9–17; task metric definitions and corresponding results table"}],"strengths":[],"limitations":[{"text":"Unlabelled target cells participate in adaptation. Comparators differ in feature modalities and access to reference information, which must remain attached to their reported results.","source_ids":["scatac-llmda-2026"],"source_locator":"Methods: Benchmark datasets; Benchmark methods; Problem definition; cached text lines 9–17; task metric definitions and corresponding results table"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Chromatin-accessibility representations in source/reference and target/query domains.","Evaluation: Cross-platform annotation transfer compared with specialized annotation methods.","Readout: Accuracy and F1 for cell-type annotation, stratified by source-to-target transfer task."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["scatac-llmda-2026"],"source_locator":"Methods: Benchmark datasets; Benchmark methods; Problem definition; cached text lines 9–17; task metric definitions and corresponding results table"},"coverage":"limited","gaps":["Uncertainty: The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Relevant full-paper computational evaluation sections, tables/captions and cited supplementary task passages were reviewed. Reporting omissions are scoped to the inspected sources. Original numerical results are unchanged."}}}}
{"id":"reported-task-dc82fcbfb44935","kind":"benchmark","name":"RNA secondary structure","description":"","status":"needs_review","facets":{"areas":["rna-transcriptomes"],"tasks":["RNA secondary structure"]},"source_ids":["bpfold-2025"],"links":[{"relation":"dataset","target_id":"reported-dataset-8317793f18b026"}],"attributes":{"entity_level":"task","version":null,"task":"RNA secondary structure","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"RNA secondary-structure prediction is evaluated separately for sequence-level and RNA-family generalization.","summary_source_ids":["bpfold-2025"],"summary_source_locator":"Introduction; Methods: Datasets and evaluation; cached text lines 9, 44, 74–77; matching task comparison table/ablation captions","sections":[{"title":"Evaluation methodology","body":"ArchiveII, bpRNA-TS0, Rfam12.3–14.10 and experimentally grounded PDB50 evaluation collections. The paper distinguishes sequence-wise assessment from cross-family evaluation. Macro-averaged precision, recall, F1 and interaction network fidelity on canonical base-pair predictions. SPOT-RNA, MXfold2, ContextFold, CONTRAfold, EternaFold, LinearFold, RNAfold, SimFold and RNAstructure. A model confidence index is correlated with observed F1; that diagnostic is not a confidence interval for benchmark performance.","source_ids":["bpfold-2025"],"source_locator":"Introduction; Methods: Datasets and evaluation; cached text lines 9, 44, 74–77; matching task comparison table/ablation captions"}],"facts":[{"label":"Datasets","value":"ArchiveII, bpRNA-TS0, Rfam12.3–14.10 and experimentally grounded PDB50 evaluation collections.","status":"source_checked","source_ids":["bpfold-2025"],"source_locator":"Introduction; Methods: Datasets and evaluation; cached text lines 9, 44, 74–77; matching task comparison table/ablation captions"},{"label":"Splits","value":"The paper distinguishes sequence-wise assessment from cross-family evaluation.","status":"source_checked","source_ids":["bpfold-2025"],"source_locator":"Introduction; Methods: Datasets and evaluation; cached text lines 9, 44, 74–77; matching task comparison table/ablation captions"},{"label":"Metrics","value":"Macro-averaged precision, recall, F1 and interaction network fidelity on canonical base-pair predictions.","status":"source_checked","source_ids":["bpfold-2025"],"source_locator":"Introduction; Methods: Datasets and evaluation; cached text lines 9, 44, 74–77; matching task comparison table/ablation captions"},{"label":"Baselines","value":"SPOT-RNA, MXfold2, ContextFold, CONTRAfold, EternaFold, LinearFold, RNAfold, SimFold and RNAstructure.","status":"source_checked","source_ids":["bpfold-2025"],"source_locator":"Introduction; Methods: Datasets and evaluation; cached text lines 9, 44, 74–77; matching task comparison table/ablation captions"},{"label":"Leakage controls","value":"The bpRNA benchmark applies an 80% sequence-similarity filter. Family-wise testing uses newly added Rfam families absent from the bpRNA training collection and separately removes similar sequences at 80%. Sequence-wise and unseen-family results therefore measure different forms of generalization.","status":"source_checked","source_ids":["bpfold-2025"],"source_locator":"Methods dataset list in full XML; Results: Evaluating BPfold on family-wise datasets; Tables 1–2"},{"label":"Uncertainty","value":"A model confidence index is correlated with observed F1; that diagnostic is not a confidence interval for benchmark performance.","status":"source_checked","source_ids":["bpfold-2025"],"source_locator":"Introduction; Methods: Datasets and evaluation; cached text lines 9, 44, 74–77; matching task comparison table/ablation captions"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["bpfold-2025"],"source_locator":"Introduction; Methods: Datasets and evaluation; cached text lines 9, 44, 74–77; matching task comparison table/ablation captions"},{"label":"Organisms","value":"The evaluation pools RNA-family datasets and PDB RNA structures. The Methods dataset list enumerates RNA families and sequence sets, not taxa. Individual bacterial and viral examples in Figure 6 do not establish the species composition of the aggregate benchmark.","status":"unreported","source_ids":["bpfold-2025"],"source_locator":"Methods dataset list; Tables 1–2; Fig.6 caption"},{"label":"Assays","value":"RNA secondary-structure references including experimentally grounded PDB structures.","status":"source_checked","source_ids":["bpfold-2025"],"source_locator":"Introduction; Methods: Datasets and evaluation; cached text lines 9, 44, 74–77; matching task comparison table/ablation captions"},{"label":"Allowed inputs","value":"RNA sequence.","status":"source_checked","source_ids":["bpfold-2025"],"source_locator":"Introduction; Methods: Datasets and evaluation; cached text lines 9, 44, 74–77; matching task comparison table/ablation captions"},{"label":"Adaptation","value":"Supervised sequence-to-structure prediction with separate sequence-wise and cross-family assessments.","status":"source_checked","source_ids":["bpfold-2025"],"source_locator":"Introduction; Methods: Datasets and evaluation; cached text lines 9, 44, 74–77; matching task comparison table/ablation captions"}],"strengths":[],"limitations":[{"text":"Sequence-wise and family-wise tests are distinct conditions. RNA-family exclusion is not a substitute for a full provenance audit of every training component and comparator.","source_ids":["bpfold-2025"],"source_locator":"Introduction; Methods: Datasets and evaluation; cached text lines 9, 44, 74–77; matching task comparison table/ablation captions"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: RNA sequence.","Evaluation: Supervised sequence-to-structure prediction with separate sequence-wise and cross-family assessments.","Readout: Macro-averaged precision, recall, F1 and interaction network fidelity on canonical base-pair predictions."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["bpfold-2025"],"source_locator":"Introduction; Methods: Datasets and evaluation; cached text lines 9, 44, 74–77; matching task comparison table/ablation captions"},"coverage":"limited","gaps":["Organisms: The evaluation pools RNA-family datasets and PDB RNA structures. The Methods dataset list enumerates RNA families and sequence sets, not taxa. Individual bacterial and viral examples in Figure 6 do not establish the species composition of the aggregate benchmark."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Relevant full-paper computational evaluation sections, tables/captions and cited supplementary task passages were reviewed. Reporting omissions are scoped to the inspected sources. Original numerical results are unchanged."}}}}
{"id":"reported-task-dd001540e0f4ec","kind":"benchmark","name":"Genome-wide prophage detection","description":"","status":"needs_review","facets":{"areas":["microbes-communities"],"tasks":["Genome-wide prophage detection"]},"source_ids":["lambda-prophage-2026"],"links":[{"relation":"dataset","target_id":"reported-dataset-1b4f6ea24c0587"}],"attributes":{"entity_level":"task","version":null,"task":"Genome-wide prophage detection","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Prophage detection is evaluated with grouped data partitions and explicit sequence-composition controls.","summary_source_ids":["lambda-prophage-2026"],"summary_source_locator":"Methods: LAMBDA Benchmark Dataset Construction; Data Splitting and Data Leakage Prevention; control datasets; cached text lines 66–67, 78–85; uncertainty/repeat-run/statistical-comparison passages","sections":[{"title":"Evaluation methodology","body":"Curated phage and bacterial reference-genome collections. Cluster/group-aware 80:10:10 training, development and test split. Separate control collections assess bacterial false positives and phage false negatives. Phage clusters and bacterial genus groups are assigned wholly to one partition. Repeated experiments are summarized with mean and standard deviation for each compared model.","source_ids":["lambda-prophage-2026"],"source_locator":"Methods: LAMBDA Benchmark Dataset Construction; Data Splitting and Data Leakage Prevention; control datasets; cached text lines 66–67, 78–85; uncertainty/repeat-run/statistical-comparison passages"}],"facts":[{"label":"Datasets","value":"Curated phage and bacterial reference-genome collections.","status":"source_checked","source_ids":["lambda-prophage-2026"],"source_locator":"Methods: LAMBDA Benchmark Dataset Construction; Data Splitting and Data Leakage Prevention; control datasets; cached text lines 66–67, 78–85; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Splits","value":"Cluster/group-aware 80:10:10 training, development and test split.","status":"source_checked","source_ids":["lambda-prophage-2026"],"source_locator":"Methods: LAMBDA Benchmark Dataset Construction; Data Splitting and Data Leakage Prevention; control datasets; cached text lines 66–67, 78–85; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Metrics","value":"Separate control collections assess bacterial false positives and phage false negatives.","status":"source_checked","source_ids":["lambda-prophage-2026"],"source_locator":"Methods: LAMBDA Benchmark Dataset Construction; Data Splitting and Data Leakage Prevention; control datasets; cached text lines 66–67, 78–85; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Baselines","value":"Genome-wide comparisons include PHASTER, geNomad, VIBRANT, Phigaro, PhiSpy and VirSorter2, plus protein-language-model-based PIDE. Embedding probes also compare pretrained and randomly initialized representations; those probes are a separate comparison from genome-wide tools.","status":"source_checked","source_ids":["lambda-prophage-2026"],"source_locator":"Comparison with Traditional and Protein-based Models; embedding-probe experiment and Tables 2–3"},{"label":"Leakage controls","value":"Phage clusters and bacterial genus groups are assigned wholly to one partition.","status":"source_checked","source_ids":["lambda-prophage-2026"],"source_locator":"Methods: LAMBDA Benchmark Dataset Construction; Data Splitting and Data Leakage Prevention; control datasets; cached text lines 66–67, 78–85; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Uncertainty","value":"Repeated experiments are summarized with mean and standard deviation for each compared model.","status":"source_checked","source_ids":["lambda-prophage-2026"],"source_locator":"Methods: LAMBDA Benchmark Dataset Construction; Data Splitting and Data Leakage Prevention; control datasets; cached text lines 66–67, 78–85; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["lambda-prophage-2026"],"source_locator":"Methods: LAMBDA Benchmark Dataset Construction; Data Splitting and Data Leakage Prevention; control datasets; cached text lines 66–67, 78–85; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Organisms","value":"Phage and bacterial reference collections.","status":"source_checked","source_ids":["lambda-prophage-2026"],"source_locator":"Methods: LAMBDA Benchmark Dataset Construction; Data Splitting and Data Leakage Prevention; control datasets; cached text lines 66–67, 78–85; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Assays","value":"Reference genome/prophage annotations.","status":"source_checked","source_ids":["lambda-prophage-2026"],"source_locator":"Methods: LAMBDA Benchmark Dataset Construction; Data Splitting and Data Leakage Prevention; control datasets; cached text lines 66–67, 78–85; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Allowed inputs","value":"Genomic sequence.","status":"source_checked","source_ids":["lambda-prophage-2026"],"source_locator":"Methods: LAMBDA Benchmark Dataset Construction; Data Splitting and Data Leakage Prevention; control datasets; cached text lines 66–67, 78–85; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Adaptation","value":"Supervised detection with a cluster/group-aware train/development/test split.","status":"source_checked","source_ids":["lambda-prophage-2026"],"source_locator":"Methods: LAMBDA Benchmark Dataset Construction; Data Splitting and Data Leakage Prevention; control datasets; cached text lines 66–67, 78–85; uncertainty/repeat-run/statistical-comparison passages"}],"strengths":[{"text":"Cluster/group-aware partitioning addresses related-sequence overlap between training and testing.","source_ids":["lambda-prophage-2026"],"source_locator":"Methods: LAMBDA Benchmark Dataset Construction; Data Splitting and Data Leakage Prevention; control datasets; cached text lines 66–67, 78–85; uncertainty/repeat-run/statistical-comparison passages"}],"limitations":[{"text":"Shuffled-sequence controls test sequence-order dependence while preserving composition. Their classification scores and genome-wide localization comparisons are separate endpoints.","source_ids":["lambda-prophage-2026"],"source_locator":"Methods: LAMBDA Benchmark Dataset Construction; Data Splitting and Data Leakage Prevention; control datasets; cached text lines 66–67, 78–85; uncertainty/repeat-run/statistical-comparison passages"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Genomic sequence.","Evaluation: Supervised detection with a cluster/group-aware train/development/test split.","Readout: Separate control collections assess bacterial false positives and phage false negatives."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["lambda-prophage-2026"],"source_locator":"Methods: LAMBDA Benchmark Dataset Construction; Data Splitting and Data Leakage Prevention; control datasets; cached text lines 66–67, 78–85; uncertainty/repeat-run/statistical-comparison passages"},"coverage":"limited","gaps":["Shuffled-sequence controls test sequence-order dependence while preserving composition. Their classification scores and genome-wide localization comparisons are separate endpoints."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Relevant full-paper computational evaluation sections, tables/captions and cited supplementary task passages were reviewed. Reporting omissions are scoped to the inspected sources. Original numerical results are unchanged."}}}}
{"id":"reported-task-dec9e0f5e3da2a","kind":"benchmark","name":"Intrinsically disordered protein ensemble docking","description":"","status":"needs_review","facets":{"areas":["molecular-interactions"],"tasks":["Intrinsically disordered protein ensemble docking"]},"source_ids":["ensemble-idp-docking-2025"],"links":[{"relation":"dataset","target_id":"reported-dataset-00201f65c32f6d"}],"attributes":{"entity_level":"task","version":null,"task":"Intrinsically disordered protein ensemble docking","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Ensemble docking is assessed against molecular-dynamics conformations using two distinct pose-matching definitions.","summary_source_ids":["ensemble-idp-docking-2025"],"summary_source_locator":"Results: cross docking; Methods: RMSD calculations; cached text lines 46, 54, 77–78; uncertainty/repeat-run/statistical-comparison passages","sections":[{"title":"Evaluation methodology","body":"Ligand-bound intrinsically disordered protein ensembles used as computational reference conformations. Cross-docking evaluates a ligand using protein conformations associated with another ligand; this is distinct from a supervised split. Frame-matched ligand RMSD compares corresponding conformations; best-matched RMSD selects the closest reference conformation within a cluster. AutoDock Vina and DiffDock. Normalized docking-score uncertainty uses bootstrap resampling within each docked ensemble; this is distinct from uncertainty across targets.","source_ids":["ensemble-idp-docking-2025"],"source_locator":"Results: cross docking; Methods: RMSD calculations; cached text lines 46, 54, 77–78; uncertainty/repeat-run/statistical-comparison passages"}],"facts":[{"label":"Datasets","value":"Ligand-bound intrinsically disordered protein ensembles used as computational reference conformations.","status":"source_checked","source_ids":["ensemble-idp-docking-2025"],"source_locator":"Results: cross docking; Methods: RMSD calculations; cached text lines 46, 54, 77–78; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Splits","value":"Cross-docking evaluates a ligand using protein conformations associated with another ligand; this is distinct from a supervised split.","status":"source_checked","source_ids":["ensemble-idp-docking-2025"],"source_locator":"Results: cross docking; Methods: RMSD calculations; cached text lines 46, 54, 77–78; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Metrics","value":"Frame-matched ligand RMSD compares corresponding conformations; best-matched RMSD selects the closest reference conformation within a cluster.","status":"source_checked","source_ids":["ensemble-idp-docking-2025"],"source_locator":"Results: cross docking; Methods: RMSD calculations; cached text lines 46, 54, 77–78; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Baselines","value":"AutoDock Vina and DiffDock.","status":"source_checked","source_ids":["ensemble-idp-docking-2025"],"source_locator":"Results: cross docking; Methods: RMSD calculations; cached text lines 46, 54, 77–78; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Leakage controls","value":"Receptor ensembles and reference binding statistics come from the same previously generated molecular-dynamics trajectories. Cross-docking changes the ligand-associated receptor ensemble, rather than defining a supervised train/test split. The inspected docking methods do not audit overlap with DiffDock’s original training structures.","status":"unreported","source_ids":["ensemble-idp-docking-2025"],"source_locator":"Results: apo/holo/cross-docking evaluation; Methods: trajectories, conformation sampling and docking"},{"label":"Uncertainty","value":"Normalized docking-score uncertainty uses bootstrap resampling within each docked ensemble; this is distinct from uncertainty across targets.","status":"source_checked","source_ids":["ensemble-idp-docking-2025"],"source_locator":"Results: cross docking; Methods: RMSD calculations; cached text lines 46, 54, 77–78; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["ensemble-idp-docking-2025"],"source_locator":"Results: cross docking; Methods: RMSD calculations; cached text lines 46, 54, 77–78; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Organisms","value":"The actual evaluation uses a C-terminal α-synuclein fragment and three ligand-associated simulation collections. Its Methods identifies the reused trajectories but does not explicitly give an organism/accession for the simulated protein; references to mouse disease models are not that metadata.","status":"unreported","source_ids":["ensemble-idp-docking-2025"],"source_locator":"Results: α-synuclein evaluation; Methods: MD simulations; Data availability"},{"label":"Assays","value":"Ligand-bound structural ensembles used as computational references.","status":"source_checked","source_ids":["ensemble-idp-docking-2025"],"source_locator":"Results: cross docking; Methods: RMSD calculations; cached text lines 46, 54, 77–78; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Allowed inputs","value":"Ligands and alternative protein conformations.","status":"source_checked","source_ids":["ensemble-idp-docking-2025"],"source_locator":"Results: cross docking; Methods: RMSD calculations; cached text lines 46, 54, 77–78; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Adaptation","value":"Cross-docking with pretrained or conventional docking algorithms; this is not a supervised dataset split.","status":"source_checked","source_ids":["ensemble-idp-docking-2025"],"source_locator":"Results: cross docking; Methods: RMSD calculations; cached text lines 46, 54, 77–78; uncertainty/repeat-run/statistical-comparison passages"}],"strengths":[{"text":"Cross-docking asks whether a method tolerates a receptor conformation associated with a different ligand.","source_ids":["ensemble-idp-docking-2025"],"source_locator":"Results: cross docking; Methods: RMSD calculations; cached text lines 46, 54, 77–78; uncertainty/repeat-run/statistical-comparison passages"}],"limitations":[{"text":"Best-match selection is more permissive than frame-matched evaluation. These MD-referenced results are not automatically experimental pose-validation results.","source_ids":["ensemble-idp-docking-2025"],"source_locator":"Results: cross docking; Methods: RMSD calculations; cached text lines 46, 54, 77–78; uncertainty/repeat-run/statistical-comparison passages"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Ligands and alternative protein conformations.","Evaluation: Cross-docking with pretrained or conventional docking algorithms; this is not a supervised dataset split.","Readout: Frame-matched ligand RMSD compares corresponding conformations; best-matched RMSD selects the closest reference conformation within a cluster."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["ensemble-idp-docking-2025"],"source_locator":"Results: cross docking; Methods: RMSD calculations; cached text lines 46, 54, 77–78; uncertainty/repeat-run/statistical-comparison passages"},"coverage":"limited","gaps":["Leakage controls: Receptor ensembles and reference binding statistics come from the same previously generated molecular-dynamics trajectories. Cross-docking changes the ligand-associated receptor ensemble, rather than defining a supervised train/test split. The inspected docking methods do not audit overlap with DiffDock’s original training structures.","Organisms: The actual evaluation uses a C-terminal α-synuclein fragment and three ligand-associated simulation collections. Its Methods identifies the reused trajectories but does not explicitly give an organism/accession for the simulated protein; references to mouse disease models are not that metadata."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Relevant full-paper computational evaluation sections, tables/captions and cited supplementary task passages were reviewed. Reporting omissions are scoped to the inspected sources. Original numerical results are unchanged."}}}}
{"id":"reported-task-df18c710f45213","kind":"benchmark","name":"RNA sequence design","description":"","status":"needs_review","facets":{"areas":["rna-transcriptomes"],"tasks":["RNA sequence design"]},"source_ids":["r3design-2025"],"links":[{"relation":"dataset","target_id":"reported-dataset-71614d99b3099f"}],"attributes":{"entity_level":"task","version":null,"task":"RNA sequence design","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"RNA inverse-folding evaluation uses structure-grouped partitions and computational structure-recovery checks.","summary_source_ids":["r3design-2025"],"summary_source_locator":"Results: benchmark dataset and structural assessment; cached text lines 10, 31–33","sections":[{"title":"Evaluation methodology","body":"Nonredundant RNA structures from PDB and RNASolo. Training, validation and test allocations are based on structural similarity. Predicted-structure RMSD is among the evaluated outcomes. Structural similarity informs the dataset partition. Three experiments with different seeds report mean and standard deviation.","source_ids":["r3design-2025"],"source_locator":"Results: benchmark dataset and structural assessment; cached text lines 10, 31–33"}],"facts":[{"label":"Datasets","value":"Nonredundant RNA structures from PDB and RNASolo.","status":"source_checked","source_ids":["r3design-2025"],"source_locator":"Results: benchmark dataset and structural assessment; cached text lines 10, 31–33"},{"label":"Splits","value":"Training, validation and test allocations are based on structural similarity.","status":"source_checked","source_ids":["r3design-2025"],"source_locator":"Results: benchmark dataset and structural assessment; cached text lines 10, 31–33"},{"label":"Metrics","value":"Predicted-structure RMSD is among the evaluated outcomes.","status":"source_checked","source_ids":["r3design-2025"],"source_locator":"Results: benchmark dataset and structural assessment; cached text lines 10, 31–33"},{"label":"Baselines","value":"The core structural benchmark compares SeqRNN and SeqLSTM at two hidden sizes, StructMLP, StructGNN, GraphTrans and PiFold. Separate Rfam/RNA-Puzzles tables include secondary-structure-oriented methods; their scores should retain their own dataset and metric context.","status":"source_checked","source_ids":["r3design-2025"],"source_locator":"Tables 1–3: method-column entries; computational evaluation results"},{"label":"Leakage controls","value":"Structural similarity informs the dataset partition.","status":"source_checked","source_ids":["r3design-2025"],"source_locator":"Results: benchmark dataset and structural assessment; cached text lines 10, 31–33"},{"label":"Uncertainty","value":"Three experiments with different seeds report mean and standard deviation.","status":"source_checked","source_ids":["r3design-2025"],"source_locator":"Results: benchmark dataset and structural assessment; cached text lines 10, 31–33"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["r3design-2025"],"source_locator":"Results: benchmark dataset and structural assessment; cached text lines 10, 31–33"},{"label":"Organisms","value":"RNA structures are selected from PDB/RNASolo, with separate Rfam and RNA-Puzzles evaluations. The curation description reports RNA structural diversity rather than organism membership or species-stratified scores.","status":"unreported","source_ids":["r3design-2025"],"source_locator":"Results: benchmark-data curation; Tables 1–3; Data availability"},{"label":"Assays","value":"Structural reference backbones and associated RNA sequences.","status":"source_checked","source_ids":["r3design-2025"],"source_locator":"Results: benchmark dataset and structural assessment; cached text lines 10, 31–33"},{"label":"Allowed inputs","value":"RNA structural context for sequence prediction.","status":"source_checked","source_ids":["r3design-2025"],"source_locator":"Results: benchmark dataset and structural assessment; cached text lines 10, 31–33"},{"label":"Adaptation","value":"Task training on structurally partitioned data; evaluated with recovery/perplexity and structural consistency criteria.","status":"source_checked","source_ids":["r3design-2025"],"source_locator":"Results: benchmark dataset and structural assessment; cached text lines 10, 31–33"}],"strengths":[],"limitations":[{"text":"Structure-predictor agreement is a computational proxy for folding, not experimental validation; predictor dependence remains an evaluation limitation.","source_ids":["r3design-2025"],"source_locator":"Results: benchmark dataset and structural assessment; cached text lines 10, 31–33"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: RNA structural context for sequence prediction.","Evaluation: Task training on structurally partitioned data; evaluated with recovery/perplexity and structural consistency criteria.","Readout: Predicted-structure RMSD is among the evaluated outcomes."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["r3design-2025"],"source_locator":"Results: benchmark dataset and structural assessment; cached text lines 10, 31–33"},"coverage":"limited","gaps":["Organisms: RNA structures are selected from PDB/RNASolo, with separate Rfam and RNA-Puzzles evaluations. The curation description reports RNA structural diversity rather than organism membership or species-stratified scores."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Relevant full-paper computational evaluation sections, tables/captions and cited supplementary task passages were reviewed. Reporting omissions are scoped to the inspected sources. Original numerical results are unchanged."}}}}
{"id":"reported-task-dfa8f2285dbfa5","kind":"benchmark","name":"protein-protein interaction prediction","description":"","status":"needs_review","facets":{"areas":["proteins-complexes"],"tasks":["protein-protein interaction prediction"]},"source_ids":["gsmformer-ppi-2026"],"links":[{"relation":"dataset","target_id":"reported-dataset-07d355c146be1f"}],"attributes":{"entity_level":"task","version":null,"task":"protein-protein interaction prediction","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Protein-pair interaction classification derives positive and sampled-negative pairs from PINDER structural partitions.","summary_source_ids":["gsmformer-ppi-2026"],"summary_source_locator":"Methods: Splits; Sampling; cached text lines 15–21; task metric definitions and corresponding results table; uncertainty/repeat-run/statistical-comparison passages; matching task comparison table/ablation captions","sections":[{"title":"Evaluation methodology","body":"Filtered PINDER dimeric systems and negative pairs checked against BioGRID. The ML-ready PINDER training, validation and test partitions are retained during task construction. Accuracy, sensitivity, F1, MCC, AUROC and AUPRC for protein-pair classification. ESM-2, ProtT5 and ProstT5 sequence-embedding variants; graph/surface-feature removals are separate ablations. PINDER uses interface/sequence similarity clustering; sampled negatives are checked against known interactions. Three-fold cross-validation with different random seeds is reported using means and standard deviations across runs.","source_ids":["gsmformer-ppi-2026"],"source_locator":"Methods: Splits; Sampling; cached text lines 15–21; task metric definitions and corresponding results table; uncertainty/repeat-run/statistical-comparison passages; matching task comparison table/ablation captions"}],"facts":[{"label":"Datasets","value":"Filtered PINDER dimeric systems and negative pairs checked against BioGRID.","status":"source_checked","source_ids":["gsmformer-ppi-2026"],"source_locator":"Methods: Splits; Sampling; cached text lines 15–21; task metric definitions and corresponding results table; uncertainty/repeat-run/statistical-comparison passages; matching task comparison table/ablation captions"},{"label":"Splits","value":"The ML-ready PINDER training, validation and test partitions are retained during task construction.","status":"source_checked","source_ids":["gsmformer-ppi-2026"],"source_locator":"Methods: Splits; Sampling; cached text lines 15–21; task metric definitions and corresponding results table; uncertainty/repeat-run/statistical-comparison passages; matching task comparison table/ablation captions"},{"label":"Metrics","value":"Accuracy, sensitivity, F1, MCC, AUROC and AUPRC for protein-pair classification.","status":"source_checked","source_ids":["gsmformer-ppi-2026"],"source_locator":"Methods: Splits; Sampling; cached text lines 15–21; task metric definitions and corresponding results table; uncertainty/repeat-run/statistical-comparison passages; matching task comparison table/ablation captions"},{"label":"Baselines","value":"ESM-2, ProtT5 and ProstT5 sequence-embedding variants; graph/surface-feature removals are separate ablations.","status":"source_checked","source_ids":["gsmformer-ppi-2026"],"source_locator":"Methods: Splits; Sampling; cached text lines 15–21; task metric definitions and corresponding results table; uncertainty/repeat-run/statistical-comparison passages; matching task comparison table/ablation captions"},{"label":"Leakage controls","value":"PINDER uses interface/sequence similarity clustering; sampled negatives are checked against known interactions.","status":"source_checked","source_ids":["gsmformer-ppi-2026"],"source_locator":"Methods: Splits; Sampling; cached text lines 15–21; task metric definitions and corresponding results table; uncertainty/repeat-run/statistical-comparison passages; matching task comparison table/ablation captions"},{"label":"Uncertainty","value":"Three-fold cross-validation with different random seeds is reported using means and standard deviations across runs.","status":"source_checked","source_ids":["gsmformer-ppi-2026"],"source_locator":"Methods: Splits; Sampling; cached text lines 15–21; task metric definitions and corresponding results table; uncertainty/repeat-run/statistical-comparison passages; matching task comparison table/ablation captions"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["gsmformer-ppi-2026"],"source_locator":"Methods: Splits; Sampling; cached text lines 15–21; task metric definitions and corresponding results table; uncertainty/repeat-run/statistical-comparison passages; matching task comparison table/ablation captions"},{"label":"Organisms","value":"The main task inherits filtered PINDER dimer systems; its Dataset section does not enumerate source organisms. The named human, yeast, bacterial and worm species in the external Baranwal evaluation describe that separate dataset and cannot be assigned as the full PINDER inventory.","status":"unreported","source_ids":["gsmformer-ppi-2026"],"source_locator":"Methods: Dataset/Sampling; external Baranwal evaluation"},{"label":"Assays","value":"PINDER interacting dimers and BioGRID-checked negative pairs.","status":"source_checked","source_ids":["gsmformer-ppi-2026"],"source_locator":"Methods: Splits; Sampling; cached text lines 15–21; task metric definitions and corresponding results table; uncertainty/repeat-run/statistical-comparison passages; matching task comparison table/ablation captions"},{"label":"Allowed inputs","value":"Protein-pair representations.","status":"source_checked","source_ids":["gsmformer-ppi-2026"],"source_locator":"Methods: Splits; Sampling; cached text lines 15–21; task metric definitions and corresponding results table; uncertainty/repeat-run/statistical-comparison passages; matching task comparison table/ablation captions"},{"label":"Adaptation","value":"Supervised interaction prediction using retained PINDER training/validation/test partitions.","status":"source_checked","source_ids":["gsmformer-ppi-2026"],"source_locator":"Methods: Splits; Sampling; cached text lines 15–21; task metric definitions and corresponding results table; uncertainty/repeat-run/statistical-comparison passages; matching task comparison table/ablation captions"}],"strengths":[],"limitations":[{"text":"Absence of a sampled protein pair from BioGRID does not prove noninteraction. Random negative construction makes the selected pair lists part of the reproducible protocol.","source_ids":["gsmformer-ppi-2026"],"source_locator":"Methods: Splits; Sampling; cached text lines 15–21; task metric definitions and corresponding results table; uncertainty/repeat-run/statistical-comparison passages; matching task comparison table/ablation captions"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Protein-pair representations.","Evaluation: Supervised interaction prediction using retained PINDER training/validation/test partitions.","Readout: Accuracy, sensitivity, F1, MCC, AUROC and AUPRC for protein-pair classification."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["gsmformer-ppi-2026"],"source_locator":"Methods: Splits; Sampling; cached text lines 15–21; task metric definitions and corresponding results table; uncertainty/repeat-run/statistical-comparison passages; matching task comparison table/ablation captions"},"coverage":"limited","gaps":["Organisms: The main task inherits filtered PINDER dimer systems; its Dataset section does not enumerate source organisms. The named human, yeast, bacterial and worm species in the external Baranwal evaluation describe that separate dataset and cannot be assigned as the full PINDER inventory."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Relevant full-paper computational evaluation sections, tables/captions and cited supplementary task passages were reviewed. Reporting omissions are scoped to the inspected sources. Original numerical results are unchanged."}}}}
{"id":"reported-task-e2009c35eabd69","kind":"benchmark","name":"microbiome disease-state classification","description":"","status":"needs_review","facets":{"areas":["microbes-communities"],"tasks":["microbiome disease-state classification"]},"source_ids":["mdl4microbiome-2022"],"links":[{"relation":"dataset","target_id":"reported-dataset-bd3f98e2eeb5d3"}],"attributes":{"entity_level":"task","version":null,"task":"microbiome disease-state classification","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Microbiome disease-state classification evaluates independent disease/control cohorts through leave-one-out prediction.","summary_source_ids":["mdl4microbiome-2022"],"summary_source_locator":"Methods: Data preparation and preprocessing; Performance evaluation; cached text lines 10–12, 28–30","sections":[{"title":"Evaluation methodology","body":"Four gut-microbiome datasets with IBD, type-2 diabetes, liver-cirrhosis or colorectal-cancer labels and controls. Leave-one-out cross-validation excludes the held-out sample from both embedding training and final classifier training. Accuracy, precision and recall. Random forest, XGBoost, principal-component regression, lasso and SVM. The source explicitly removes the held-out sample from both learning stages. The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","source_ids":["mdl4microbiome-2022"],"source_locator":"Methods: Data preparation and preprocessing; Performance evaluation; cached text lines 10–12, 28–30"}],"facts":[{"label":"Datasets","value":"Four gut-microbiome datasets with IBD, type-2 diabetes, liver-cirrhosis or colorectal-cancer labels and controls.","status":"source_checked","source_ids":["mdl4microbiome-2022"],"source_locator":"Methods: Data preparation and preprocessing; Performance evaluation; cached text lines 10–12, 28–30"},{"label":"Splits","value":"Leave-one-out cross-validation excludes the held-out sample from both embedding training and final classifier training.","status":"source_checked","source_ids":["mdl4microbiome-2022"],"source_locator":"Methods: Data preparation and preprocessing; Performance evaluation; cached text lines 10–12, 28–30"},{"label":"Metrics","value":"Accuracy, precision and recall.","status":"source_checked","source_ids":["mdl4microbiome-2022"],"source_locator":"Methods: Data preparation and preprocessing; Performance evaluation; cached text lines 10–12, 28–30"},{"label":"Baselines","value":"Random forest, XGBoost, principal-component regression, lasso and SVM.","status":"source_checked","source_ids":["mdl4microbiome-2022"],"source_locator":"Methods: Data preparation and preprocessing; Performance evaluation; cached text lines 10–12, 28–30"},{"label":"Leakage controls","value":"The source explicitly removes the held-out sample from both learning stages.","status":"source_checked","source_ids":["mdl4microbiome-2022"],"source_locator":"Methods: Data preparation and preprocessing; Performance evaluation; cached text lines 10–12, 28–30"},{"label":"Uncertainty","value":"The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","status":"unreported","source_ids":["mdl4microbiome-2022"],"source_locator":"Methods: Data preparation and preprocessing; Performance evaluation; cached text lines 10–12, 28–30"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["mdl4microbiome-2022"],"source_locator":"Methods: Data preparation and preprocessing; Performance evaluation; cached text lines 10–12, 28–30"},{"label":"Organisms","value":"Human gut microbial communities.","status":"source_checked","source_ids":["mdl4microbiome-2022"],"source_locator":"Methods: Data preparation and preprocessing; Performance evaluation; cached text lines 10–12, 28–30"},{"label":"Assays","value":"Microbiome profiles with disease/control labels.","status":"source_checked","source_ids":["mdl4microbiome-2022"],"source_locator":"Methods: Data preparation and preprocessing; Performance evaluation; cached text lines 10–12, 28–30"},{"label":"Allowed inputs","value":"Microbiome composition/features.","status":"source_checked","source_ids":["mdl4microbiome-2022"],"source_locator":"Methods: Data preparation and preprocessing; Performance evaluation; cached text lines 10–12, 28–30"},{"label":"Adaptation","value":"Embedding and classifier training are repeated inside leave-one-out folds, excluding the held-out subject sample.","status":"source_checked","source_ids":["mdl4microbiome-2022"],"source_locator":"Methods: Data preparation and preprocessing; Performance evaluation; cached text lines 10–12, 28–30"}],"strengths":[{"text":"The held-out sample is excluded from both representation fitting and the final classifier.","source_ids":["mdl4microbiome-2022"],"source_locator":"Methods: Data preparation and preprocessing; Performance evaluation; cached text lines 10–12, 28–30"}],"limitations":[{"text":"Within-cohort leave-one-out performance does not establish transfer across cohorts or clinical validity. Cohort confounding remain unextracted.","source_ids":["mdl4microbiome-2022"],"source_locator":"Methods: Data preparation and preprocessing; Performance evaluation; cached text lines 10–12, 28–30"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Microbiome composition/features.","Evaluation: Embedding and classifier training are repeated inside leave-one-out folds, excluding the held-out subject sample.","Readout: Accuracy, precision and recall."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["mdl4microbiome-2022"],"source_locator":"Methods: Data preparation and preprocessing; Performance evaluation; cached text lines 10–12, 28–30"},"coverage":"limited","gaps":["Within-cohort leave-one-out performance does not establish transfer across cohorts or clinical validity. Cohort confounding remain unextracted."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Task-specific computational methodology and field context checked in the cited primary-source artifact. Source-backed fields, inapplicable evaluator dimensions and unresolved details are distinguished. Numerical results were not reproduced."}}}}
{"id":"reported-task-e5c34f686ac403","kind":"benchmark","name":"E. coli sigma70 promoter prediction","description":"","status":"needs_review","facets":{"areas":["microbes-communities"],"tasks":["E. coli sigma70 promoter prediction"]},"source_ids":["cyaprombert-2022"],"links":[{"relation":"dataset","target_id":"reported-dataset-a1da4a37eb46a5"}],"attributes":{"entity_level":"task","version":null,"task":"E. coli sigma70 promoter prediction","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"The E. coli promoter benchmark is an independent comparison within a broader promoter-prediction paper.","summary_source_ids":["cyaprombert-2022"],"summary_source_locator":"Datasets; Evaluation criteria; Results: independent datasets from E. coli; cached text lines 12, 28–30, 42; comparative evaluation and ablation passages","sections":[{"title":"Evaluation methodology","body":"Independent E. coli promoter datasets linked to the chenli-bioinfo/promoter repository. Precision, recall, F1, AUROC and average-precision-based AUPRC are described in the evaluation methods. iPro70-FMWin and iPromoter-2L2.0 evaluated on the independent E. coli sigma70 benchmark. The checked E. coli comparison identifies an independent dataset but does not specify sequence/homology overlap checks against comparator training data. The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","source_ids":["cyaprombert-2022"],"source_locator":"Datasets; Evaluation criteria; Results: independent datasets from E. coli; cached text lines 12, 28–30, 42; comparative evaluation and ablation passages"}],"facts":[{"label":"Datasets","value":"Independent E. coli promoter datasets linked to the chenli-bioinfo/promoter repository.","status":"source_checked","source_ids":["cyaprombert-2022"],"source_locator":"Datasets; Evaluation criteria; Results: independent datasets from E. coli; cached text lines 12, 28–30, 42; comparative evaluation and ablation passages"},{"label":"Splits","value":"The E. coli comparison uses the independent σ70-promoter dataset from the linked Zhang/chenli-bioinfo study and reports class support in Table 3. The E. coli subsection does not specify the exact training/test file allocation; cyanobacterial cross-species partitions elsewhere are not its split definition.","status":"unreported","source_ids":["cyaprombert-2022"],"source_locator":"Datasets; Results: independent E. coli benchmarking; Table 3"},{"label":"Metrics","value":"Precision, recall, F1, AUROC and average-precision-based AUPRC are described in the evaluation methods.","status":"source_checked","source_ids":["cyaprombert-2022"],"source_locator":"Datasets; Evaluation criteria; Results: independent datasets from E. coli; cached text lines 12, 28–30, 42; comparative evaluation and ablation passages"},{"label":"Baselines","value":"iPro70-FMWin and iPromoter-2L2.0 evaluated on the independent E. coli sigma70 benchmark.","status":"source_checked","source_ids":["cyaprombert-2022"],"source_locator":"Datasets; Evaluation criteria; Results: independent datasets from E. coli; cached text lines 12, 28–30, 42; comparative evaluation and ablation passages"},{"label":"Leakage controls","value":"The checked E. coli comparison identifies an independent dataset but does not specify sequence/homology overlap checks against comparator training data.","status":"unreported","source_ids":["cyaprombert-2022"],"source_locator":"Datasets; Evaluation criteria; Results: independent datasets from E. coli; cached text lines 12, 28–30, 42; comparative evaluation and ablation passages"},{"label":"Uncertainty","value":"The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","status":"unreported","source_ids":["cyaprombert-2022"],"source_locator":"Datasets; Evaluation criteria; Results: independent datasets from E. coli; cached text lines 12, 28–30, 42; comparative evaluation and ablation passages"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["cyaprombert-2022"],"source_locator":"Datasets; Evaluation criteria; Results: independent datasets from E. coli; cached text lines 12, 28–30, 42; comparative evaluation and ablation passages"},{"label":"Organisms","value":"Escherichia coli for this sigma70 task; cyanobacterial tasks elsewhere in the paper are separate.","status":"source_checked","source_ids":["cyaprombert-2022"],"source_locator":"Datasets; Evaluation criteria; Results: independent datasets from E. coli; cached text lines 12, 28–30, 42; comparative evaluation and ablation passages"},{"label":"Assays","value":"Curated sigma70 promoter/non-promoter labels.","status":"source_checked","source_ids":["cyaprombert-2022"],"source_locator":"Datasets; Evaluation criteria; Results: independent datasets from E. coli; cached text lines 12, 28–30, 42; comparative evaluation and ablation passages"},{"label":"Allowed inputs","value":"DNA promoter-window sequences.","status":"source_checked","source_ids":["cyaprombert-2022"],"source_locator":"Datasets; Evaluation criteria; Results: independent datasets from E. coli; cached text lines 12, 28–30, 42; comparative evaluation and ablation passages"},{"label":"Adaptation","value":"BERT-based supervised classification trained on the E. coli benchmark; evaluated against existing promoter web servers.","status":"source_checked","source_ids":["cyaprombert-2022"],"source_locator":"Datasets; Evaluation criteria; Results: independent datasets from E. coli; cached text lines 12, 28–30, 42; comparative evaluation and ablation passages"}],"strengths":[],"limitations":[{"text":"The paper also evaluates cyanobacterial data, but those species-transfer partitions do not specify the E. coli test. Table 3 reports the E. coli comparator setting separately.","source_ids":["cyaprombert-2022"],"source_locator":"Datasets; Evaluation criteria; Results: independent datasets from E. coli; cached text lines 12, 28–30, 42; comparative evaluation and ablation passages"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: DNA promoter-window sequences.","Evaluation: BERT-based supervised classification trained on the E. coli benchmark; evaluated against existing promoter web servers.","Readout: Precision, recall, F1, AUROC and average-precision-based AUPRC are described in the evaluation methods."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["cyaprombert-2022"],"source_locator":"Datasets; Evaluation criteria; Results: independent datasets from E. coli; cached text lines 12, 28–30, 42; comparative evaluation and ablation passages"},"coverage":"limited","gaps":["Splits: The E. coli comparison uses the independent σ70-promoter dataset from the linked Zhang/chenli-bioinfo study and reports class support in Table 3. The E. coli subsection does not specify the exact training/test file allocation; cyanobacterial cross-species partitions elsewhere are not its split definition.","Leakage controls: The checked E. coli comparison identifies an independent dataset but does not specify sequence/homology overlap checks against comparator training data.","Uncertainty: The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Relevant full-paper computational evaluation sections, tables/captions and cited supplementary task passages were reviewed. Reporting omissions are scoped to the inspected sources. Original numerical results are unchanged."}}}}
{"id":"reported-task-ed3dd3b83c4505","kind":"benchmark","name":"ClinVar 3-prime UTR variant classification","description":"","status":"needs_review","facets":{"areas":["dna-genomes"],"tasks":["ClinVar 3-prime UTR variant classification"]},"source_ids":["phylogpn-2025"],"links":[{"relation":"dataset","target_id":"reported-dataset-a28180d33f7a23"}],"attributes":{"entity_level":"task","version":null,"task":"ClinVar 3-prime UTR variant classification","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"ClinVar variant classification assesses whether zero-shot allele scores distinguish pathogenic from benign annotations.","summary_source_ids":["phylogpn-2025"],"summary_source_locator":"Results: Log Likelihood Ratio Evaluation; ClinVar and OMIM tests; cached text lines 46–48, 52, 56","sections":[{"title":"Evaluation methodology","body":"ClinVar variants meeting a review-status filter, with pathogenic/likely-pathogenic positives and benign/likely-benign negatives; the catalogue retains the 3-prime UTR subset. Zero-shot evaluation, distinct from supervised embedding tasks elsewhere in the paper. ROC-based assessment for ClinVar; AUPRC described for the separate imbalanced OMIM comparison must not be imported automatically. Nucleotide Transformer, HyenaDNA, Caduceus and GPN-MSA comparisons. The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","source_ids":["phylogpn-2025"],"source_locator":"Results: Log Likelihood Ratio Evaluation; ClinVar and OMIM tests; cached text lines 46–48, 52, 56"}],"facts":[{"label":"Datasets","value":"ClinVar variants meeting a review-status filter, with pathogenic/likely-pathogenic positives and benign/likely-benign negatives; the catalogue retains the 3-prime UTR subset.","status":"source_checked","source_ids":["phylogpn-2025"],"source_locator":"Results: Log Likelihood Ratio Evaluation; ClinVar and OMIM tests; cached text lines 46–48, 52, 56"},{"label":"Splits","value":"Zero-shot evaluation, distinct from supervised embedding tasks elsewhere in the paper.","status":"source_checked","source_ids":["phylogpn-2025"],"source_locator":"Results: Log Likelihood Ratio Evaluation; ClinVar and OMIM tests; cached text lines 46–48, 52, 56"},{"label":"Metrics","value":"ROC-based assessment for ClinVar; AUPRC described for the separate imbalanced OMIM comparison must not be imported automatically.","status":"source_checked","source_ids":["phylogpn-2025"],"source_locator":"Results: Log Likelihood Ratio Evaluation; ClinVar and OMIM tests; cached text lines 46–48, 52, 56"},{"label":"Baselines","value":"Nucleotide Transformer, HyenaDNA, Caduceus and GPN-MSA comparisons.","status":"source_checked","source_ids":["phylogpn-2025"],"source_locator":"Results: Log Likelihood Ratio Evaluation; ClinVar and OMIM tests; cached text lines 46–48, 52, 56"},{"label":"Leakage controls","value":"ClinVar variants are selected by review status and pathogenic/benign labels for zero-shot scoring. The main ClinVar experiment does not report exclusion of those loci from genome pretraining. The odd/even chromosome held-out experiment is a separate ablation, not the stated split of the main ClinVar result.","status":"unreported","source_ids":["phylogpn-2025"],"source_locator":"Classifying ClinVar variants and Table 1; held-out-human-sequence ablation; cached paragraphs 36,51–53,72"},{"label":"Uncertainty","value":"The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","status":"unreported","source_ids":["phylogpn-2025"],"source_locator":"Results: Log Likelihood Ratio Evaluation; ClinVar and OMIM tests; cached text lines 46–48, 52, 56"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["phylogpn-2025"],"source_locator":"Results: Log Likelihood Ratio Evaluation; ClinVar and OMIM tests; cached text lines 46–48, 52, 56"},{"label":"Organisms","value":"Human ClinVar variants.","status":"source_checked","source_ids":["phylogpn-2025"],"source_locator":"Results: Log Likelihood Ratio Evaluation; ClinVar and OMIM tests; cached text lines 46–48, 52, 56"},{"label":"Assays","value":"Clinical pathogenic/benign assertions meeting a review-status filter.","status":"source_checked","source_ids":["phylogpn-2025"],"source_locator":"Results: Log Likelihood Ratio Evaluation; ClinVar and OMIM tests; cached text lines 46–48, 52, 56"},{"label":"Allowed inputs","value":"Reference/alternate genomic sequences for the 3-prime UTR subset.","status":"source_checked","source_ids":["phylogpn-2025"],"source_locator":"Results: Log Likelihood Ratio Evaluation; ClinVar and OMIM tests; cached text lines 46–48, 52, 56"},{"label":"Adaptation","value":"Zero-shot variant scoring; supervised embedding experiments elsewhere are not this protocol.","status":"source_checked","source_ids":["phylogpn-2025"],"source_locator":"Results: Log Likelihood Ratio Evaluation; ClinVar and OMIM tests; cached text lines 46–48, 52, 56"}],"strengths":[{"text":"The clinical evaluation uses an explicit review-status filter and separates clinical labels from zero-shot scoring.","source_ids":["phylogpn-2025"],"source_locator":"Results: Log Likelihood Ratio Evaluation; ClinVar and OMIM tests; cached text lines 46–48, 52, 56"}],"limitations":[{"text":"The main ClinVar result uses review-status-filtered variants; it is distinct from the held-out-chromosome ablation. Genome pretraining and downstream label use are separate sources of possible overlap.","source_ids":["phylogpn-2025"],"source_locator":"Results: Log Likelihood Ratio Evaluation; ClinVar and OMIM tests; cached text lines 46–48, 52, 56"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Reference/alternate genomic sequences for the 3-prime UTR subset.","Evaluation: Zero-shot variant scoring; supervised embedding experiments elsewhere are not this protocol.","Readout: ROC-based assessment for ClinVar; AUPRC described for the separate imbalanced OMIM comparison must not be imported automatically."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["phylogpn-2025"],"source_locator":"Results: Log Likelihood Ratio Evaluation; ClinVar and OMIM tests; cached text lines 46–48, 52, 56"},"coverage":"limited","gaps":["Leakage controls: ClinVar variants are selected by review status and pathogenic/benign labels for zero-shot scoring. The main ClinVar experiment does not report exclusion of those loci from genome pretraining. The odd/even chromosome held-out experiment is a separate ablation, not the stated split of the main ClinVar result.","Uncertainty: The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Relevant full-paper computational evaluation sections, tables/captions and cited supplementary task passages were reviewed. Reporting omissions are scoped to the inspected sources. Original numerical results are unchanged."}}}}
{"id":"reported-task-ee34721cf55590","kind":"benchmark","name":"gene fusion breakpoint classification","description":"","status":"needs_review","facets":{"areas":["dna-genomes"],"tasks":["gene fusion breakpoint classification"]},"source_ids":["fusion-breakpoint-foundation-models-2026"],"links":[{"relation":"dataset","target_id":"reported-dataset-f6922a9744ba27"}],"attributes":{"entity_level":"task","version":null,"task":"gene fusion breakpoint classification","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Fusion-breakpoint classification evaluates curated sequence labels rather than locating breakpoints in raw sequencing data.","summary_source_ids":["fusion-breakpoint-foundation-models-2026"],"summary_source_locator":"Methods: Evaluation metrics; implementation; Discussion; cached text lines 34–35, 45, 88–90","sections":[{"title":"Evaluation methodology","body":"The curated FusionAI benchmark. Accuracy, class-weighted precision/recall/F1 and ROC-AUC. Foundation-model embeddings with supervised heads are compared with FusionAI.","source_ids":["fusion-breakpoint-foundation-models-2026"],"source_locator":"Methods: Evaluation metrics; implementation; Discussion; cached text lines 34–35, 45, 88–90; Methods: Evaluation metrics; implementation; Discussion; cached text lines 34–35, 45, 88–90; Methods: Evaluation metrics; implementation; Discussion; cached text lines 34–35, 45, 88–90"}],"facts":[{"label":"Datasets","value":"The curated FusionAI benchmark.","status":"source_checked","source_ids":["fusion-breakpoint-foundation-models-2026"],"source_locator":"Methods: Evaluation metrics; implementation; Discussion; cached text lines 34–35, 45, 88–90"},{"label":"Splits","value":"Sampling and partitioning use a fixed random seed so models share the same examples. The reported split is not described as holding out entire genes or breakpoint families.","status":"source_checked","source_ids":["fusion-breakpoint-foundation-models-2026"],"source_locator":"Methods: implementation/reproducibility, random seed 42; Dataset and Classification"},{"label":"Metrics","value":"Accuracy, class-weighted precision/recall/F1 and ROC-AUC.","status":"source_checked","source_ids":["fusion-breakpoint-foundation-models-2026"],"source_locator":"Methods: Evaluation metrics; implementation; Discussion; cached text lines 34–35, 45, 88–90"},{"label":"Baselines","value":"Foundation-model embeddings with supervised heads are compared with FusionAI.","status":"source_checked","source_ids":["fusion-breakpoint-foundation-models-2026"],"source_locator":"Methods: Evaluation metrics; implementation; Discussion; cached text lines 34–35, 45, 88–90"},{"label":"Leakage controls","value":"All model comparisons use the same fixed-seed partitions. This supports matched comparisons but does not by itself prevent related genomic loci crossing the split.","status":"source_checked","source_ids":["fusion-breakpoint-foundation-models-2026"],"source_locator":"Methods: Evaluation metrics; implementation; Discussion; cached text lines 34–35, 45, 88–90"},{"label":"Uncertainty","value":"The paper reports final-epoch neural results and single-run SVM results; repeated-run uncertainty is not established in the reviewed passage.","status":"source_checked","source_ids":["fusion-breakpoint-foundation-models-2026"],"source_locator":"Methods: Evaluation metrics; implementation; Discussion; cached text lines 34–35, 45, 88–90"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["fusion-breakpoint-foundation-models-2026"],"source_locator":"Methods: Evaluation metrics; implementation; Discussion; cached text lines 34–35, 45, 88–90"},{"label":"Organisms","value":"Human gene fusions from the reused FusionAI classification dataset; the foundation models’ multispecies pretraining coverage is a different property.","status":"source_checked","source_ids":["fusion-breakpoint-foundation-models-2026"],"source_locator":"Dataset/methods description; Discussion of human gene-fusion classification; cached paragraphs 77,84"},{"label":"Assays","value":"Fusion breakpoint labels.","status":"source_checked","source_ids":["fusion-breakpoint-foundation-models-2026"],"source_locator":"Methods: Evaluation metrics; implementation; Discussion; cached text lines 34–35, 45, 88–90"},{"label":"Allowed inputs","value":"Genomic breakpoint sequence represented by foundation-model embeddings.","status":"source_checked","source_ids":["fusion-breakpoint-foundation-models-2026"],"source_locator":"Methods: Evaluation metrics; implementation; Discussion; cached text lines 34–35, 45, 88–90"},{"label":"Adaptation","value":"Supervised prediction heads are compared with FusionAI.","status":"source_checked","source_ids":["fusion-breakpoint-foundation-models-2026"],"source_locator":"Methods: Evaluation metrics; implementation; Discussion; cached text lines 34–35, 45, 88–90"}],"strengths":[],"limitations":[{"text":"Breakpoint localization is described as future work. This classification task should not be represented as validated coordinate prediction.","source_ids":["fusion-breakpoint-foundation-models-2026"],"source_locator":"Methods: Evaluation metrics; implementation; Discussion; cached text lines 34–35, 45, 88–90"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Genomic breakpoint sequence represented by foundation-model embeddings.","Evaluation: Supervised prediction heads are compared with FusionAI.","Readout: Accuracy, class-weighted precision/recall/F1 and ROC-AUC."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["fusion-breakpoint-foundation-models-2026"],"source_locator":"Methods: Evaluation metrics; implementation; Discussion; cached text lines 34–35, 45, 88–90"},"coverage":"limited","gaps":["Breakpoint localization is described as future work. This classification task should not be represented as validated coordinate prediction."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Relevant full-paper computational evaluation sections, tables/captions and cited supplementary task passages were reviewed. Reporting omissions are scoped to the inspected sources. Original numerical results are unchanged."}}}}
{"id":"reported-task-f0ed5188dbb6d4","kind":"benchmark","name":"protein-protein binding-site prediction","description":"","status":"needs_review","facets":{"areas":["proteins-complexes"],"tasks":["protein-protein binding-site prediction"]},"source_ids":["protein-binding-sites-2023"],"links":[{"relation":"dataset","target_id":"reported-dataset-f08b1a60aebeeb"}],"attributes":{"entity_level":"task","version":null,"task":"protein-protein binding-site prediction","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Protein-interaction binding-site prediction evaluates residue labels on nonredundant protein collections.","summary_source_ids":["protein-binding-sites-2023"],"summary_source_locator":"Methods: Datasets; Evaluation performance; cached text lines 57–59, 110–112; matching task comparison table/ablation captions","sections":[{"title":"Evaluation methodology","body":"PDB-derived Dset collections and a BioLip-derived Dset_1291 collection. Dset_843 supplies training sequences and Dset_448 an independent test subset for the BioLip setting. Sensitivity, specificity, precision, accuracy, F1, MCC, AUROC and average precision. Feature-descriptor ablations, ESM-1b/ProGen2/ProtT5 embeddings and task-specific SCRIBER/DELPHI comparisons. The dataset construction describes sequence-similarity reduction before the train/test subdivision. The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","source_ids":["protein-binding-sites-2023"],"source_locator":"Methods: Datasets; Evaluation performance; cached text lines 57–59, 110–112; matching task comparison table/ablation captions"}],"facts":[{"label":"Datasets","value":"PDB-derived Dset collections and a BioLip-derived Dset_1291 collection.","status":"source_checked","source_ids":["protein-binding-sites-2023"],"source_locator":"Methods: Datasets; Evaluation performance; cached text lines 57–59, 110–112; matching task comparison table/ablation captions"},{"label":"Splits","value":"Dset_843 supplies training sequences and Dset_448 an independent test subset for the BioLip setting.","status":"source_checked","source_ids":["protein-binding-sites-2023"],"source_locator":"Methods: Datasets; Evaluation performance; cached text lines 57–59, 110–112; matching task comparison table/ablation captions"},{"label":"Metrics","value":"Sensitivity, specificity, precision, accuracy, F1, MCC, AUROC and average precision.","status":"source_checked","source_ids":["protein-binding-sites-2023"],"source_locator":"Methods: Datasets; Evaluation performance; cached text lines 57–59, 110–112; matching task comparison table/ablation captions"},{"label":"Baselines","value":"Feature-descriptor ablations, ESM-1b/ProGen2/ProtT5 embeddings and task-specific SCRIBER/DELPHI comparisons.","status":"source_checked","source_ids":["protein-binding-sites-2023"],"source_locator":"Methods: Datasets; Evaluation performance; cached text lines 57–59, 110–112; matching task comparison table/ablation captions"},{"label":"Leakage controls","value":"The dataset construction describes sequence-similarity reduction before the train/test subdivision.","status":"source_checked","source_ids":["protein-binding-sites-2023"],"source_locator":"Methods: Datasets; Evaluation performance; cached text lines 57–59, 110–112; matching task comparison table/ablation captions"},{"label":"Uncertainty","value":"The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","status":"unreported","source_ids":["protein-binding-sites-2023"],"source_locator":"Methods: Datasets; Evaluation performance; cached text lines 57–59, 110–112; matching task comparison table/ablation captions"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["protein-binding-sites-2023"],"source_locator":"Methods: Datasets; Evaluation performance; cached text lines 57–59, 110–112; matching task comparison table/ablation captions"},{"label":"Organisms","value":"The benchmark uses PDB- and BioLip-derived protein collections selected for structure quality, sequence redundancy and interaction annotations. The Datasets section does not report their species distribution or a species-specific sampling rule.","status":"unreported","source_ids":["protein-binding-sites-2023"],"source_locator":"Methods: Datasets; Supplementary Table 1 description"},{"label":"Assays","value":"Protein–protein binding-residue annotations.","status":"source_checked","source_ids":["protein-binding-sites-2023"],"source_locator":"Methods: Datasets; Evaluation performance; cached text lines 57–59, 110–112; matching task comparison table/ablation captions"},{"label":"Allowed inputs","value":"Protein sequence/representation for binding-site prediction.","status":"source_checked","source_ids":["protein-binding-sites-2023"],"source_locator":"Methods: Datasets; Evaluation performance; cached text lines 57–59, 110–112; matching task comparison table/ablation captions"},{"label":"Adaptation","value":"Supervised residue classification using the defined Dset training and independent test sets.","status":"source_checked","source_ids":["protein-binding-sites-2023"],"source_locator":"Methods: Datasets; Evaluation performance; cached text lines 57–59, 110–112; matching task comparison table/ablation captions"}],"strengths":[],"limitations":[{"text":"Multiple Dset collections differ in sequence coverage and training provenance. Results must retain the named training/test combination.","source_ids":["protein-binding-sites-2023"],"source_locator":"Methods: Datasets; Evaluation performance; cached text lines 57–59, 110–112; matching task comparison table/ablation captions"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Protein sequence/representation for binding-site prediction.","Evaluation: Supervised residue classification using the defined Dset training and independent test sets.","Readout: Sensitivity, specificity, precision, accuracy, F1, MCC, AUROC and average precision."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["protein-binding-sites-2023"],"source_locator":"Methods: Datasets; Evaluation performance; cached text lines 57–59, 110–112; matching task comparison table/ablation captions"},"coverage":"limited","gaps":["Uncertainty: The cited text-accessible evaluation sections give no confidence-interval, resampling or repeat-run error-bar specification. Image-only tables and uninspected supplements are outside this absence claim.","Organisms: The benchmark uses PDB- and BioLip-derived protein collections selected for structure quality, sequence redundancy and interaction annotations. The Datasets section does not report their species distribution or a species-specific sampling rule."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Relevant full-paper computational evaluation sections, tables/captions and cited supplementary task passages were reviewed. Reporting omissions are scoped to the inspected sources. Original numerical results are unchanged."}}}}
{"id":"reported-task-f3a12dbc0e0439","kind":"benchmark","name":"Antibody loop structure prediction","description":"","status":"needs_review","facets":{"areas":["molecular-interactions"],"tasks":["Antibody loop structure prediction"]},"source_ids":["ibex-2025"],"links":[{"relation":"dataset","target_id":"reported-dataset-b7204b005bd476"}],"attributes":{"entity_level":"task","version":null,"task":"Antibody loop structure prediction","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Antibody loop structure assessment uses public ImmuneBuilder tests and a separate private structural collection.","summary_source_ids":["ibex-2025"],"summary_source_locator":"Results: Benchmarking on the ImmuneBuilder test set; Benchmarking on a private dataset; cached text lines 22–28","sections":[{"title":"Evaluation methodology","body":"ImmuneBuilder antibody, nanobody and TCR tests; an additional private antibody-structure dataset. Training excludes clusters containing public test structures; the private set is characterized by loop distance to public structures. Mean region-specific RMSD; private-set loop RMSD is also stratified by sequence edit distance. ESMFold, Chai-1, Boltz-1 and specialized immune-protein structure models. Boltz-2 is excluded from the public test comparison because its later training cutoff includes much of that test set. Chai-1 and Boltz-1 use one seed/trajectory in the public comparison; this is not a repeated-sampling uncertainty analysis.","source_ids":["ibex-2025"],"source_locator":"Results: Benchmarking on the ImmuneBuilder test set; Benchmarking on a private dataset; cached text lines 22–28"}],"facts":[{"label":"Datasets","value":"ImmuneBuilder antibody, nanobody and TCR tests; an additional private antibody-structure dataset.","status":"source_checked","source_ids":["ibex-2025"],"source_locator":"Results: Benchmarking on the ImmuneBuilder test set; Benchmarking on a private dataset; cached text lines 22–28"},{"label":"Splits","value":"Training excludes clusters containing public test structures; the private set is characterized by loop distance to public structures.","status":"source_checked","source_ids":["ibex-2025"],"source_locator":"Results: Benchmarking on the ImmuneBuilder test set; Benchmarking on a private dataset; cached text lines 22–28"},{"label":"Metrics","value":"Mean region-specific RMSD; private-set loop RMSD is also stratified by sequence edit distance.","status":"source_checked","source_ids":["ibex-2025"],"source_locator":"Results: Benchmarking on the ImmuneBuilder test set; Benchmarking on a private dataset; cached text lines 22–28"},{"label":"Baselines","value":"ESMFold, Chai-1, Boltz-1 and specialized immune-protein structure models.","status":"source_checked","source_ids":["ibex-2025"],"source_locator":"Results: Benchmarking on the ImmuneBuilder test set; Benchmarking on a private dataset; cached text lines 22–28"},{"label":"Leakage controls","value":"Boltz-2 is excluded from the public test comparison because its later training cutoff includes much of that test set.","status":"source_checked","source_ids":["ibex-2025"],"source_locator":"Results: Benchmarking on the ImmuneBuilder test set; Benchmarking on a private dataset; cached text lines 22–28"},{"label":"Uncertainty","value":"Chai-1 and Boltz-1 use one seed/trajectory in the public comparison; this is not a repeated-sampling uncertainty analysis.","status":"source_checked","source_ids":["ibex-2025"],"source_locator":"Results: Benchmarking on the ImmuneBuilder test set; Benchmarking on a private dataset; cached text lines 22–28"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["ibex-2025"],"source_locator":"Results: Benchmarking on the ImmuneBuilder test set; Benchmarking on a private dataset; cached text lines 22–28"},{"label":"Organisms","value":"Antibody, nanobody and T-cell receptor collections.","status":"source_checked","source_ids":["ibex-2025"],"source_locator":"Results: Benchmarking on the ImmuneBuilder test set; Benchmarking on a private dataset; cached text lines 22–28"},{"label":"Assays","value":"Experimentally determined immune-protein structures.","status":"source_checked","source_ids":["ibex-2025"],"source_locator":"Results: Benchmarking on the ImmuneBuilder test set; Benchmarking on a private dataset; cached text lines 22–28"},{"label":"Allowed inputs","value":"Immune-protein sequence for loop/structure prediction.","status":"source_checked","source_ids":["ibex-2025"],"source_locator":"Results: Benchmarking on the ImmuneBuilder test set; Benchmarking on a private dataset; cached text lines 22–28"},{"label":"Adaptation","value":"Structure prediction trained with clusters containing public test examples excluded.","status":"source_checked","source_ids":["ibex-2025"],"source_locator":"Results: Benchmarking on the ImmuneBuilder test set; Benchmarking on a private dataset; cached text lines 22–28"}],"strengths":[{"text":"Training excludes clusters containing public test structures rather than only deleting exact test entries.","source_ids":["ibex-2025"],"source_locator":"Results: Benchmarking on the ImmuneBuilder test set; Benchmarking on a private dataset; cached text lines 22–28"}],"limitations":[{"text":"Private evaluation structures limit independent reproducibility. Antigen context is absent in the described private comparison.","source_ids":["ibex-2025"],"source_locator":"Results: Benchmarking on the ImmuneBuilder test set; Benchmarking on a private dataset; cached text lines 22–28"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Immune-protein sequence for loop/structure prediction.","Evaluation: Structure prediction trained with clusters containing public test examples excluded.","Readout: Mean region-specific RMSD; private-set loop RMSD is also stratified by sequence edit distance."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["ibex-2025"],"source_locator":"Results: Benchmarking on the ImmuneBuilder test set; Benchmarking on a private dataset; cached text lines 22–28"},"coverage":"limited","gaps":["Private evaluation structures limit independent reproducibility. Antigen context is absent in the described private comparison."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Task-specific computational methodology and field context checked in the cited primary-source artifact. Source-backed fields, inapplicable evaluator dimensions and unresolved details are distinguished. Numerical results were not reproduced."}}}}
{"id":"reported-task-f4b1c9373f0929","kind":"benchmark","name":"Hierarchical metagenomic taxonomy classification","description":"","status":"needs_review","facets":{"areas":["microbes-communities"],"tasks":["Hierarchical metagenomic taxonomy classification"]},"source_ids":["icctax-2025"],"links":[{"relation":"dataset","target_id":"reported-dataset-561834dfa1682c"}],"attributes":{"entity_level":"task","version":null,"task":"Hierarchical metagenomic taxonomy classification","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Hierarchical taxonomic classification evaluates sequence labels under separate in-distribution and out-of-distribution settings.","summary_source_ids":["icctax-2025"],"summary_source_locator":"Results §§3.1–3.2.1; cached text lines 47–48, 51, 53–54; uncertainty/repeat-run/statistical-comparison passages","sections":[{"title":"Evaluation methodology","body":"BERTax-derived ID, OOD and Complete datasets spanning major taxonomic groups. Accuracy and macro average precision at superkingdom and phylum levels; micro/macro AUC appears in supplements. MMseqs2, Minimap2, Kraken2, sourmash, MetaPhlAn4, CAT, DeepMicrobes and BERTax.","source_ids":["icctax-2025"],"source_locator":"Results §§3.1–3.2.1; cached text lines 47–48, 51, 53–54; uncertainty/repeat-run/statistical-comparison passages; Results §§3.1–3.2.1; cached text lines 47–48, 51, 53–54; uncertainty/repeat-run/statistical-comparison passages; Results §§3.1–3.2.1; cached text lines 47–48, 51, 53–54; uncertainty/repeat-run/statistical-comparison passages"}],"facts":[{"label":"Datasets","value":"BERTax-derived ID, OOD and Complete datasets spanning major taxonomic groups.","status":"source_checked","source_ids":["icctax-2025"],"source_locator":"Results §§3.1–3.2.1; cached text lines 47–48, 51, 53–54; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Splits","value":"ID reserves 2,000 sequences per phylum for testing and uses the rest for training. OOD assigns every sequence in a genus to a single side of the train/test partition. The Complete-dataset classifier is a separate setting.","status":"source_checked","source_ids":["icctax-2025"],"source_locator":"§3.1 Datasets; Fig.2 caption (a–c)"},{"label":"Metrics","value":"Accuracy and macro average precision at superkingdom and phylum levels; micro/macro AUC appears in supplements.","status":"source_checked","source_ids":["icctax-2025"],"source_locator":"Results §§3.1–3.2.1; cached text lines 47–48, 51, 53–54; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Baselines","value":"MMseqs2, Minimap2, Kraken2, sourmash, MetaPhlAn4, CAT, DeepMicrobes and BERTax.","status":"source_checked","source_ids":["icctax-2025"],"source_locator":"Results §§3.1–3.2.1; cached text lines 47–48, 51, 53–54; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Leakage controls","value":"Figure 2 defines ID testing by sampling sequences within phyla, while OOD places all sequences from a genus wholly in training or wholly in testing. The same genome collection is partitioned differently, so ID scores do not establish performance on unseen genera.","status":"source_checked","source_ids":["icctax-2025"],"source_locator":"§3.1 Datasets; full-XML Fig.2 caption (a–c)"},{"label":"Uncertainty","value":"The paper gives bootstrap confidence intervals for accuracy and average precision on both ID and OOD datasets.","status":"source_checked","source_ids":["icctax-2025"],"source_locator":"Results §§3.1–3.2.1; cached text lines 47–48, 51, 53–54; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["icctax-2025"],"source_locator":"Results §§3.1–3.2.1; cached text lines 47–48, 51, 53–54; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Organisms","value":"Major taxonomic groups in BERTax-derived ID/OOD datasets.","status":"source_checked","source_ids":["icctax-2025"],"source_locator":"Results §§3.1–3.2.1; cached text lines 47–48, 51, 53–54; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Assays","value":"Reference sequence/taxonomy annotations.","status":"source_checked","source_ids":["icctax-2025"],"source_locator":"Results §§3.1–3.2.1; cached text lines 47–48, 51, 53–54; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Allowed inputs","value":"DNA sequence fragments.","status":"source_checked","source_ids":["icctax-2025"],"source_locator":"Results §§3.1–3.2.1; cached text lines 47–48, 51, 53–54; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Adaptation","value":"Supervised hierarchical classification tested in in-distribution and out-of-distribution regimes.","status":"source_checked","source_ids":["icctax-2025"],"source_locator":"Results §§3.1–3.2.1; cached text lines 47–48, 51, 53–54; uncertainty/repeat-run/statistical-comparison passages"}],"strengths":[],"limitations":[{"text":"ID and OOD scores use different taxonomic partition strategies. Reference-database coverage also varies between external comparison tools.","source_ids":["icctax-2025"],"source_locator":"Results §§3.1–3.2.1; cached text lines 47–48, 51, 53–54; uncertainty/repeat-run/statistical-comparison passages"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: DNA sequence fragments.","Evaluation: Supervised hierarchical classification tested in in-distribution and out-of-distribution regimes.","Readout: Accuracy and macro average precision at superkingdom and phylum levels; micro/macro AUC appears in supplements."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["icctax-2025"],"source_locator":"Results §§3.1–3.2.1; cached text lines 47–48, 51, 53–54; uncertainty/repeat-run/statistical-comparison passages"},"coverage":"limited","gaps":["ID and OOD scores use different taxonomic partition strategies. Reference-database coverage also varies between external comparison tools."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Relevant full-paper computational evaluation sections, tables/captions and cited supplementary task passages were reviewed. Reporting omissions are scoped to the inspected sources. Original numerical results are unchanged."}}}}
{"id":"reported-task-f7142c3b3e0f3c","kind":"benchmark","name":"translation-efficiency prediction","description":"","status":"needs_review","facets":{"areas":["rna-transcriptomes"],"tasks":["translation-efficiency prediction"]},"source_ids":["mrnabert-2025"],"links":[{"relation":"dataset","target_id":"reported-dataset-1744719eef145b"}],"attributes":{"entity_level":"task","version":null,"task":"translation-efficiency prediction","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Translation-efficiency assessment uses published full-transcript measurements and is distinct from the paper’s CDS-only tasks.","summary_source_ids":["mrnabert-2025"],"summary_source_locator":"Methods: full-length mRNA datasets and translation efficiency; cached text lines 119–120; task metric definitions and corresponding results table","sections":[{"title":"Evaluation methodology","body":"PERSIST-seq full-length mRNA measurements and a ribosome-profiling atlas of human/mouse translation efficiency. R-squared and Spearman correlation for the human/mouse translation-efficiency atlas; separate mRNA tasks use their own metrics. Comparison models are fine-tuned using a common model library for the full-length mRNA task.","source_ids":["mrnabert-2025"],"source_locator":"Methods: full-length mRNA datasets and translation efficiency; cached text lines 119–120; task metric definitions and corresponding results table"}],"facts":[{"label":"Datasets","value":"PERSIST-seq full-length mRNA measurements and a ribosome-profiling atlas of human/mouse translation efficiency.","status":"source_checked","source_ids":["mrnabert-2025"],"source_locator":"Methods: full-length mRNA datasets and translation efficiency; cached text lines 119–120; task metric definitions and corresponding results table"},{"label":"Splits","value":"The supplementary full-length section distinguishes PERSIST-seq traits from the human/mouse TE atlas, but gives no TE-atlas train/validation/test membership or partition rule. Five-fold reporting for PERSIST-seq or other downstream tasks cannot be assigned to the atlas result.","status":"unreported","source_ids":["mrnabert-2025","evidence-task-final-b-mrnabert-supplement"],"source_locator":"Main full-length mRNA results/Methods; Supplementary Information Full-Length mRNA, pp19–21, Tables 12–13"},{"label":"Metrics","value":"R-squared and Spearman correlation for the human/mouse translation-efficiency atlas; separate mRNA tasks use their own metrics.","status":"source_checked","source_ids":["mrnabert-2025"],"source_locator":"Methods: full-length mRNA datasets and translation efficiency; cached text lines 119–120; task metric definitions and corresponding results table"},{"label":"Baselines","value":"Comparison models are fine-tuned using a common model library for the full-length mRNA task.","status":"source_checked","source_ids":["mrnabert-2025"],"source_locator":"Methods: full-length mRNA datasets and translation efficiency; cached text lines 119–120; task metric definitions and corresponding results table"},{"label":"Leakage controls","value":"The main and supplementary TE-atlas descriptions provide sequence-length and dataset summaries but do not describe exclusion of homologous transcripts, shared genes or pretraining overlap for that experiment. Controls described for other downstream datasets are not atlas controls.","status":"unreported","source_ids":["mrnabert-2025","evidence-task-final-b-mrnabert-supplement"],"source_locator":"Methods: translation efficiency; Supplementary Information Full-Length mRNA, pp19–21, Tables 12–13"},{"label":"Uncertainty","value":"The TE-atlas comparison has no task-specific repeat count or confidence-interval method in the inspected main and supplementary sections. The five-fold means/standard deviations for the separate PERSIST-seq evaluation do not define atlas uncertainty.","status":"unreported","source_ids":["mrnabert-2025","evidence-task-final-b-mrnabert-supplement"],"source_locator":"Main TE-atlas results; Supplementary Information Full-Length mRNA, pp19–21"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["mrnabert-2025"],"source_locator":"Methods: full-length mRNA datasets and translation efficiency; cached text lines 119–120; task metric definitions and corresponding results table"},{"label":"Organisms","value":"Human and mouse.","status":"source_checked","source_ids":["mrnabert-2025"],"source_locator":"Methods: full-length mRNA datasets and translation efficiency; cached text lines 119–120; task metric definitions and corresponding results table"},{"label":"Assays","value":"PERSIST-seq and ribosome-profiling translation-efficiency measurements.","status":"source_checked","source_ids":["mrnabert-2025"],"source_locator":"Methods: full-length mRNA datasets and translation efficiency; cached text lines 119–120; task metric definitions and corresponding results table"},{"label":"Allowed inputs","value":"Full-length mRNA sequence.","status":"source_checked","source_ids":["mrnabert-2025"],"source_locator":"Methods: full-length mRNA datasets and translation efficiency; cached text lines 119–120; task metric definitions and corresponding results table"},{"label":"Adaptation","value":"Task fine-tuning through a common comparison-model library.","status":"source_checked","source_ids":["mrnabert-2025"],"source_locator":"Methods: full-length mRNA datasets and translation efficiency; cached text lines 119–120; task metric definitions and corresponding results table"}],"strengths":[],"limitations":[{"text":"PERSIST-seq full-transcript traits and the human/mouse TE atlas are different evaluations. The supplementary section identifies them separately but does not supply an atlas partition or uncertainty protocol.","source_ids":["mrnabert-2025","evidence-task-final-b-mrnabert-supplement"],"source_locator":"Methods: full-length mRNA datasets and translation efficiency; cached text lines 119–120; task metric definitions and corresponding results table ; Supplementary Information task sections"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Full-length mRNA sequence.","Evaluation: Task fine-tuning through a common comparison-model library.","Readout: R-squared and Spearman correlation for the human/mouse translation-efficiency atlas; separate mRNA tasks use their own metrics."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["mrnabert-2025"],"source_locator":"Methods: full-length mRNA datasets and translation efficiency; cached text lines 119–120; task metric definitions and corresponding results table"},"coverage":"limited","gaps":["Splits: The supplementary full-length section distinguishes PERSIST-seq traits from the human/mouse TE atlas, but gives no TE-atlas train/validation/test membership or partition rule. Five-fold reporting for PERSIST-seq or other downstream tasks cannot be assigned to the atlas result.","Leakage controls: The main and supplementary TE-atlas descriptions provide sequence-length and dataset summaries but do not describe exclusion of homologous transcripts, shared genes or pretraining overlap for that experiment. Controls described for other downstream datasets are not atlas controls.","Uncertainty: The TE-atlas comparison has no task-specific repeat count or confidence-interval method in the inspected main and supplementary sections. The five-fold means/standard deviations for the separate PERSIST-seq evaluation do not define atlas uncertainty."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Relevant full-paper computational evaluation sections, tables/captions and cited supplementary task passages were reviewed. Reporting omissions are scoped to the inspected sources. Original numerical results are unchanged."}}}}
{"id":"reported-task-ff2dec63c5a3dd","kind":"benchmark","name":"Lipid–protein binding pose","description":"","status":"needs_review","facets":{"areas":["molecular-interactions"],"tasks":["Lipid–protein binding pose"]},"source_ids":["lipp-2026"],"links":[{"relation":"dataset","target_id":"reported-dataset-6a44f5946cd7ab"}],"attributes":{"entity_level":"task","version":null,"task":"Lipid–protein binding pose","scope_note":"Paper-specific evaluation task; protocol completeness requires further extraction.","historical_missing_metadata":{"protocol_version":"not_reported_in_legacy_extract","split":"not_reported_in_legacy_extract"},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"Lipid–protein pose prediction evaluates curated experimental structures while distinguishing pre- and post-training-cutoff examples.","summary_source_ids":["lipp-2026"],"summary_source_locator":"Methods: Curation of LiPP; test/precutoff separation; Evaluation of Scoring Power; cached text lines 12–16, 50–51; uncertainty/repeat-run/statistical-comparison passages","sections":[{"title":"Evaluation methodology","body":"BioDolphin 1.1-derived lipid–protein complexes curated against PDB structural information. Post-cutoff structures form a test subset; older structures remain in a separately labelled precutoff subset. Top-ranked lipid RMSD is compared with each method’s confidence or affinity score. AlphaFold3, Chai-1, RoseTTAFold All-Atom, DiffDock-L and AutoDock Vina. The temporal cutoff is selected after the most recent stated training/calibration cutoff among compared tools. Success-rate intervals are computed with the exact binomial distribution.","source_ids":["lipp-2026"],"source_locator":"Methods: Curation of LiPP; test/precutoff separation; Evaluation of Scoring Power; cached text lines 12–16, 50–51; uncertainty/repeat-run/statistical-comparison passages"}],"facts":[{"label":"Datasets","value":"BioDolphin 1.1-derived lipid–protein complexes curated against PDB structural information.","status":"source_checked","source_ids":["lipp-2026"],"source_locator":"Methods: Curation of LiPP; test/precutoff separation; Evaluation of Scoring Power; cached text lines 12–16, 50–51; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Splits","value":"Post-cutoff structures form a test subset; older structures remain in a separately labelled precutoff subset.","status":"source_checked","source_ids":["lipp-2026"],"source_locator":"Methods: Curation of LiPP; test/precutoff separation; Evaluation of Scoring Power; cached text lines 12–16, 50–51; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Metrics","value":"Top-ranked lipid RMSD is compared with each method’s confidence or affinity score.","status":"source_checked","source_ids":["lipp-2026"],"source_locator":"Methods: Curation of LiPP; test/precutoff separation; Evaluation of Scoring Power; cached text lines 12–16, 50–51; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Baselines","value":"AlphaFold3, Chai-1, RoseTTAFold All-Atom, DiffDock-L and AutoDock Vina.","status":"source_checked","source_ids":["lipp-2026"],"source_locator":"Methods: Curation of LiPP; test/precutoff separation; Evaluation of Scoring Power; cached text lines 12–16, 50–51; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Leakage controls","value":"The temporal cutoff is selected after the most recent stated training/calibration cutoff among compared tools.","status":"source_checked","source_ids":["lipp-2026"],"source_locator":"Methods: Curation of LiPP; test/precutoff separation; Evaluation of Scoring Power; cached text lines 12–16, 50–51; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Uncertainty","value":"Success-rate intervals are computed with the exact binomial distribution.","status":"source_checked","source_ids":["lipp-2026"],"source_locator":"Methods: Curation of LiPP; test/precutoff separation; Evaluation of Scoring Power; cached text lines 12–16, 50–51; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["lipp-2026"],"source_locator":"Methods: Curation of LiPP; test/precutoff separation; Evaluation of Scoring Power; cached text lines 12–16, 50–51; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Organisms","value":"LiPP selects BioDolphin/PDB lipid–protein structures by quality, ligand class and redundancy. Dataset curation and coverage analyses describe protein functions/families and lipid classes but do not tabulate organism composition.","status":"unreported","source_ids":["lipp-2026"],"source_locator":"Materials and Methods: LiPP curation; Results: dataset coverage"},{"label":"Assays","value":"BioDolphin/PDB structural references.","status":"source_checked","source_ids":["lipp-2026"],"source_locator":"Methods: Curation of LiPP; test/precutoff separation; Evaluation of Scoring Power; cached text lines 12–16, 50–51; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Allowed inputs","value":"Lipid and protein inputs for pose prediction.","status":"source_checked","source_ids":["lipp-2026"],"source_locator":"Methods: Curation of LiPP; test/precutoff separation; Evaluation of Scoring Power; cached text lines 12–16, 50–51; uncertainty/repeat-run/statistical-comparison passages"},{"label":"Adaptation","value":"Pretrained structure/docking methods are tested on pre-cutoff and post-cutoff structures separately.","status":"source_checked","source_ids":["lipp-2026"],"source_locator":"Methods: Curation of LiPP; test/precutoff separation; Evaluation of Scoring Power; cached text lines 12–16, 50–51; uncertainty/repeat-run/statistical-comparison passages"}],"strengths":[],"limitations":[{"text":"Temporal exclusion does not establish complete sequence/ligand dissimilarity. Scores from different tools have different meanings and are not interchangeable uncertainty estimates.","source_ids":["lipp-2026"],"source_locator":"Methods: Curation of LiPP; test/precutoff separation; Evaluation of Scoring Power; cached text lines 12–16, 50–51; uncertainty/repeat-run/statistical-comparison passages"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Lipid and protein inputs for pose prediction.","Evaluation: Pretrained structure/docking methods are tested on pre-cutoff and post-cutoff structures separately.","Readout: Top-ranked lipid RMSD is compared with each method’s confidence or affinity score."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["lipp-2026"],"source_locator":"Methods: Curation of LiPP; test/precutoff separation; Evaluation of Scoring Power; cached text lines 12–16, 50–51; uncertainty/repeat-run/statistical-comparison passages"},"coverage":"limited","gaps":["Organisms: LiPP selects BioDolphin/PDB lipid–protein structures by quality, ligand class and redundancy. Dataset curation and coverage analyses describe protein functions/families and lipid classes but do not tabulate organism composition."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Relevant full-paper computational evaluation sections, tables/captions and cited supplementary task passages were reviewed. Reporting omissions are scoped to the inspected sources. Original numerical results are unchanged."}}}}
{"id":"rewire-evaluation-baseline-kmer-position-v2","kind":"evaluation","name":"Corrected k-mer / position baseline on MFASS v2","description":"","status":"reproduced","facets":{"areas":["dna-genomes"]},"source_ids":["rewire-mfass-v2-source"],"links":[{"relation":"model","target_id":"rewire-model-baseline-kmer-position-v2"},{"relation":"benchmark","target_id":"rewire-mfass-v2"},{"relation":"dataset","target_id":"rewire-mfass-v2-dataset"}],"attributes":{"origin":"rewire_run","protocol":"Assay-oriented 21 bp k-mer window, exon position, allele identity and conservation features; gradient-boosted trees trained on the MFASS training split.","version":"bee9133b83f3aedaf2bbb9013f1875515845607e","run_url":"/benchmarks/runs/mfass-v2/","comparison":{"protocol_id":"mfass-v2","dataset_version":"bee9133b83f3aedaf2bbb9013f1875515845607e","split":"split-v2.tsv","population":"8324/8324","inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"coverage":"8324/8324","missing_metadata":{"paired_comparison":"Refer to paired bootstrap artifacts; marginal scores are not paired comparisons."}}}
{"id":"rewire-evaluation-dnabert2-117m-frozen-pair-logreg","kind":"evaluation","name":"DNABERT-2 117M · frozen pair embeddings on MFASS v2","description":"","status":"reproduced","facets":{"areas":["dna-genomes"]},"source_ids":["rewire-mfass-v2-source"],"links":[{"relation":"model","target_id":"rewire-model-dnabert2-117m-frozen-pair-logreg"},{"relation":"benchmark","target_id":"rewire-mfass-v2"},{"relation":"dataset","target_id":"rewire-mfass-v2-dataset"}],"attributes":{"origin":"rewire_run","protocol":"Masked mean of frozen last hidden states for 170 bp reference and mutant sequences; concatenate reference and mutant-minus-reference embeddings; fixed balanced L2 logistic head trained only on the MFASS training split.","version":"bee9133b83f3aedaf2bbb9013f1875515845607e","run_url":"/benchmarks/runs/mfass-v2/","comparison":{"protocol_id":"mfass-v2","dataset_version":"bee9133b83f3aedaf2bbb9013f1875515845607e","split":"split-v2.tsv","population":"8324/8324","inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"coverage":"8324/8324","missing_metadata":{"paired_comparison":"Refer to paired bootstrap artifacts; marginal scores are not paired comparisons."}}}
{"id":"rewire-evaluation-pangolin-maskfalse","kind":"evaluation","name":"Pangolin · mask=False on MFASS v2","description":"","status":"reproduced","facets":{"areas":["dna-genomes"]},"source_ids":["rewire-mfass-v2-source"],"links":[{"relation":"model","target_id":"rewire-model-pangolin-maskfalse"},{"relation":"benchmark","target_id":"rewire-mfass-v2"},{"relation":"dataset","target_id":"rewire-mfass-v2-dataset"}],"attributes":{"origin":"rewire_run","protocol":"Unchanged specialist run in genomic context with GENCODE v44; zero-shot on MFASS assay labels. Point metrics use the scored subset.","version":"bee9133b83f3aedaf2bbb9013f1875515845607e","run_url":"/benchmarks/runs/mfass-v2/","comparison":{"protocol_id":"mfass-v2","dataset_version":"bee9133b83f3aedaf2bbb9013f1875515845607e","split":"split-v2.tsv","population":"8301/8324","inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"coverage":"8301/8324","missing_metadata":{"paired_comparison":"Refer to paired bootstrap artifacts; marginal scores are not paired comparisons."}}}
{"id":"rewire-evaluation-spliceai-1-3-1","kind":"evaluation","name":"SpliceAI 1.3.1 on MFASS v2","description":"","status":"reproduced","facets":{"areas":["dna-genomes"]},"source_ids":["rewire-mfass-v2-source"],"links":[{"relation":"model","target_id":"rewire-model-spliceai-1-3-1"},{"relation":"benchmark","target_id":"rewire-mfass-v2"},{"relation":"dataset","target_id":"rewire-mfass-v2-dataset"}],"attributes":{"origin":"rewire_run","protocol":"Unchanged specialist run in genomic context with bundled annotation; zero-shot on MFASS assay labels. Point metrics use the scored subset.","version":"bee9133b83f3aedaf2bbb9013f1875515845607e","run_url":"/benchmarks/runs/mfass-v2/","comparison":{"protocol_id":"mfass-v2","dataset_version":"bee9133b83f3aedaf2bbb9013f1875515845607e","split":"split-v2.tsv","population":"8194/8324","inputs":null,"adaptation":null,"metric_implementation":null,"aggregation":null,"budget":null},"coverage":"8194/8324","missing_metadata":{"paired_comparison":"Refer to paired bootstrap artifacts; marginal scores are not paired comparisons."}}}
{"id":"rewire-mfass-v1","kind":"benchmark","name":"MFASS v1 (superseded)","description":"The mfass-v1 baseline used a mis-centred k-mer window for 7,770 assay variants whose raw sequence was reverse-complemented. mfass-v2 validates assay-oriented reference and mutant pairs and rebuilds the baseline; mfass-v1 remains a historical record.","status":"superseded","facets":{"areas":["dna-genomes"]},"source_ids":["rewire-mfass-v2-source"],"links":[{"relation":"evaluates_task","target_id":"catalog-task-mfass-splice"}],"attributes":{"entity_level":"protocol","version":"v1","task":"Splice-variant prioritisation","scope_note":"The mfass-v1 baseline used a mis-centred k-mer window for 7,770 assay variants whose raw sequence was reverse-complemented. mfass-v2 validates assay-oriented reference and mutant pairs and rebuilds the baseline; mfass-v1 remains a historical record.","historical_missing_metadata":{},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"MFASS v1 is a superseded historical protocol whose baseline used an incorrectly centred sequence window.","summary_source_ids":["evidence-benchmark-mfass-pinned-readme"],"summary_source_locator":"Pinned README: correction notice; Dataset; Cohort reconciliation; Split; MFASS-v2 results; Paired comparisons; Limits; Archived mfass-v1 snapshot","sections":[{"title":"Evaluation methodology","body":"Historical MFASS exon-recognition evaluation; the correction notice documents a sequence-orientation mismatch in the v1 baseline. The source distinguishes protocol versions from split versions. The earlier exon-only split-cost experiment requires a corrected rerun; its conclusions are withdrawn. Precision at a fixed review capacity, average precision and AUROC. Each method’s point estimates use its scored subset; paired comparisons use common scored variants. Historical baseline and specialist results are preserved for audit. The corrected v2 baseline and DNABERT-2 protocol must not be relabelled as v1. The v1 windowing defect invalidates its baseline comparisons as current evidence; this is separate from the grouped-split controls.","source_ids":["evidence-benchmark-mfass-pinned-readme"],"source_locator":"Pinned README: correction notice; Dataset; Cohort reconciliation; Split; MFASS-v2 results; Paired comparisons; Limits; Archived mfass-v1 snapshot"}],"facts":[{"label":"Datasets","value":"Historical MFASS exon-recognition evaluation; the correction notice documents a sequence-orientation mismatch in the v1 baseline.","status":"source_checked","source_ids":["evidence-benchmark-mfass-pinned-readme"],"source_locator":"Pinned README: correction notice; Dataset; Cohort reconciliation; Split; MFASS-v2 results; Paired comparisons; Limits; Archived mfass-v1 snapshot"},{"label":"Splits","value":"The source distinguishes protocol versions from split versions. The earlier exon-only split-cost experiment requires a corrected rerun; its conclusions are withdrawn.","status":"source_checked","source_ids":["evidence-benchmark-mfass-pinned-readme"],"source_locator":"Pinned README: correction notice; Dataset; Cohort reconciliation; Split; MFASS-v2 results; Paired comparisons; Limits; Archived mfass-v1 snapshot"},{"label":"Metrics","value":"Precision at a fixed review capacity, average precision and AUROC. Each method’s point estimates use its scored subset; paired comparisons use common scored variants.","status":"source_checked","source_ids":["evidence-benchmark-mfass-pinned-readme"],"source_locator":"Pinned README: correction notice; Dataset; Cohort reconciliation; Split; MFASS-v2 results; Paired comparisons; Limits; Archived mfass-v1 snapshot"},{"label":"Baselines","value":"Historical baseline and specialist results are preserved for audit. The corrected v2 baseline and DNABERT-2 protocol must not be relabelled as v1.","status":"source_checked","source_ids":["evidence-benchmark-mfass-pinned-readme"],"source_locator":"Pinned README: correction notice; Dataset; Cohort reconciliation; Split; MFASS-v2 results; Paired comparisons; Limits; Archived mfass-v1 snapshot"},{"label":"Leakage controls","value":"The v1 windowing defect invalidates its baseline comparisons as current evidence; this is separate from the grouped-split controls.","status":"source_checked","source_ids":["evidence-benchmark-mfass-pinned-readme"],"source_locator":"Pinned README: correction notice; Dataset; Cohort reconciliation; Split; MFASS-v2 results; Paired comparisons; Limits; Archived mfass-v1 snapshot"},{"label":"Uncertainty","value":"The archived SpliceAI comparison records 2,000 paired bootstrap draws, 454 independent groups and 95% interval bounds. These quantify sampling uncertainty in the historical run; they do not correct its sequence-window error and must not be reused as current baseline evidence.","status":"source_checked","source_ids":["evidence-mfass-final-v1-comparison","evidence-benchmark-mfass-pinned-readme"],"source_locator":"Archived JSON: independent_groups and paired.*.{draws,ci95_low,ci95_high}; README correction notice"},{"label":"Entity type","value":"Superseded historical computational protocol.","status":"source_checked","source_ids":["evidence-benchmark-mfass-pinned-readme"],"source_locator":"Pinned README: correction notice; Dataset; Cohort reconciliation; Split; MFASS-v2 results; Paired comparisons; Limits; Archived mfass-v1 snapshot"},{"label":"Organisms","value":"Human (Homo sapiens): naturally occurring ExAC variants in or beside human exons, measured in the MFASS minigene assay.","status":"source_checked","source_ids":["evidence-mfass-final-geo"],"source_locator":"GSE120695 SOFT: Series_summary and Sample_organism_ch1 fields"},{"label":"Assays","value":"MFASS functional exon-recognition labels.","status":"source_checked","source_ids":["evidence-benchmark-mfass-pinned-readme"],"source_locator":"Pinned README: correction notice; Dataset; Cohort reconciliation; Split; MFASS-v2 results; Paired comparisons; Limits; Archived mfass-v1 snapshot"},{"label":"Allowed inputs","value":"Historical baseline uses the incorrectly centered sequence window; corrected assay-oriented pairs belong to v2.","status":"source_checked","source_ids":["evidence-benchmark-mfass-pinned-readme"],"source_locator":"Pinned README: correction notice; Dataset; Cohort reconciliation; Split; MFASS-v2 results; Paired comparisons; Limits; Archived mfass-v1 snapshot"},{"label":"Adaptation","value":"Historical supervised baseline; its comparisons and split-cost claims are withdrawn pending corrected reruns.","status":"source_checked","source_ids":["evidence-benchmark-mfass-pinned-readme"],"source_locator":"Pinned README: correction notice; Dataset; Cohort reconciliation; Split; MFASS-v2 results; Paired comparisons; Limits; Archived mfass-v1 snapshot"}],"strengths":[],"limitations":[{"text":"Superseded baseline comparisons and split-cost claims must not be used as current evidence. This profile documents the correction and does not reproduce or rehabilitate the historical results.","source_ids":["evidence-benchmark-mfass-pinned-readme"],"source_locator":"Pinned README: correction notice; Dataset; Cohort reconciliation; Split; MFASS-v2 results; Paired comparisons; Limits; Archived mfass-v1 snapshot"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Historical baseline uses the incorrectly centered sequence window; corrected assay-oriented pairs belong to v2.","Evaluation: Historical supervised baseline; its comparisons and split-cost claims are withdrawn pending corrected reruns.","Readout: Precision at a fixed review capacity, average precision and AUROC. Each method’s point estimates use its scored subset; paired comparisons use common scored variants."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["evidence-benchmark-mfass-pinned-readme"],"source_locator":"Pinned README: correction notice; Dataset; Cohort reconciliation; Split; MFASS-v2 results; Paired comparisons; Limits; Archived mfass-v1 snapshot"},"coverage":"limited","gaps":["Superseded baseline comparisons and split-cost claims must not be used as current evidence. This profile documents the correction and does not reproduce or rehabilitate the historical results."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Task-specific computational methodology and field context checked in the cited primary-source artifact. Source-backed fields, inapplicable evaluator dimensions and unresolved details are distinguished. Numerical results were not reproduced. Original assay organism checked against GEO SOFT metadata; v1 interval procedure checked against its immutable historical JSON without rehabilitating superseded comparisons."}}}}
{"id":"rewire-mfass-v2","kind":"benchmark","name":"MFASS v2","description":"","status":"source_checked","facets":{"areas":["dna-genomes"]},"source_ids":["rewire-mfass-v2-source"],"links":[{"relation":"supersedes","target_id":"rewire-mfass-v1"},{"relation":"evaluates_task","target_id":"catalog-task-mfass-splice"}],"attributes":{"entity_level":"protocol","version":"bee9133b83f3aedaf2bbb9013f1875515845607e","task":"Splice-variant prioritisation","scope_note":"The mfass-v1 baseline used a mis-centred k-mer window for 7,770 assay variants whose raw sequence was reverse-complemented. mfass-v2 validates assay-oriented reference and mutant pairs and rebuilds the baseline; mfass-v1 remains a historical record.","historical_missing_metadata":{},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"MFASS v2 evaluates splice-variant prioritisation using a corrected sequence baseline and a fixed grouped holdout.","summary_source_ids":["evidence-benchmark-mfass-pinned-readme"],"summary_source_locator":"Pinned README: correction notice; Dataset; Cohort reconciliation; Split; MFASS-v2 results; Paired comparisons; Limits; Archived mfass-v1 snapshot","sections":[{"title":"Evaluation methodology","body":"Functional exon-recognition labels from the MFASS assay; validated assay-oriented reference/mutant sequence pairs define the sequence-model inputs. The canonical split assigns whole connected exon/gene groups to training or test. Prevalence matching is selected before model runs and uses no predictions. Precision at a fixed review capacity, average precision and AUROC. Each method’s point estimates use its scored subset; paired comparisons use common scored variants. Corrected k-mer/position baseline, frozen DNABERT-2 pair embeddings with a fixed logistic head, and SpliceAI/Pangolin genomic-context specialists. Baseline and logistic-head fitting use training labels only. Exon/gene connected components prevent linked variants crossing arms; exact DNABERT-2 pretraining overlap has not been checked. Paired intervals resample whole connected exon/gene groups; precision intervals account for fixed review-list fraction during group resampling.","source_ids":["evidence-benchmark-mfass-pinned-readme"],"source_locator":"Pinned README: correction notice; Dataset; Cohort reconciliation; Split; MFASS-v2 results; Paired comparisons; Limits; Archived mfass-v1 snapshot"}],"facts":[{"label":"Datasets","value":"Functional exon-recognition labels from the MFASS assay; validated assay-oriented reference/mutant sequence pairs define the sequence-model inputs.","status":"source_checked","source_ids":["evidence-benchmark-mfass-pinned-readme"],"source_locator":"Pinned README: correction notice; Dataset; Cohort reconciliation; Split; MFASS-v2 results; Paired comparisons; Limits; Archived mfass-v1 snapshot"},{"label":"Splits","value":"The canonical split assigns whole connected exon/gene groups to training or test. Prevalence matching is selected before model runs and uses no predictions.","status":"source_checked","source_ids":["evidence-benchmark-mfass-pinned-readme"],"source_locator":"Pinned README: correction notice; Dataset; Cohort reconciliation; Split; MFASS-v2 results; Paired comparisons; Limits; Archived mfass-v1 snapshot"},{"label":"Metrics","value":"Precision at a fixed review capacity, average precision and AUROC. Each method’s point estimates use its scored subset; paired comparisons use common scored variants.","status":"source_checked","source_ids":["evidence-benchmark-mfass-pinned-readme"],"source_locator":"Pinned README: correction notice; Dataset; Cohort reconciliation; Split; MFASS-v2 results; Paired comparisons; Limits; Archived mfass-v1 snapshot"},{"label":"Baselines","value":"Corrected k-mer/position baseline, frozen DNABERT-2 pair embeddings with a fixed logistic head, and SpliceAI/Pangolin genomic-context specialists.","status":"source_checked","source_ids":["evidence-benchmark-mfass-pinned-readme"],"source_locator":"Pinned README: correction notice; Dataset; Cohort reconciliation; Split; MFASS-v2 results; Paired comparisons; Limits; Archived mfass-v1 snapshot"},{"label":"Leakage controls","value":"Baseline and logistic-head fitting use training labels only. Exon/gene connected components prevent linked variants crossing arms; exact DNABERT-2 pretraining overlap has not been checked.","status":"source_checked","source_ids":["evidence-benchmark-mfass-pinned-readme"],"source_locator":"Pinned README: correction notice; Dataset; Cohort reconciliation; Split; MFASS-v2 results; Paired comparisons; Limits; Archived mfass-v1 snapshot"},{"label":"Uncertainty","value":"Paired intervals resample whole connected exon/gene groups; precision intervals account for fixed review-list fraction during group resampling.","status":"source_checked","source_ids":["evidence-benchmark-mfass-pinned-readme"],"source_locator":"Pinned README: correction notice; Dataset; Cohort reconciliation; Split; MFASS-v2 results; Paired comparisons; Limits; Archived mfass-v1 snapshot"},{"label":"Entity type","value":"Paper-specific computational evaluation protocol.","status":"source_checked","source_ids":["evidence-benchmark-mfass-pinned-readme"],"source_locator":"Pinned README: correction notice; Dataset; Cohort reconciliation; Split; MFASS-v2 results; Paired comparisons; Limits; Archived mfass-v1 snapshot"},{"label":"Organisms","value":"Human (Homo sapiens): naturally occurring ExAC variants in or beside human exons, measured in the MFASS minigene assay.","status":"source_checked","source_ids":["evidence-mfass-final-geo"],"source_locator":"GSE120695 SOFT: Series_summary and Sample_organism_ch1 fields"},{"label":"Assays","value":"MFASS functional exon-recognition labels.","status":"source_checked","source_ids":["evidence-benchmark-mfass-pinned-readme"],"source_locator":"Pinned README: correction notice; Dataset; Cohort reconciliation; Split; MFASS-v2 results; Paired comparisons; Limits; Archived mfass-v1 snapshot"},{"label":"Allowed inputs","value":"Validated assay-oriented sequence pairs for baseline/DNABERT-2; specialists use genomic context.","status":"source_checked","source_ids":["evidence-benchmark-mfass-pinned-readme"],"source_locator":"Pinned README: correction notice; Dataset; Cohort reconciliation; Split; MFASS-v2 results; Paired comparisons; Limits; Archived mfass-v1 snapshot"},{"label":"Adaptation","value":"Baseline and logistic head use MFASS training labels; specialists are zero-shot on the assay.","status":"source_checked","source_ids":["evidence-benchmark-mfass-pinned-readme"],"source_locator":"Pinned README: correction notice; Dataset; Cohort reconciliation; Split; MFASS-v2 results; Paired comparisons; Limits; Archived mfass-v1 snapshot"}],"strengths":[{"text":"Connected exon/gene groups are preserved in both splitting and paired bootstrap uncertainty.","source_ids":["evidence-benchmark-mfass-pinned-readme"],"source_locator":"Pinned README: correction notice; Dataset; Cohort reconciliation; Split; MFASS-v2 results; Paired comparisons; Limits; Archived mfass-v1 snapshot"}],"limitations":[{"text":"The assay measures an artificial construct rather than patient RNA. Supervised assay-input models and zero-shot genomic-context specialists have different input/adaptation regimes; specialist annotation versions also differ.","source_ids":["evidence-benchmark-mfass-pinned-readme"],"source_locator":"Pinned README: correction notice; Dataset; Cohort reconciliation; Split; MFASS-v2 results; Paired comparisons; Limits; Archived mfass-v1 snapshot"}],"diagram":{"title":"Computational evaluation flow","steps":["Input: Validated assay-oriented sequence pairs for baseline/DNABERT-2; specialists use genomic context.","Evaluation: Baseline and logistic head use MFASS training labels; specialists are zero-shot on the assay.","Readout: Precision at a fixed review capacity, average precision and AUROC. Each method’s point estimates use its scored subset; paired comparisons use common scored variants."],"caption":"Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.","source_ids":["evidence-benchmark-mfass-pinned-readme"],"source_locator":"Pinned README: correction notice; Dataset; Cohort reconciliation; Split; MFASS-v2 results; Paired comparisons; Limits; Archived mfass-v1 snapshot"},"coverage":"limited","gaps":["The assay measures an artificial construct rather than patient RNA. Supervised assay-input models and zero-shot genomic-context specialists have different input/adaptation regimes; specialist annotation versions also differ."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Task-specific computational methodology and field context checked in the cited primary-source artifact. Source-backed fields, inapplicable evaluator dimensions and unresolved details are distinguished. Numerical results were not reproduced. Original assay organism checked against GEO SOFT metadata; v1 interval procedure checked against its immutable historical JSON without rehabilitating superseded comparisons."}}}}
{"id":"rewire-mfass-v2-dataset","kind":"dataset","name":"MFASS v2 eligible assay cohort","description":"","status":"source_checked","facets":{"areas":["dna-genomes"]},"source_ids":["rewire-mfass-v2-source"],"links":[],"attributes":{"version":"bee9133b83f3aedaf2bbb9013f1875515845607e","split":"split-v2.tsv","cohort_variants":27733,"train_variants":19409,"test_variants":8324,"test_positives":315,"independent_test_groups":463,"missing_metadata":{}}}
{"id":"rewire-mfass-v2-source","kind":"source","name":"MFASS v2 pinned rewire artifacts","description":"","status":"source_checked","facets":{},"source_ids":[],"links":[],"attributes":{"url":"https://github.com/timini/rewire-benchmarks/tree/bee9133b83f3aedaf2bbb9013f1875515845607e","version":"bee9133b83f3aedaf2bbb9013f1875515845607e","retrieved_at":"2026-09-16T10:50:02Z","artifact_sha256":"a3af693afc070b39b334e359beed7f9affd76f7456a4f9264e2d1e9dab26111d"}}
{"id":"rewire-model-baseline-kmer-position-v2","kind":"model","name":"Corrected k-mer / position baseline","description":"","status":"source_checked","facets":{"areas":["dna-genomes"]},"source_ids":["rewire-mfass-v2-source"],"links":[],"attributes":{"entity_level":"method","version":"bee9133b83f3aedaf2bbb9013f1875515845607e","reported_name":"Corrected k-mer / position baseline","historical_missing_metadata":{},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"The corrected MFASS baseline combines local sequence composition, allele identity, exon position and conservation in gradient-boosted trees.","summary_source_ids":["evidence-reported-rewire-run-baseline"],"summary_source_locator":"run_baseline.py module documentation and recorded runner configuration","sections":[{"title":"Evaluated procedure","body":"Features use an assay-oriented 21-base window centred on the validated variant position. The model is trained on the fixed MFASS training split, then scores the held-out variants. The corrected orientation is part of the method identity.","source_ids":["rewire-mfass-v2-source"],"source_locator":"benchmarks/mfass/src/mfass/run_baseline.py, featurise; results/baseline-kmer-position-v2.json at bee9133b83f3aedaf2bbb9013f1875515845607e"}],"facts":[{"label":"Model type","value":"Train-only gradient-boosted-tree baseline","status":"source_checked","source_ids":["evidence-reported-rewire-run-baseline"],"source_locator":"run_baseline.py; module documentation, main() configuration and scoring procedure"},{"label":"Architecture / procedure","value":"HistGradientBoostingClassifier on allele, exon-distance, conservation and 3-mer features. Corrected windows use validated assay-oriented mutant sequences.","status":"source_checked","source_ids":["evidence-reported-rewire-run-baseline"],"source_locator":"run_baseline.py; module documentation, main() configuration and scoring procedure"},{"label":"Biological inputs","value":"Assay-oriented allele pairs, variant/exon positions, phyloP and phastCons annotations","status":"source_checked","source_ids":["evidence-reported-rewire-run-baseline"],"source_locator":"run_baseline.py; module documentation, main() configuration and scoring procedure"},{"label":"Outputs","value":"Probability of MFASS splice disruption","status":"source_checked","source_ids":["evidence-reported-rewire-run-baseline"],"source_locator":"run_baseline.py; module documentation, main() configuration and scoring procedure"},{"label":"Parameters","value":"Not a pretrained neural parameter count; fitted tree structure is controlled by the runner configuration.","status":"source_checked","source_ids":["evidence-reported-rewire-run-baseline"],"source_locator":"run_baseline.py; module documentation, main() configuration and scoring procedure"},{"label":"Training data / fitting","value":"19,409 MFASS train variants in split-v2; feature fitting and model training use train rows only.","status":"source_checked","source_ids":["evidence-reported-rewire-run-baseline"],"source_locator":"run_baseline.py; module documentation, main() configuration and scoring procedure"},{"label":"Context limits","value":"A 21-bp variant-centred sequence window supplies 3-mer composition; other positional and conservation features are also provided.","status":"source_checked","source_ids":["evidence-reported-rewire-run-baseline"],"source_locator":"run_baseline.py; module documentation, main() configuration and scoring procedure"},{"label":"Known versions / configuration","value":"bee9133b83f3aedaf2bbb9013f1875515845607e","status":"source_checked","source_ids":["evidence-reported-rewire-run-baseline"],"source_locator":"run_baseline.py; module documentation, main() configuration and scoring procedure"},{"label":"Access","value":"Public pinned benchmark source; no genome retrieval or pretrained weight download required.","status":"source_checked","source_ids":["evidence-reported-rewire-run-baseline"],"source_locator":"run_baseline.py; module documentation, main() configuration and scoring procedure"},{"label":"Code licence","value":"MIT for the Rewire runner at the reviewed revision.","status":"source_checked","source_ids":["evidence-reported-rewire-license"],"source_locator":"LICENSE; complete MIT grant"},{"label":"Weights licence","value":"Not applicable to a pretrained model; the fitted baseline is an experiment artifact.","status":"inapplicable","source_ids":["evidence-reported-rewire-run-baseline"],"source_locator":"run_baseline.py; module documentation, main() configuration and scoring procedure"}],"strengths":[{"text":"Provides an interpretable feature-based reference for asking whether a more complex model adds value.","source_ids":["rewire-mfass-v2-source"],"source_locator":"benchmarks/mfass/src/mfass/run_baseline.py, featurise; results/baseline-kmer-position-v2.json at bee9133b83f3aedaf2bbb9013f1875515845607e"}],"limitations":[{"text":"It uses assay-specific labels and engineered annotation features. It is not a zero-shot baseline with the same inputs as every specialist.","source_ids":["rewire-mfass-v2-source"],"source_locator":"benchmarks/mfass/src/mfass/run_baseline.py, featurise; results/baseline-kmer-position-v2.json at bee9133b83f3aedaf2bbb9013f1875515845607e"}],"diagram":{"title":"Evaluated pipeline","steps":["Validated assay sequence","21-base window and features","Training split","Gradient-boosted trees","Held-out ranking"],"caption":"Schematic of the pinned MFASS configuration; this does not generalise to every member of the model family.","source_ids":["rewire-mfass-v2-source"],"source_locator":"benchmarks/mfass/src/mfass/run_baseline.py, featurise; results/baseline-kmer-position-v2.json at bee9133b83f3aedaf2bbb9013f1875515845607e"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Pinned MFASS runner source rechecked at bee9133b83f3aedaf2bbb9013f1875515845607e. Existing numerical records and historical artifacts remain unchanged; no new run was performed."}}}}
{"id":"rewire-model-dnabert2-117m-frozen-pair-logreg","kind":"model","name":"DNABERT-2 117M · frozen pair embeddings","description":"","status":"source_checked","facets":{"areas":["dna-genomes"]},"source_ids":["rewire-mfass-v2-source"],"links":[{"relation":"uses_model","target_id":"discovery-model-dnabert-2"}],"attributes":{"entity_level":"method","version":"bee9133b83f3aedaf2bbb9013f1875515845607e","reported_name":"DNABERT-2 117M · frozen pair embeddings","historical_missing_metadata":{},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"This MFASS pipeline uses a frozen DNABERT-2 encoder with an assay-specific logistic-regression head. It is distinct from the base encoder and from end-to-end fine-tuning.","summary_source_ids":["evidence-reported-rewire-run-dnabert2"],"summary_source_locator":"run_dnabert2.py module documentation and recorded runner configuration","sections":[{"title":"Evaluated procedure","body":"Validated 170-base reference and mutant sequences are embedded separately. Attention-mask means of final hidden states are combined as reference plus mutant-minus-reference. Standardisation and balanced L2 logistic regression are fitted only on the training arm.","source_ids":["rewire-mfass-v2-source"],"source_locator":"benchmarks/mfass/README.md, DNABERT-2 protocol; results/dnabert2-117m-frozen-pair-logreg.json, config at bee9133b83f3aedaf2bbb9013f1875515845607e"}],"facts":[{"label":"Model type","value":"Frozen DNA encoder plus fitted logistic-regression pipeline","status":"source_checked","source_ids":["evidence-reported-rewire-run-dnabert2"],"source_locator":"run_dnabert2.py; module documentation, main() configuration and scoring procedure"},{"label":"Architecture / procedure","value":"Attention-mask mean pooling of last hidden states for reference and mutant sequences; concatenate reference embedding with mutant-minus-reference, scale features and fit balanced L2 logistic regression with C=0.1 on train rows only.","status":"source_checked","source_ids":["evidence-reported-rewire-run-dnabert2"],"source_locator":"run_dnabert2.py; module documentation, main() configuration and scoring procedure"},{"label":"Biological inputs","value":"Validated 170-bp assay-oriented reference/mutant pairs","status":"source_checked","source_ids":["evidence-reported-rewire-run-dnabert2"],"source_locator":"run_dnabert2.py; module documentation, main() configuration and scoring procedure"},{"label":"Outputs","value":"MFASS splice-disruption probabilities from the fitted logistic head","status":"source_checked","source_ids":["evidence-reported-rewire-run-dnabert2"],"source_locator":"run_dnabert2.py; module documentation, main() configuration and scoring procedure"},{"label":"Parameters","value":"117M frozen DNABERT-2 encoder; the logistic head is a separate fitted component.","status":"source_checked","source_ids":["evidence-reported-rewire-run-dnabert2"],"source_locator":"run_dnabert2.py; module documentation, main() configuration and scoring procedure"},{"label":"Training data / fitting","value":"The encoder remains frozen; StandardScaler and logistic regression fit split-v2 training variants without test tuning.","status":"source_checked","source_ids":["evidence-reported-rewire-run-dnabert2"],"source_locator":"run_dnabert2.py; module documentation, main() configuration and scoring procedure"},{"label":"Context limits","value":"170 bases per sequence; tokenizer truncation disabled.","status":"source_checked","source_ids":["evidence-reported-rewire-run-dnabert2"],"source_locator":"run_dnabert2.py; module documentation, main() configuration and scoring procedure"},{"label":"Known versions / configuration","value":"zhihan1996/DNABERT-2-117M; runner revision bee9133b83f3aedaf2bbb9013f1875515845607e","status":"source_checked","source_ids":["evidence-reported-rewire-run-dnabert2"],"source_locator":"run_dnabert2.py; module documentation, main() configuration and scoring procedure"},{"label":"Access","value":"Pinned public runner; model and tokenizer identities are validated against recorded artifact hashes.","status":"source_checked","source_ids":["evidence-reported-rewire-run-dnabert2"],"source_locator":"run_dnabert2.py; module documentation, main() configuration and scoring procedure"},{"label":"Code licence","value":"MIT for the Rewire runner at the reviewed revision. Upstream DNABERT-2: Apache 2.0. These are separate component licences.","status":"source_checked","source_ids":["evidence-reported-rewire-license","evidence-reported-base-dnabert2-license"],"source_locator":"LICENSE; complete MIT grant; upstream LICENSE"},{"label":"Weights licence","value":"The official DNABERT-2-117M checkpoint repository carries Apache 2.0. The separately fitted logistic-regression artifact is not assigned an additional inferred licence.","status":"source_checked","source_ids":["evidence-reported-base-dnabert2-card-license"],"source_locator":"LICENSE in official checkpoint repository"}],"strengths":[{"text":"Separates the utility of a frozen representation from encoder fine-tuning and pins the checkpoint and head settings.","source_ids":["rewire-mfass-v2-source"],"source_locator":"benchmarks/mfass/README.md, DNABERT-2 protocol; results/dnabert2-117m-frozen-pair-logreg.json, config at bee9133b83f3aedaf2bbb9013f1875515845607e"}],"limitations":[{"text":"The result tests this short-context supervised pipeline. It does not establish how all DNABERT-2 adaptations or all foundation models perform.","source_ids":["rewire-mfass-v2-source"],"source_locator":"benchmarks/mfass/README.md, DNABERT-2 protocol; results/dnabert2-117m-frozen-pair-logreg.json, config at bee9133b83f3aedaf2bbb9013f1875515845607e"}],"diagram":{"title":"Evaluated pipeline","steps":["Assay-oriented sequence pair","Frozen DNABERT-2","Masked mean embeddings","Reference plus difference","Trained logistic head"],"caption":"Schematic of the pinned MFASS configuration; this does not generalise to every member of the model family.","source_ids":["rewire-mfass-v2-source"],"source_locator":"benchmarks/mfass/README.md, DNABERT-2 protocol; results/dnabert2-117m-frozen-pair-logreg.json, config at bee9133b83f3aedaf2bbb9013f1875515845607e"},"coverage":"reviewed","gaps":[],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Pinned MFASS runner source rechecked at bee9133b83f3aedaf2bbb9013f1875515845607e. Existing numerical records and historical artifacts remain unchanged; no new run was performed."}}}}
{"id":"rewire-model-pangolin-maskfalse","kind":"model","name":"Pangolin · mask=False","description":"","status":"source_checked","facets":{"areas":["dna-genomes"]},"source_ids":["rewire-mfass-v2-source"],"links":[{"relation":"uses_model","target_id":"discovery-model-pangolin"}],"attributes":{"entity_level":"method","version":"bee9133b83f3aedaf2bbb9013f1875515845607e","reported_name":"Pangolin · mask=False","historical_missing_metadata":{},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"This MFASS evaluation uses Pangolin with masking disabled, scoring variants in genomic context.","summary_source_ids":["evidence-reported-rewire-run-pangolin"],"summary_source_locator":"run_pangolin.py module documentation and recorded runner configuration","sections":[{"title":"Evaluated procedure","body":"The runner uses the specified reference and GENCODE annotation to predict changes in splice strength. The mask=False choice retains changes that annotation-based masking would remove. These specialist predictions were retained unchanged in the corrected v2 comparison.","source_ids":["rewire-mfass-v2-source"],"source_locator":"benchmarks/mfass/README.md, MFASS-v2 results; results/pangolin-maskFalse.json, config and notes at bee9133b83f3aedaf2bbb9013f1875515845607e"}],"facts":[{"label":"Model type","value":"Pretrained splice-usage specialist ensemble","status":"source_checked","source_ids":["evidence-reported-rewire-run-pangolin"],"source_locator":"run_pangolin.py; module documentation, main() configuration and scoring procedure"},{"label":"Architecture / procedure","value":"Official 12-model Pangolin ensemble; use maximum absolute predicted splice-usage change with masking disabled.","status":"source_checked","source_ids":["evidence-reported-rewire-run-pangolin"],"source_locator":"run_pangolin.py; module documentation, main() configuration and scoring procedure"},{"label":"Biological inputs","value":"Variant genomic coordinates, GRCh38 reference and GENCODE annotation","status":"source_checked","source_ids":["evidence-reported-rewire-run-pangolin"],"source_locator":"run_pangolin.py; module documentation, main() configuration and scoring procedure"},{"label":"Outputs","value":"Maximum absolute predicted splice-usage change","status":"source_checked","source_ids":["evidence-reported-rewire-run-pangolin"],"source_locator":"run_pangolin.py; module documentation, main() configuration and scoring procedure"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["evidence-reported-rewire-run-pangolin"],"source_locator":"Pinned runner configuration and implementation; inspected for aggregate parameter count (component sizes are not added without an exact configuration)"},{"label":"Training data / fitting","value":"No fitting on MFASS assay outcomes in this runner. Upstream training overlap at assayed exons remains unchecked.","status":"source_checked","source_ids":["evidence-reported-rewire-run-pangolin"],"source_locator":"run_pangolin.py; module documentation, main() configuration and scoring procedure"},{"label":"Context limits","value":"10,000-bp genomic context and 50-base scoring distance in the recorded configuration.","status":"source_checked","source_ids":["evidence-reported-rewire-run-pangolin"],"source_locator":"run_pangolin.py; module documentation, main() configuration and scoring procedure"},{"label":"Known versions / configuration","value":"Pangolin official final.{1,2,3}.{0,2,4,6}.3.v2 ensemble; mask=False","status":"source_checked","source_ids":["evidence-reported-rewire-run-pangolin"],"source_locator":"run_pangolin.py; module documentation, main() configuration and scoring procedure"},{"label":"Access","value":"Public pinned runner with external reference, annotation and model dependencies.","status":"source_checked","source_ids":["evidence-reported-rewire-run-pangolin"],"source_locator":"run_pangolin.py; module documentation, main() configuration and scoring procedure"},{"label":"Code licence","value":"MIT for the Rewire runner at the reviewed revision. Upstream Pangolin: GPLv3. These are separate component licences.","status":"source_checked","source_ids":["evidence-reported-rewire-license","evidence-reported-base-pangolin-license"],"source_locator":"LICENSE; complete MIT grant; upstream LICENSE"},{"label":"Weights licence","value":"A separate grant specifically naming the Pangolin weight files is not stated in the inspected README or repository GPL licence; code and weight scope require clarification.","status":"unreported","source_ids":["evidence-reported-base-pangolin-readme-md"],"source_locator":"README.md; model installation and full repository LICENSE"}],"strengths":[{"text":"Tests a pretrained splice specialist without learning from MFASS assay labels.","source_ids":["rewire-mfass-v2-source"],"source_locator":"benchmarks/mfass/README.md, MFASS-v2 results; results/pangolin-maskFalse.json, config and notes at bee9133b83f3aedaf2bbb9013f1875515845607e"}],"limitations":[{"text":"Results depend on annotation, masking and scored coverage. This configuration is not interchangeable with Pangolin’s masked default.","source_ids":["rewire-mfass-v2-source"],"source_locator":"benchmarks/mfass/README.md, MFASS-v2 results; results/pangolin-maskFalse.json, config and notes at bee9133b83f3aedaf2bbb9013f1875515845607e"}],"diagram":{"title":"Evaluated pipeline","steps":["Variant and genomic context","Pangolin predictor","Splice-strength changes","mask=False aggregation","Variant ranking"],"caption":"Schematic of the pinned MFASS configuration; this does not generalise to every member of the model family.","source_ids":["rewire-mfass-v2-source"],"source_locator":"benchmarks/mfass/README.md, MFASS-v2 results; results/pangolin-maskFalse.json, config and notes at bee9133b83f3aedaf2bbb9013f1875515845607e"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","A separate grant specifically naming the Pangolin weight files is not stated in the inspected README or repository GPL licence; code and weight scope require clarification."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Pinned MFASS runner source rechecked at bee9133b83f3aedaf2bbb9013f1875515845607e. Existing numerical records and historical artifacts remain unchanged; no new run was performed."}}}}
{"id":"rewire-model-spliceai-1-3-1","kind":"model","name":"SpliceAI 1.3.1","description":"","status":"source_checked","facets":{"areas":["dna-genomes"]},"source_ids":["rewire-mfass-v2-source"],"links":[{"relation":"uses_model","target_id":"discovery-model-spliceai"}],"attributes":{"entity_level":"method","version":"bee9133b83f3aedaf2bbb9013f1875515845607e","reported_name":"SpliceAI 1.3.1","historical_missing_metadata":{},"metadata_review_scope":"historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.","profile":{"summary":"This MFASS evaluation uses the official SpliceAI 1.3.1 five-model ensemble in genomic context.","summary_source_ids":["evidence-reported-rewire-run-spliceai"],"summary_source_locator":"run_spliceai.py module documentation and recorded runner configuration","sections":[{"title":"Evaluated procedure","body":"The runner scores variants against GRCh38 with bundled annotations and unmasked outputs. It ranks variants by the largest acceptor/donor gain or loss delta score. Its specialist predictions were retained unchanged for the corrected v2 comparison.","source_ids":["rewire-mfass-v2-source"],"source_locator":"benchmarks/mfass/README.md, MFASS-v2 results; results/spliceai-1.3.1.json, description, config and notes at bee9133b83f3aedaf2bbb9013f1875515845607e"}],"facts":[{"label":"Model type","value":"Pretrained splice-site specialist ensemble","status":"source_checked","source_ids":["evidence-reported-rewire-run-spliceai"],"source_locator":"run_spliceai.py; module documentation, main() configuration and scoring procedure"},{"label":"Architecture / procedure","value":"Official five-model SpliceAI ensemble; score is max(DS_AG, DS_AL, DS_DG, DS_DL), with masking disabled.","status":"source_checked","source_ids":["evidence-reported-rewire-run-spliceai"],"source_locator":"run_spliceai.py; module documentation, main() configuration and scoring procedure"},{"label":"Biological inputs","value":"Variant genomic coordinates and GRCh38 reference/annotation","status":"source_checked","source_ids":["evidence-reported-rewire-run-spliceai"],"source_locator":"run_spliceai.py; module documentation, main() configuration and scoring procedure"},{"label":"Outputs","value":"Maximum splice acceptor/donor gain/loss delta score","status":"source_checked","source_ids":["evidence-reported-rewire-run-spliceai"],"source_locator":"run_spliceai.py; module documentation, main() configuration and scoring procedure"},{"label":"Parameters","value":"An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources.","status":"unreported","source_ids":["evidence-reported-rewire-run-spliceai"],"source_locator":"Pinned runner configuration and implementation; inspected for aggregate parameter count (component sizes are not added without an exact configuration)"},{"label":"Training data / fitting","value":"No MFASS-label fitting by the runner. Upstream transcript overlap with assayed exons remains unchecked.","status":"source_checked","source_ids":["evidence-reported-rewire-run-spliceai"],"source_locator":"run_spliceai.py; module documentation, main() configuration and scoring procedure"},{"label":"Context limits","value":"10,000-bp genomic context; distance_D=50; mask_M=0.","status":"source_checked","source_ids":["evidence-reported-rewire-run-spliceai"],"source_locator":"run_spliceai.py; module documentation, main() configuration and scoring procedure"},{"label":"Known versions / configuration","value":"SpliceAI 1.3.1; bundled spliceai1–5.h5 weights","status":"source_checked","source_ids":["evidence-reported-rewire-run-spliceai"],"source_locator":"run_spliceai.py; module documentation, main() configuration and scoring procedure"},{"label":"Access","value":"Public pinned benchmark runner with external genome and SpliceAI dependencies.","status":"source_checked","source_ids":["evidence-reported-rewire-run-spliceai"],"source_locator":"run_spliceai.py; module documentation, main() configuration and scoring procedure"},{"label":"Code licence","value":"MIT for the Rewire runner at the reviewed revision. Upstream SpliceAI: PolyForm Strict 1.0.0. These are separate component licences.","status":"source_checked","source_ids":["evidence-reported-rewire-license","evidence-reported-base-spliceai-license"],"source_locator":"LICENSE; complete MIT grant; upstream LICENSE"},{"label":"Weights licence","value":"SpliceAI trained models are CC BY-NC 4.0 for academic/non-commercial use; commercial use requires separate Illumina terms.","status":"source_checked","source_ids":["evidence-reported-base-spliceai-readme-md"],"source_locator":"README.md; License"}],"strengths":[{"text":"Uses a pretrained splice specialist without fitting to MFASS labels.","source_ids":["rewire-mfass-v2-source"],"source_locator":"benchmarks/mfass/README.md, MFASS-v2 results; results/spliceai-1.3.1.json, description, config and notes at bee9133b83f3aedaf2bbb9013f1875515845607e"}],"limitations":[{"text":"The genomic context differs from the assay construct. Missing predictions and possible exon overlap with specialist training data must be considered.","source_ids":["rewire-mfass-v2-source"],"source_locator":"benchmarks/mfass/README.md, MFASS-v2 results; results/spliceai-1.3.1.json, description, config and notes at bee9133b83f3aedaf2bbb9013f1875515845607e"}],"diagram":{"title":"Evaluated pipeline","steps":["Variant in GRCh38","SpliceAI ensemble","Acceptor / donor delta scores","Maximum delta","Variant ranking"],"caption":"Schematic of the pinned MFASS configuration; this does not generalise to every member of the model family.","source_ids":["rewire-mfass-v2-source"],"source_locator":"benchmarks/mfass/README.md, MFASS-v2 results; results/spliceai-1.3.1.json, description, config and notes at bee9133b83f3aedaf2bbb9013f1875515845607e"},"coverage":"limited","gaps":["An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources."],"review":{"method":"automated_source_review","date":"2026-09-16","note":"Pinned MFASS runner source rechecked at bee9133b83f3aedaf2bbb9013f1875515845607e. Existing numerical records and historical artifacts remain unchanged; no new run was performed."}}}}
{"id":"rewire-result-baseline-kmer-position-v2-auroc","kind":"result","name":"Corrected k-mer / position baseline · auroc","description":"","status":"reproduced","facets":{"areas":["dna-genomes"]},"source_ids":["rewire-mfass-v2-source"],"links":[{"relation":"evaluation","target_id":"rewire-evaluation-baseline-kmer-position-v2"}],"attributes":{"printed_value":"0.7779498064677238","numeric_value":"0.7779498064677238","metric":"auroc","metric_direction":"higher","unit":"fraction","uncertainty":null,"source_locator":"benchmarks/mfass/results/baseline-kmer-position-v2.json :: auroc","review":{"method":"existing_run_import","reviewer":"rewire documented MFASS v2 run","reviewed_at":"2026-09-16T10:50:02Z","notes":"Imported existing documented own-run record, not a newly executed reproduction."},"coverage":"8324/8324","missing_metadata":{"uncertainty":"Point estimate; see paired-comparison artifacts."}}}
{"id":"rewire-result-baseline-kmer-position-v2-average-precision","kind":"result","name":"Corrected k-mer / position baseline · average_precision","description":"","status":"reproduced","facets":{"areas":["dna-genomes"]},"source_ids":["rewire-mfass-v2-source"],"links":[{"relation":"evaluation","target_id":"rewire-evaluation-baseline-kmer-position-v2"}],"attributes":{"printed_value":"0.28641674595892375","numeric_value":"0.28641674595892375","metric":"average_precision","metric_direction":"higher","unit":"fraction","uncertainty":null,"source_locator":"benchmarks/mfass/results/baseline-kmer-position-v2.json :: average_precision","review":{"method":"existing_run_import","reviewer":"rewire documented MFASS v2 run","reviewed_at":"2026-09-16T10:50:02Z","notes":"Imported existing documented own-run record, not a newly executed reproduction."},"coverage":"8324/8324","missing_metadata":{"uncertainty":"Point estimate; see paired-comparison artifacts."}}}
{"id":"rewire-result-baseline-kmer-position-v2-precision-at-100","kind":"result","name":"Corrected k-mer / position baseline · precision_at_100","description":"","status":"reproduced","facets":{"areas":["dna-genomes"]},"source_ids":["rewire-mfass-v2-source"],"links":[{"relation":"evaluation","target_id":"rewire-evaluation-baseline-kmer-position-v2"}],"attributes":{"printed_value":"0.61","numeric_value":"0.61","metric":"precision_at_100","metric_direction":"higher","unit":"fraction","uncertainty":null,"source_locator":"benchmarks/mfass/results/baseline-kmer-position-v2.json :: precision_at_100","review":{"method":"existing_run_import","reviewer":"rewire documented MFASS v2 run","reviewed_at":"2026-09-16T10:50:02Z","notes":"Imported existing documented own-run record, not a newly executed reproduction."},"coverage":"8324/8324","missing_metadata":{"uncertainty":"Point estimate; see paired-comparison artifacts."}}}
{"id":"rewire-result-dnabert2-117m-frozen-pair-logreg-auroc","kind":"result","name":"DNABERT-2 117M · frozen pair embeddings · auroc","description":"","status":"reproduced","facets":{"areas":["dna-genomes"]},"source_ids":["rewire-mfass-v2-source"],"links":[{"relation":"evaluation","target_id":"rewire-evaluation-dnabert2-117m-frozen-pair-logreg"}],"attributes":{"printed_value":"0.5500324040216661","numeric_value":"0.5500324040216661","metric":"auroc","metric_direction":"higher","unit":"fraction","uncertainty":null,"source_locator":"benchmarks/mfass/results/dnabert2-117m-frozen-pair-logreg.json :: auroc","review":{"method":"existing_run_import","reviewer":"rewire documented MFASS v2 run","reviewed_at":"2026-09-16T10:50:02Z","notes":"Imported existing documented own-run record, not a newly executed reproduction."},"coverage":"8324/8324","missing_metadata":{"uncertainty":"Point estimate; see paired-comparison artifacts."}}}
{"id":"rewire-result-dnabert2-117m-frozen-pair-logreg-average-precision","kind":"result","name":"DNABERT-2 117M · frozen pair embeddings · average_precision","description":"","status":"reproduced","facets":{"areas":["dna-genomes"]},"source_ids":["rewire-mfass-v2-source"],"links":[{"relation":"evaluation","target_id":"rewire-evaluation-dnabert2-117m-frozen-pair-logreg"}],"attributes":{"printed_value":"0.04508654312652131","numeric_value":"0.04508654312652131","metric":"average_precision","metric_direction":"higher","unit":"fraction","uncertainty":null,"source_locator":"benchmarks/mfass/results/dnabert2-117m-frozen-pair-logreg.json :: average_precision","review":{"method":"existing_run_import","reviewer":"rewire documented MFASS v2 run","reviewed_at":"2026-09-16T10:50:02Z","notes":"Imported existing documented own-run record, not a newly executed reproduction."},"coverage":"8324/8324","missing_metadata":{"uncertainty":"Point estimate; see paired-comparison artifacts."}}}
{"id":"rewire-result-dnabert2-117m-frozen-pair-logreg-precision-at-100","kind":"result","name":"DNABERT-2 117M · frozen pair embeddings · precision_at_100","description":"","status":"reproduced","facets":{"areas":["dna-genomes"]},"source_ids":["rewire-mfass-v2-source"],"links":[{"relation":"evaluation","target_id":"rewire-evaluation-dnabert2-117m-frozen-pair-logreg"}],"attributes":{"printed_value":"0.03","numeric_value":"0.03","metric":"precision_at_100","metric_direction":"higher","unit":"fraction","uncertainty":null,"source_locator":"benchmarks/mfass/results/dnabert2-117m-frozen-pair-logreg.json :: precision_at_100","review":{"method":"existing_run_import","reviewer":"rewire documented MFASS v2 run","reviewed_at":"2026-09-16T10:50:02Z","notes":"Imported existing documented own-run record, not a newly executed reproduction."},"coverage":"8324/8324","missing_metadata":{"uncertainty":"Point estimate; see paired-comparison artifacts."}}}
{"id":"rewire-result-pangolin-maskfalse-auroc","kind":"result","name":"Pangolin · mask=False · auroc","description":"","status":"reproduced","facets":{"areas":["dna-genomes"]},"source_ids":["rewire-mfass-v2-source"],"links":[{"relation":"evaluation","target_id":"rewire-evaluation-pangolin-maskfalse"}],"attributes":{"printed_value":"0.8756851300560864","numeric_value":"0.8756851300560864","metric":"auroc","metric_direction":"higher","unit":"fraction","uncertainty":null,"source_locator":"benchmarks/mfass/results/pangolin-maskFalse.json :: auroc","review":{"method":"existing_run_import","reviewer":"rewire documented MFASS v2 run","reviewed_at":"2026-09-16T10:50:02Z","notes":"Imported existing documented own-run record, not a newly executed reproduction."},"coverage":"8301/8324","missing_metadata":{"uncertainty":"Point estimate; see paired-comparison artifacts."}}}
{"id":"rewire-result-pangolin-maskfalse-average-precision","kind":"result","name":"Pangolin · mask=False · average_precision","description":"","status":"reproduced","facets":{"areas":["dna-genomes"]},"source_ids":["rewire-mfass-v2-source"],"links":[{"relation":"evaluation","target_id":"rewire-evaluation-pangolin-maskfalse"}],"attributes":{"printed_value":"0.3887617543064248","numeric_value":"0.3887617543064248","metric":"average_precision","metric_direction":"higher","unit":"fraction","uncertainty":null,"source_locator":"benchmarks/mfass/results/pangolin-maskFalse.json :: average_precision","review":{"method":"existing_run_import","reviewer":"rewire documented MFASS v2 run","reviewed_at":"2026-09-16T10:50:02Z","notes":"Imported existing documented own-run record, not a newly executed reproduction."},"coverage":"8301/8324","missing_metadata":{"uncertainty":"Point estimate; see paired-comparison artifacts."}}}
{"id":"rewire-result-pangolin-maskfalse-precision-at-100","kind":"result","name":"Pangolin · mask=False · precision_at_100","description":"","status":"reproduced","facets":{"areas":["dna-genomes"]},"source_ids":["rewire-mfass-v2-source"],"links":[{"relation":"evaluation","target_id":"rewire-evaluation-pangolin-maskfalse"}],"attributes":{"printed_value":"0.65","numeric_value":"0.65","metric":"precision_at_100","metric_direction":"higher","unit":"fraction","uncertainty":null,"source_locator":"benchmarks/mfass/results/pangolin-maskFalse.json :: precision_at_100","review":{"method":"existing_run_import","reviewer":"rewire documented MFASS v2 run","reviewed_at":"2026-09-16T10:50:02Z","notes":"Imported existing documented own-run record, not a newly executed reproduction."},"coverage":"8301/8324","missing_metadata":{"uncertainty":"Point estimate; see paired-comparison artifacts."}}}
{"id":"rewire-result-spliceai-1-3-1-auroc","kind":"result","name":"SpliceAI 1.3.1 · auroc","description":"","status":"reproduced","facets":{"areas":["dna-genomes"]},"source_ids":["rewire-mfass-v2-source"],"links":[{"relation":"evaluation","target_id":"rewire-evaluation-spliceai-1-3-1"}],"attributes":{"printed_value":"0.8055241740253153","numeric_value":"0.8055241740253153","metric":"auroc","metric_direction":"higher","unit":"fraction","uncertainty":null,"source_locator":"benchmarks/mfass/results/spliceai-1.3.1.json :: auroc","review":{"method":"existing_run_import","reviewer":"rewire documented MFASS v2 run","reviewed_at":"2026-09-16T10:50:02Z","notes":"Imported existing documented own-run record, not a newly executed reproduction."},"coverage":"8194/8324","missing_metadata":{"uncertainty":"Point estimate; see paired-comparison artifacts."}}}
{"id":"rewire-result-spliceai-1-3-1-average-precision","kind":"result","name":"SpliceAI 1.3.1 · average_precision","description":"","status":"reproduced","facets":{"areas":["dna-genomes"]},"source_ids":["rewire-mfass-v2-source"],"links":[{"relation":"evaluation","target_id":"rewire-evaluation-spliceai-1-3-1"}],"attributes":{"printed_value":"0.2986855472760137","numeric_value":"0.2986855472760137","metric":"average_precision","metric_direction":"higher","unit":"fraction","uncertainty":null,"source_locator":"benchmarks/mfass/results/spliceai-1.3.1.json :: average_precision","review":{"method":"existing_run_import","reviewer":"rewire documented MFASS v2 run","reviewed_at":"2026-09-16T10:50:02Z","notes":"Imported existing documented own-run record, not a newly executed reproduction."},"coverage":"8194/8324","missing_metadata":{"uncertainty":"Point estimate; see paired-comparison artifacts."}}}
{"id":"rewire-result-spliceai-1-3-1-precision-at-100","kind":"result","name":"SpliceAI 1.3.1 · precision_at_100","description":"","status":"reproduced","facets":{"areas":["dna-genomes"]},"source_ids":["rewire-mfass-v2-source"],"links":[{"relation":"evaluation","target_id":"rewire-evaluation-spliceai-1-3-1"}],"attributes":{"printed_value":"0.64","numeric_value":"0.64","metric":"precision_at_100","metric_direction":"higher","unit":"fraction","uncertainty":null,"source_locator":"benchmarks/mfass/results/spliceai-1.3.1.json :: precision_at_100","review":{"method":"existing_run_import","reviewer":"rewire documented MFASS v2 run","reviewed_at":"2026-09-16T10:50:02Z","notes":"Imported existing documented own-run record, not a newly executed reproduction."},"coverage":"8194/8324","missing_metadata":{"uncertainty":"Point estimate; see paired-comparison artifacts."}}}
{"id":"rlsite-rna-binding-2025","kind":"source","name":"RNA language model and graph attention network for RNA and small molecule binding sites prediction","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["rna-transcriptomes"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12417085/","version":"journal full text in PMC","retrieved_at":"2026-09-15T23:33:26Z","doi":"10.1093/bioinformatics/btaf447","publication_status":"peer_reviewed","year":2025,"artifact_sha256":"a50f344e253162ae43f51d7120cfb35a1d0f6114fd8176d760aceb6d05fd95bd","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12417085/fullTextXML","artifact_retrieved_at":"2026-09-16T10:38:57.558222+00:00","legacy_paper":{"id":"rlsite-rna-binding-2025","title":"RNA language model and graph attention network for RNA and small molecule binding sites prediction","year":2025,"publication_status":"peer_reviewed","version":"journal full text in PMC","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12417085/","primary_domain":"rna-transcriptomes","retrieved_utc":"2026-09-15T23:33:26Z","doi":"10.1093/bioinformatics/btaf447","notes":"Numeric result checked against Table 1. in primary full-text XML; journal/source: Bioinformatics."},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"rnaret-2026","kind":"source","name":"Retentive Network promotes efficient RNA language modeling of long sequences","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["rna-transcriptomes"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC13111708/","version":"journal full text in PMC","retrieved_at":"2026-09-15T23:33:26Z","doi":"10.1038/s42003-026-09757-x","publication_status":"peer_reviewed","year":2026,"artifact_sha256":"e970e7322e07fb3c9d12efd315691cc5de5575a3f2616f4b788614c8c706dd0b","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC13111708/fullTextXML","artifact_retrieved_at":"2026-09-16T10:38:57.558224+00:00","legacy_paper":{"id":"rnaret-2026","title":"Retentive Network promotes efficient RNA language modeling of long sequences","year":2026,"publication_status":"peer_reviewed","version":"journal full text in PMC","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC13111708/","primary_domain":"rna-transcriptomes","retrieved_utc":"2026-09-15T23:33:26Z","doi":"10.1038/s42003-026-09757-x","notes":"Numeric result checked against Table 1 in primary full-text XML; journal/source: Communications Biology."},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"scalr-2025","kind":"source","name":"scaLR: a low-resource deep neural network-based platform for single cell analysis and biomarker discovery","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["cells-tissues"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12121358/","version":"version of record","retrieved_at":"2026-09-15T23:29:32Z","doi":"10.1093/bib/bbaf243","publication_status":"peer_reviewed","year":2025,"artifact_sha256":"829afab6a4e30997c608745d3eb280105b8c5c4e601020ffdc55c866144527ca","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12121358/fullTextXML","artifact_retrieved_at":"2026-09-16T10:44:03.403844+00:00","legacy_paper":{"id":"scalr-2025","title":"scaLR: a low-resource deep neural network-based platform for single cell analysis and biomarker discovery","year":2025,"publication_status":"peer_reviewed","version":"version of record","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12121358/","primary_domain":"cells-tissues","retrieved_utc":"2026-09-15T23:29:32Z","notes":"Primary full text verified using Europe PMC XML; venue: Briefings in Bioinformatics; PMC ID: PMC12121358.","doi":"10.1093/bib/bbaf243"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"scatac-llmda-2026","kind":"source","name":"Cell type annotation for scATAC-seq via DNA large language model and graph domain adaptation","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["cells-tissues"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC13132462/","version":"version of record","retrieved_at":"2026-09-15T23:29:32Z","doi":"10.1371/journal.pcbi.1014226","publication_status":"peer_reviewed","year":2026,"artifact_sha256":"f1cdc7d54c6b2d491e4a74a44a1188a3679c555262c988e95a1c0de4614fe3cb","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC13132462/fullTextXML","artifact_retrieved_at":"2026-09-16T10:44:03.395850+00:00","legacy_paper":{"id":"scatac-llmda-2026","title":"Cell type annotation for scATAC-seq via DNA large language model and graph domain adaptation","year":2026,"publication_status":"peer_reviewed","version":"version of record","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC13132462/","primary_domain":"cells-tissues","retrieved_utc":"2026-09-15T23:29:32Z","notes":"Primary full text verified using Europe PMC XML; venue: PLOS Computational Biology; PMC ID: PMC13132462.","doi":"10.1371/journal.pcbi.1014226"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"scelmo-2025","kind":"source","name":"scELMo: Embeddings from Language Models are Good Learners for Single-cell Data Analysis","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["cells-tissues"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12393277/","version":"preprint archived 2025-08-23","retrieved_at":"2026-09-15T23:25:00Z","doi":"10.1101/2023.12.07.569910","publication_status":"preprint","year":2025,"artifact_sha256":"ef75f0d63a567f5e9d7132fd847f44838a82a9741ae55323437e1d1812d86316","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12393277/fullTextXML","artifact_retrieved_at":"2026-09-16T10:41:16.537541+00:00","legacy_paper":{"id":"scelmo-2025","title":"scELMo: Embeddings from Language Models are Good Learners for Single-cell Data Analysis","year":2025,"publication_status":"preprint","version":"preprint archived 2025-08-23","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12393277/","primary_domain":"cells-tissues","retrieved_utc":"2026-09-15T23:25:00Z","notes":"Primary full text via Europe PMC XML; venue: bioRxiv; PMC ID: PMC12393277.","doi":"10.1101/2023.12.07.569910"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"scregnet-2025","kind":"source","name":"Prediction of Gene Regulatory Connections with Joint Single-Cell Foundation Models and Graph-Based Learning","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["cells-tissues"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11838224/","version":"PMC11838224.2","retrieved_at":"2026-09-15T23:25:00Z","doi":"10.1101/2024.12.16.628715","publication_status":"preprint","year":2025,"artifact_sha256":"65b3272d47bb9c4ee1e7a965169bef63add9dbeb31508d4076e5145b761af4ec","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11838224/fullTextXML","artifact_retrieved_at":"2026-09-16T10:41:16.541287+00:00","legacy_paper":{"id":"scregnet-2025","title":"Prediction of Gene Regulatory Connections with Joint Single-Cell Foundation Models and Graph-Based Learning","year":2025,"publication_status":"preprint","version":"PMC11838224.2","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11838224/","primary_domain":"cells-tissues","retrieved_utc":"2026-09-15T23:25:00Z","notes":"Primary full text via Europe PMC XML; venue: bioRxiv; PMC ID: PMC11838224.","doi":"10.1101/2024.12.16.628715"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"scxdr-2026","kind":"source","name":"scXDR: drug response prediction across single-cell datasets via heterogeneous network transfer learning","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["cells-tissues"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12859067/","version":"PMC archival version PMC12859067.1","retrieved_at":"2026-09-15T23:29:32Z","doi":"10.1038/s42003-025-09418-5","publication_status":"peer_reviewed","year":2026,"artifact_sha256":"47b5925e9887d87fc8288d29288802b1d67d54f064d913151df92171f7c68d33","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12859067/fullTextXML","artifact_retrieved_at":"2026-09-16T10:33:50.056Z","legacy_paper":{"id":"scxdr-2026","title":"scXDR: drug response prediction across single-cell datasets via heterogeneous network transfer learning","year":2026,"publication_status":"peer_reviewed","version":"PMC archival version PMC12859067.1","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12859067/","primary_domain":"cells-tissues","retrieved_utc":"2026-09-15T23:29:32Z","notes":"Primary full text verified using Europe PMC XML; venue: Communications Biology; PMC ID: PMC12859067.","doi":"10.1038/s42003-025-09418-5"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"single-cell-aging-probes-2026","kind":"source","name":"Inflammation-linked aging signals in frozen single-cell foundation models: donor-aware detection and robustness testing","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["cells-tissues"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC13407579/","version":"PMC archival version PMC13407579.1","retrieved_at":"2026-09-15T23:29:32Z","doi":"10.1007/s10522-026-10471-8","publication_status":"peer_reviewed","year":2026,"artifact_sha256":"d6cfb13933ceefed630954f804e7dc979b747bc4aec4c8c5518232f25772736a","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC13407579/fullTextXML","artifact_retrieved_at":"2026-09-16T10:33:44.421Z","legacy_paper":{"id":"single-cell-aging-probes-2026","title":"Inflammation-linked aging signals in frozen single-cell foundation models: donor-aware detection and robustness testing","year":2026,"publication_status":"peer_reviewed","version":"PMC archival version PMC13407579.1","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC13407579/","primary_domain":"cells-tissues","retrieved_utc":"2026-09-15T23:29:32Z","notes":"Primary full text verified using Europe PMC XML; venue: Biogerontology; PMC ID: PMC13407579.","doi":"10.1007/s10522-026-10471-8"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"single-cell-peft-2024","kind":"source","name":"Parameter-Efficient Fine-Tuning Enhances Adaptation of Single Cell Large Language Model for Cell Type Identification","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["cells-tissues"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC10862733/","version":"preprint archived 2024-01-30","retrieved_at":"2026-09-15T23:25:00Z","doi":"10.1101/2024.01.27.577455","publication_status":"preprint","year":2024,"artifact_sha256":"77a4a859010259eadf2187465db6ab385efa4927a5eadb95c1e01991044c283f","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC10862733/fullTextXML","artifact_retrieved_at":"2026-09-16T10:41:16.530269+00:00","legacy_paper":{"id":"single-cell-peft-2024","title":"Parameter-Efficient Fine-Tuning Enhances Adaptation of Single Cell Large Language Model for Cell Type Identification","year":2024,"publication_status":"preprint","version":"preprint archived 2024-01-30","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC10862733/","primary_domain":"cells-tissues","retrieved_utc":"2026-09-15T23:25:00Z","notes":"Primary full text via Europe PMC XML; venue: bioRxiv; PMC ID: PMC10862733.","doi":"10.1101/2024.01.27.577455"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"single-cell-residual-geometry-2026","kind":"source","name":"Residual-stream geometry of single-cell foundation models carries incremental gene-regulatory signal across tissues","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["cells-tissues"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC13418759/","version":"version of record","retrieved_at":"2026-09-15T23:37:05Z","doi":"10.1186/s12859-026-06538-5","publication_status":"peer_reviewed","year":2026,"artifact_sha256":"77546faec51cfb5c78b73c5943f940c4e0097d130b1ddde4ab8287b507b36df6","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC13418759/fullTextXML","artifact_retrieved_at":"2026-09-16T10:41:06Z","legacy_paper":{"id":"single-cell-residual-geometry-2026","title":"Residual-stream geometry of single-cell foundation models carries incremental gene-regulatory signal across tissues","year":2026,"publication_status":"peer_reviewed","version":"version of record","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC13418759/","primary_domain":"cells-tissues","retrieved_utc":"2026-09-15T23:37:05Z","notes":"Primary full text verified via Europe PMC fullTextXML; venue: BMC Bioinformatics; PMC ID: PMC13418759. Score reflects scGPT plus paper geometry features, not raw scGPT.","doi":"10.1186/s12859-026-06538-5"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"spin-protein-function-2026","kind":"source","name":"Scaling the profile of life by function with SPIN","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["proteins-complexes"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12970593/","version":"journal full text in PMC","retrieved_at":"2026-09-15T23:33:26Z","doi":"10.1093/bioadv/vbag064","publication_status":"peer_reviewed","year":2026,"artifact_sha256":"9701843e93bf7fa3ead71e19693fb07d483f1022379871adfb04486783722a9d","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12970593/fullTextXML","artifact_retrieved_at":"2026-09-16T10:38:57.558225+00:00","legacy_paper":{"id":"spin-protein-function-2026","title":"Scaling the profile of life by function with SPIN","year":2026,"publication_status":"peer_reviewed","version":"journal full text in PMC","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12970593/","primary_domain":"proteins-complexes","retrieved_utc":"2026-09-15T23:33:26Z","doi":"10.1093/bioadv/vbag064","notes":"Numeric result checked against Table 1 in primary full-text XML; journal/source: Bioinformatics Advances."},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"attributes":{"artifact_sha256":"045aa63715acf615327d45ff42a897e77fa6d4b88084d2ad4e2ea836eb5fb48a","licence":null,"locator":"README.md","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:31:32.390263+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/aertslab/GENIE3/blob/54bc15636322e8773357de6e0b6683c6bc802825/README.md","version":"54bc15636322e8773357de6e0b6683c6bc802825"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-aertslab-genie3","kind":"source","links":[],"name":"aertslab/GENIE3 official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"f2befd99acf59576a22b8a44abd2345e8ed7304cf470609f27d311e08ed3f066","licence":null,"locator":"README.md","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:30:28.363182+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/aertslab/GRNBoost/blob/26c836b3dcbb85852d3c6f4b8340e8655434da02/README.md","version":"26c836b3dcbb85852d3c6f4b8340e8655434da02"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-aertslab-grnboost","kind":"source","links":[],"name":"aertslab/GRNBoost official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"4aa29fa1d4333a72013fd2c60545218f0015b85b95b76878cbdb88962ff4e55d","licence":null,"locator":"README.md","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:30:22.733246+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/altoslabs/perturbench/blob/c84038bc1ea409aa54f3832cfa6f34f5059adf0c/README.md","version":"c84038bc1ea409aa54f3832cfa6f34f5059adf0c"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-altoslabs-perturbench","kind":"source","links":[],"name":"altoslabs/perturbench official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"50a634fe3236b6272b70928ac41bdebd4a230467b151f59597742ccd56ac8909","licence":null,"locator":"README.md","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:31:33.087504+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/aqlaboratory/genie3/blob/d77ae5ac04212ff1e8b29b585859a3244c614804/README.md","version":"d77ae5ac04212ff1e8b29b585859a3244c614804"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-aqlaboratory-genie3","kind":"source","links":[],"name":"aqlaboratory/genie3 official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"985b6dc09144ea14378dd2c543b288e8d2a05cb342a77cbc56b7a39c5170f638","licence":null,"locator":"README.md","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:30:25.490370+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/aqlaboratory/openfold/blob/be2ec1841f16c966c65ae0e7599ebbadc725757d/README.md","version":"be2ec1841f16c966c65ae0e7599ebbadc725757d"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-aqlaboratory-openfold","kind":"source","links":[],"name":"aqlaboratory/openfold official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"58787c8ef5cb4fba4c04322a4ceb9f174e2233ec22d4193622fb6bc67d651d89","licence":null,"locator":"README.md","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:30:24.685703+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/ArcInstitute/evo2/blob/53f195997257c56c00e5ef8d33a54f5baad143a6/README.md","version":"53f195997257c56c00e5ef8d33a54f5baad143a6"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-arcinstitute-evo2","kind":"source","links":[],"name":"ArcInstitute/evo2 official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"568c0b4f9d93374f7ebc7467c226fdde0050abfd60147bce0b316163bf4199d6","licence":null,"locator":"README.md","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:31:32.302509+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/ArcInstitute/state/blob/9bbfe78a434a55205e4de834e1ea99f85f7a3add/README.md","version":"9bbfe78a434a55205e4de834e1ea99f85f7a3add"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-arcinstitute-state","kind":"source","links":[],"name":"ArcInstitute/state official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"5a089ca429ed2a314e257fddacd70251c863e62a3547b793c38f56785861cac1","licence":null,"locator":"README.md","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:30:24.477301+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/Benchmarking-Initiative/Benchmark-Models-PEtab/blob/ddaa86d13f708926c57ec8918ce75a6b50e2e562/README.md","version":"ddaa86d13f708926c57ec8918ce75a6b50e2e562"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-benchmarking-initiative-benchmark-models-petab","kind":"source","links":[],"name":"Benchmarking-Initiative/Benchmark-Models-PEtab official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"96260519b594ae22cba9f28d1f64622de001f2abf11d406c9da572bfaf145727","licence":null,"locator":"readme.md","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:30:28.315898+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/biobakery/humann/blob/e07b3a34d0b94c09a8ac5d28ff95009611178be2/readme.md","version":"e07b3a34d0b94c09a8ac5d28ff95009611178be2"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-biobakery-humann","kind":"source","links":[],"name":"biobakery/humann official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"ce491bb2d686145e0773c685d0d02e8a5fabc7daaea60eef07cf54298561fba7","licence":null,"locator":"README.md","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:30:27.980580+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/biobakery/MetaPhlAn/blob/424f3e6e30618266404353e1083c6405a9f02f48/README.md","version":"424f3e6e30618266404353e1083c6405a9f02f48"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-biobakery-metaphlan","kind":"source","links":[],"name":"biobakery/MetaPhlAn official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"2736a9546e94b367e1bb3d1052e22460bb2188229d432b71eb9b01fe6b2a9b1a","licence":null,"locator":"readme.md","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:30:21.987781+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/biomap-research/PFMBench/blob/53758ffcbdf1d79b5d125383e4dd52d6fd59d2a1/readme.md","version":"53758ffcbdf1d79b5d125383e4dd52d6fd59d2a1"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-biomap-research-pfmbench","kind":"source","links":[],"name":"biomap-research/PFMBench official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"4d907a6723f3f56b14b13e35eeb0833ddf9a2f3512a1fafd07d79259105b264c","licence":null,"locator":"README.md","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:30:27.678733+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/BojarLab/glycowork/blob/3d63f1ec25c850da3cde4d25cb602d50b6b5732b/README.md","version":"3d63f1ec25c850da3cde4d25cb602d50b6b5732b"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-bojarlab-glycowork","kind":"source","links":[],"name":"BojarLab/glycowork official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"b0503e8ca789f19f1fc2350c5aaf57b1b323bbae43b354655231b5f4a1586c83","licence":null,"locator":"README.md","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:30:26.283123+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/bowang-lab/scGPT/blob/cebd6fae655b9c585a4807daa3ac31bb764f06b4/README.md","version":"cebd6fae655b9c585a4807daa3ac31bb764f06b4"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-bowang-lab-scgpt","kind":"source","links":[],"name":"bowang-lab/scGPT official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"1fbb077b23329940ae6ee1dbcc578831765cc6a3c21e61c830d189a1b4be1fa3","licence":null,"locator":"HTML main page","missing_metadata":{"immutable_source_version":"unavailable","licence":"unextracted","version":"unreported"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:31:55.874128+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","scope_note":"Mutable official page: content hash and retrieval time recorded, but the original page is not redistributed.","url":"https://biofunctionprediction.org/cafa/","version":null},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-cafa","kind":"source","links":[],"name":"cafa official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"b77d93b4ff352874ecd21e074357b0bb7f4f09e7bfd53d8f7f152d05063c30c1","licence":null,"locator":"README.md","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:30:27.117271+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/calico/basenji/blob/06ce5d387e20b47184d05433b3983163c5f923cd/README.md","version":"06ce5d387e20b47184d05433b3983163c5f923cd"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-calico-basenji","kind":"source","links":[],"name":"calico/basenji official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"17825bf33280f40b596a104c547b57fae5ee5c07f8d60b396d0f4780d47ef9a5","licence":null,"locator":"HTML main page","missing_metadata":{"immutable_source_version":"unavailable","licence":"unextracted","version":"unreported"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:31:54.339676+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","scope_note":"Mutable official page: content hash and retrieval time recorded, but the original page is not redistributed.","url":"https://cami-challenge.org/","version":null},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-cami","kind":"source","links":[],"name":"cami official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"5908389c2f1f5797ffb73d45776684ff0b34c0fc61a69a2b5b9dbd88df58dcec","licence":null,"locator":"README.md","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:30:23.152881+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/CAMI-challenge/AMBER/blob/f8b3a601043d13fc4227d5691c13561eb4490e50/README.md","version":"f8b3a601043d13fc4227d5691c13561eb4490e50"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-cami-challenge-amber","kind":"source","links":[],"name":"CAMI-challenge/AMBER official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"ec5309bf16daab6c9a0adb393b191a3e716b9c627834cf3e08655c250a63da61","licence":null,"locator":"README.md","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:30:22.965882+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/CAMI-challenge/CAMISIM/blob/7ce6013c6d5a0fac8ba8a80e52a03560d3546fa6/README.md","version":"7ce6013c6d5a0fac8ba8a80e52a03560d3546fa6"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-cami-challenge-camisim","kind":"source","links":[],"name":"CAMI-challenge/CAMISIM official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"d123b202234d3ed520911982dcf88e2fcfd888c83e65378effb2d2e269d0abb1","licence":null,"locator":"README.md","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:30:23.429372+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/CAMI-challenge/OPAL/blob/98120c326eef08e391899e4bd3a362e0e6558b4a/README.md","version":"98120c326eef08e391899e4bd3a362e0e6558b4a"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-cami-challenge-opal","kind":"source","links":[],"name":"CAMI-challenge/OPAL official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"5e22606895cf0c30565ed4456bc680ca97c10c8c5d38054f82a6e4677c8d8c24","licence":null,"locator":"HTML main page","missing_metadata":{"immutable_source_version":"unavailable","licence":"unextracted","version":"unreported"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:31:54.812767+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","scope_note":"Mutable official page: content hash and retrieval time recorded, but the original page is not redistributed.","url":"https://www.capri-docking.org/","version":null},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-capri","kind":"source","links":[],"name":"capri official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"de391b68636462ddb78d0659d8784128909d23e9c015832cca94d883f404a3f7","licence":null,"locator":"HTML main page","missing_metadata":{"immutable_source_version":"unavailable","licence":"unextracted","version":"unreported"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:31:54.981498+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","scope_note":"Mutable official page: content hash and retrieval time recorded, but the original page is not redistributed.","url":"https://predictioncenter.org/","version":null},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-casp","kind":"source","links":[],"name":"casp official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"4f1728521ab79de1c33e1cf8605b31037effed5de2a2fbbccba58d7b0a005ae7","licence":null,"locator":"README.md","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:30:28.428792+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/ccsb-scripps/AutoDock-Vina/blob/3c65c0b3e6c2c1d183f6a175ecb65e3c5ba91645/README.md","version":"3c65c0b3e6c2c1d183f6a175ecb65e3c5ba91645"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-ccsb-scripps-autodock-vina","kind":"source","links":[],"name":"ccsb-scripps/AutoDock-Vina official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"0f2074b432ba2516d0ea05fafab833e953201a3fbe0129c5b6beed9f5f089e81","licence":null,"locator":"README.md","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:30:21.083930+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/darlednik/GENEB/blob/9642d481e40c0af23995dcd162b779613f789f97/README.md","version":"9642d481e40c0af23995dcd162b779613f789f97"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-darlednik-geneb","kind":"source","links":[],"name":"darlednik/GENEB official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"772ebe52d2ba5100a28a888910c6f0c9fd4ded1d1372e3d89f6f1c48707e0365","licence":null,"locator":"README.md","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:30:25.524269+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/dauparas/ProteinMPNN/blob/8907e6671bfbfc92303b5f79c4b5e6ce47cdef57/README.md","version":"8907e6671bfbfc92303b5f79c4b5e6ce47cdef57"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-dauparas-proteinmpnn","kind":"source","links":[],"name":"dauparas/ProteinMPNN official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"2ea33af266b4268a55fd750d0f3265cd3165d61f6be375c5ea3b3ff5c58c7c8c","licence":null,"locator":"README.md","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:30:27.907349+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/DerrickWood/kraken2/blob/8c190b1b668825935dbf6dee5f969227dc8269bb/README.md","version":"8c190b1b668825935dbf6dee5f969227dc8269bb"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-derrickwood-kraken2","kind":"source","links":[],"name":"DerrickWood/kraken2 official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"6e404699412cb687bd737c4423c568f6ece1bad9f62a13e15ac1785062ccccd1","licence":null,"locator":"README.md","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:31:32.435757+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/drorlab/atom3d/blob/4c2f3b7e9efe128791b83f03b2e8cae91e78b018/README.md","version":"4c2f3b7e9efe128791b83f03b2e8cae91e78b018"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-drorlab-atom3d","kind":"source","links":[],"name":"drorlab/atom3d official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"74a897f0e97e3d4256f7cff11424dd264df4daaf8cabd9a6fb74b661421de038","licence":null,"locator":"README.md","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:30:25.632814+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/evolutionaryscale/esm/blob/bf343ba264b650dff7a073643725f9aaa1fdbe8d/README.md","version":"bf343ba264b650dff7a073643725f9aaa1fdbe8d"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-evolutionaryscale-esm","kind":"source","links":[],"name":"evolutionaryscale/esm official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"8b273c21a322fc9473d1b68d0dd40c8166ab2f89e4a190aa26ca87251b97cba9","licence":null,"locator":"README.md","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:30:24.844614+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/facebookresearch/esm/blob/2b369911bb5b4b0dda914521b9475cad1656b2ac/README.md","version":"2b369911bb5b4b0dda914521b9475cad1656b2ac"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-facebookresearch-esm","kind":"source","links":[],"name":"facebookresearch/esm official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"cd991ae6e76a5e84ea5449f91c4ed86ba4f57942682dc3c50d872c366bbbd4b7","licence":null,"locator":"HTML main page","missing_metadata":{"immutable_source_version":"unavailable","licence":"unextracted","version":"unreported"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:31:54.381289+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","scope_note":"Mutable official page: content hash and retrieval time recorded, but the original page is not redistributed.","url":"https://flip.protein.properties/","version":null},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-flip2","kind":"source","links":[],"name":"flip2 official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"a40358504726ee4e086f0623348fb2206c24ab8166d04a83d944579e9c62bdc8","licence":null,"locator":"README.md","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:30:20.985691+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/frederikkemarin/BEND/blob/ac6e80c75e09d83cf47a7b4bcf0e44599c5706cf/README.md","version":"ac6e80c75e09d83cf47a7b4bcf0e44599c5706cf"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-frederikkemarin-bend","kind":"source","links":[],"name":"frederikkemarin/BEND official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"5237cd1af3f1d7b9cf07cfe3cf722987bb27e10ee356ec108584943446bbb836","licence":null,"locator":"README.md","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:30:24.094166+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/GlycanML/GlycanML/blob/9f392aa6f9c6d74a296a250199beb347923d04e0/README.md","version":"9f392aa6f9c6d74a296a250199beb347923d04e0"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-glycanml-glycanml","kind":"source","links":[],"name":"GlycanML/GlycanML official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"cc16ba436ea8a967c764ac034a655d632b4c1ef8b4065868b8d07914c7cb88be","licence":null,"locator":"README.md","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:30:25.220155+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/google-deepmind/alphafold3/blob/c0f97eda2f1f482fd94d3a38bece18c7069b4a5c/README.md","version":"c0f97eda2f1f482fd94d3a38bece18c7069b4a5c"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-google-deepmind-alphafold3","kind":"source","links":[],"name":"google-deepmind/alphafold3 official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"8e5203afe343100832391e6155c7112f15cfe60bf0c21681d64e3420f854ef4d","licence":null,"locator":"README.md","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:30:24.355879+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/illumina/SpliceAI/blob/03f42437aaf56dc5dfd822c4ccee5aec1a705079/README.md","version":"03f42437aaf56dc5dfd822c4ccee5aec1a705079"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-illumina-spliceai","kind":"source","links":[],"name":"illumina/SpliceAI official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"9f51bbb20c4c5c36e77fb03ca1c5c36236e287c48a1ee31f53150545d421ec25","licence":null,"locator":"README.md","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:30:27.121929+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/instadeepai/nucleotide-transformer/blob/2dc37b86e16a6970fbc731751f7719d9f676f7f9/README.md","version":"2dc37b86e16a6970fbc731751f7719d9f676f7f9"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-instadeepai-nucleotide-transformer","kind":"source","links":[],"name":"instadeepai/nucleotide-transformer official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"f6e3b46a5f6744806846ccb4a054bcf3bec3da3d4249ada42fb9acac2a32024d","licence":null,"locator":"README.md","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:30:21.863141+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/J-SNACKKB/FLIP/blob/62cace8735f5610e2743cf06ce0f944b37fffaa6/README.md","version":"62cace8735f5610e2743cf06ce0f944b37fffaa6"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-j-snackkb-flip","kind":"source","links":[],"name":"J-SNACKKB/FLIP official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"79149435313841e6f9cc0c8b581259f3952a2469cbe6f9a733f7143b557025c2","licence":null,"locator":"README.md","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:30:25.416237+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/jwohlwend/boltz/blob/b1ebfc46ecf57f5414e0d1a6f9027bbb122c53bc/README.md","version":"b1ebfc46ecf57f5414e0d1a6f9027bbb122c53bc"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-jwohlwend-boltz","kind":"source","links":[],"name":"jwohlwend/boltz official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"058c4f98218ea2ed854681126b6f682c9f3beec91275781fb37e39c55d0c92ea","licence":null,"locator":"README.md","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:30:27.172955+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/kundajelab/chrombpnet/blob/09938fdb4397ec0006510e5251e48920a505d4de/README.md","version":"09938fdb4397ec0006510e5251e48920a505d4de"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-kundajelab-chrombpnet","kind":"source","links":[],"name":"kundajelab/chrombpnet official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"d21ee86b56c9b794f5b58f3c39c3e27c51d027a3b280d848457abf53f521f052","licence":null,"locator":"README.md","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:30:21.153480+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/kundajelab/DART-Eval/blob/af2a86d666c35304257c2fa7e15180e1fbcabb01/README.md","version":"af2a86d666c35304257c2fa7e15180e1fbcabb01"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-kundajelab-dart-eval","kind":"source","links":[],"name":"kundajelab/DART-Eval official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"c48be391c26a1ba35004265169c85661c4a96b0089d86383e922e25f6432a6f7","licence":null,"locator":"HTML main page","missing_metadata":{"immutable_source_version":"unavailable","licence":"unextracted","version":"unreported"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:31:55.415067+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","scope_note":"Mutable official page: content hash and retrieval time recorded, but the original page is not redistributed.","url":"https://fiehnlab.ucdavis.edu/projects/LipidBlast/","version":null},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-lipidblast","kind":"source","links":[],"name":"lipidblast official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"cfd3a044e568f9465d4365d8aa5e182041720301d70006e9f8f561608c660e70","licence":null,"locator":"HTML main page","missing_metadata":{"immutable_source_version":"unavailable","licence":"unextracted","version":"unreported"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:31:54.464572+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","scope_note":"Mutable official page: content hash and retrieval time recorded, but the original page is not redistributed.","url":"https://www.lipidmaps.org/resources/tools/lipidfinder/","version":null},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-lipidfinder","kind":"source","links":[],"name":"lipidfinder official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"8bcbe6e0632af4072d5124064e4032f814894185afcd684d85248836349f155c","licence":null,"locator":"HTML main page","missing_metadata":{"immutable_source_version":"unavailable","licence":"unextracted","version":"unreported"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:31:54.217588+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","scope_note":"Mutable official page: content hash and retrieval time recorded, but the original page is not redistributed.","url":"https://www.lipidmaps.org/databases/standardspectraDB/overview","version":null},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-lipidmaps-spectra","kind":"source","links":[],"name":"lipidmaps-spectra official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"71fee661aba53ea2285b776d572982a57c8637275687dfe4c5b7f5113a807d0e","licence":null,"locator":"README.md","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:30:23.520671+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/maabuu/posebusters/blob/6236d07017493531851cce775e8ef834d4763d2f/README.md","version":"6236d07017493531851cce775e8ef834d4763d2f"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-maabuu-posebusters","kind":"source","links":[],"name":"maabuu/posebusters official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"734a8cec5f667d74d421bf3b273ad7e256216109636da45aa7ceba21cd34de16","licence":null,"locator":"README.md","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:30:20.936532+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/MAGICS-LAB/DNABERT_2/blob/f25bed9ee20db966dff39e5c1571249d04e36404/README.md","version":"f25bed9ee20db966dff39e5c1571249d04e36404"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-magics-lab-dnabert-2","kind":"source","links":[],"name":"MAGICS-LAB/DNABERT_2 official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"3d002564045d3c493f30386df7983ee34b0f2f7ac89bf32ac63a5f780395d811","licence":null,"locator":"README.md","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:30:22.733154+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/mahmoodlab/HEST/blob/3ddb5eaf5bd2a8133e0c0e8015816489a3d99dc3/README.md","version":"3ddb5eaf5bd2a8133e0c0e8015816489a3d99dc3"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-mahmoodlab-hest","kind":"source","links":[],"name":"mahmoodlab/HEST official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"06a0c6b7adf444d7af53d20ce94c3cef483bd7be8db02fd8960c0704680d61ee","licence":null,"locator":"README.rst","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:30:29.085499+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/matchms/matchms/blob/066608589587c8d089afd2e8d55ceadb2766ea62/README.rst","version":"066608589587c8d089afd2e8d55ceadb2766ea62"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-matchms-matchms","kind":"source","links":[],"name":"matchms/matchms official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"fa39bffca31211baedb3b63df5dede775fcffff7b411d4261d5b18f526ae1153","licence":null,"locator":"README.md","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:30:27.482216+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/matsui-lab/GlycanGT/blob/96611518c971deb89215ca163deaf9de3a59fa32/README.md","version":"96611518c971deb89215ca163deaf9de3a59fa32"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-matsui-lab-glycangt","kind":"source","links":[],"name":"matsui-lab/GlycanGT official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"c0acb8f22d51b718d5c3a253e81b9882d804fa2aa153b6fcf3e826e22f085dd3","licence":null,"locator":"HTML main page","missing_metadata":{"immutable_source_version":"unavailable","licence":"unextracted","version":"unreported"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:31:56.576460+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","scope_note":"Mutable official page: content hash and retrieval time recorded, but the original page is not redistributed.","url":"https://meme-suite.org/meme/tools/fimo","version":null},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-meme","kind":"source","links":[],"name":"meme official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"e2dacff9ad56bca50c31373a1e87041eef948721b8aa0be001a95185c375c15f","licence":null,"locator":"README.md","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:30:23.885647+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/mims-harvard/TDC/blob/c310c35f27e3f506411018ac43d97b8ba23ca652/README.md","version":"c310c35f27e3f506411018ac43d97b8ba23ca652"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-mims-harvard-tdc","kind":"source","links":[],"name":"mims-harvard/TDC official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"926f0f196439564b095cbabe65f6a0acee3f22afbb0f6c8fe50bb3964082d15b","licence":null,"locator":"README.md","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:30:20.990833+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/ML-Bioinfo-CEITEC/genomic_benchmarks/blob/605d8539830e16c85abe7826990958303ffc5e1c/README.md","version":"605d8539830e16c85abe7826990958303ffc5e1c"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-ml-bioinfo-ceitec-genomic-benchmarks","kind":"source","links":[],"name":"ML-Bioinfo-CEITEC/genomic_benchmarks official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"f9f1c1d62adc471661ca98b30c0250e9f3ce0cff7433830f149f5f48ea41c3da","licence":null,"locator":"README.md","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:30:26.372370+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/ml4bio/RNA-FM/blob/348951516e0963d22bbb33b3c9fc18c89081d38e/README.md","version":"348951516e0963d22bbb33b3c9fc18c89081d38e"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-ml4bio-rna-fm","kind":"source","links":[],"name":"ml4bio/RNA-FM official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"f0c67304e20ced42938829dfee39480cef51ee3a4357ee8b53eafc89c305fa60","licence":null,"locator":"README.md","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:30:21.823561+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/morrislab/mRNABench/blob/74f96b8e6ae9f41cc3cccff089d826a62d5604b8/README.md","version":"74f96b8e6ae9f41cc3cccff089d826a62d5604b8"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-morrislab-mrnabench","kind":"source","links":[],"name":"morrislab/mRNABench official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"e9c0b76d743af53198b0197bfa58305bf26cff822538658d0366880fcf56a8a9","licence":null,"locator":"README.md","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:30:21.898594+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/mrzzmrzz/NABench/blob/99c8681ec1eab706e10ff90a5c329dcf184cc1d1/README.md","version":"99c8681ec1eab706e10ff90a5c329dcf184cc1d1"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-mrzzmrzz-nabench","kind":"source","links":[],"name":"mrzzmrzz/NABench official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"e503f496841fd1d3836ebdab573de645257fc92fe7b5479056592da512e0ef3f","licence":null,"locator":"HTML main page","missing_metadata":{"immutable_source_version":"unavailable","licence":"unextracted","version":"unreported"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:31:54.933176+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","scope_note":"Mutable official page: content hash and retrieval time recorded, but the original page is not redistributed.","url":"https://arxiv.org/abs/2605.19752","version":null},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-msalign","kind":"source","links":[],"name":"msalign official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"b9e620179f6b9a9aafd9eaf8b874b8f1fa2c8c4ffdced39f2e075391ec3d476d","licence":null,"locator":"README.md","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:30:24.835167+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/Murali-group/Beeline/blob/37464085eb8a95d6cc6a3d3a3c649d36db6052ed/README.md","version":"37464085eb8a95d6cc6a3d3a3c649d36db6052ed"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-murali-group-beeline","kind":"source","links":[],"name":"Murali-group/Beeline official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"321487a8de52c6cfa647a658f61150dd72e0acb0125470524fb94d1f8b23321a","licence":null,"locator":"README.md","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:30:21.811694+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/OATML-Markslab/ProteinGym/blob/144fe22b07dfaeec2b366f2346203a9838a55b4c/README.md","version":"144fe22b07dfaeec2b366f2346203a9838a55b4c"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-oatml-markslab-proteingym","kind":"source","links":[],"name":"OATML-Markslab/ProteinGym official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"ae2da4ad3ae1d822e0010b6c023b26ef1d02b97007e17611dd4b1e22c70d4998","licence":null,"locator":"README.rst","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:30:28.394293+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/opencobra/cobrapy/blob/5aa19300fbf5dd632a9ac5c39ca28c1c621b943d/README.rst","version":"5aa19300fbf5dd632a9ac5c39ca28c1c621b943d"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-opencobra-cobrapy","kind":"source","links":[],"name":"opencobra/cobrapy official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"ee26d791b60868c3e7701357831a49d8325f8d4d7cb0769b5177e5a67713114b","licence":null,"locator":"README.md","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:30:22.429615+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/openproblems-bio/openproblems/blob/0ca5d0cd040b741c1b6cc2e4cad7230cb2c50131/README.md","version":"0ca5d0cd040b741c1b6cc2e4cad7230cb2c50131"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-openproblems-bio-openproblems","kind":"source","links":[],"name":"openproblems-bio/openproblems official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"1d53c3b89030dc4651d3e7bf4749256b7e579330fc7a660cffaa992d646da34a","licence":null,"locator":"README.md","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:30:23.659162+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/plinder-org/plinder/blob/85b3f1cb1763530a6cfd934f4263a1777c41afa4/README.md","version":"85b3f1cb1763530a6cfd934f4263a1777c41afa4"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-plinder-org-plinder","kind":"source","links":[],"name":"plinder-org/plinder official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"a6afe934f9894cc71fb9b561c7a122d043b377c27b3fc6f18505fe65b6226256","licence":null,"locator":"README.md","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:30:27.232940+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/pluskal-lab/DreaMS/blob/dbec3a0b514a99e5056cfccde4559fda8cfe8129/README.md","version":"dbec3a0b514a99e5056cfccde4559fda8cfe8129"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-pluskal-lab-dreams","kind":"source","links":[],"name":"pluskal-lab/DreaMS official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"08bf3607e6e2e5462b81eac85d0e71d9d23ce1c9bf1a370c9d1079ecd60ee2d8","licence":null,"locator":"README.md","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:30:23.954980+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/pluskal-lab/MassSpecGym/blob/f259fe3780d5bd227fc6ece36ce6f397c2eef716/README.md","version":"f259fe3780d5bd227fc6ece36ce6f397c2eef716"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-pluskal-lab-massspecgym","kind":"source","links":[],"name":"pluskal-lab/MassSpecGym official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"2e488850a6557bb57407615f2df9194351718b3dc0298a03c0c97d8e93460712","licence":null,"locator":"HTML main page","missing_metadata":{"immutable_source_version":"unavailable","licence":"unextracted","version":"unreported"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:31:54.349409+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","scope_note":"Mutable official page: content hash and retrieval time recorded, but the original page is not redistributed.","url":"https://proteinbench.github.io/","version":null},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-proteinbench","kind":"source","links":[],"name":"proteinbench official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"d89eb9790ce805ce23e1b1a6804f4d67c81056025e4852b875f8ddfd87a08125","licence":null,"locator":"README.md","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:30:25.975464+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/RosettaCommons/RFdiffusion/blob/86507b6538f51fce57b5a72477165f03999ed7ae/README.md","version":"86507b6538f51fce57b5a72477165f03999ed7ae"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-rosettacommons-rfdiffusion","kind":"source","links":[],"name":"RosettaCommons/RFdiffusion official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"eb46b8a54e60643ca0cd8cb375ede05b01dcbd2380ca17ce8027a92ba13cebbb","licence":null,"locator":"README.md","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:30:26.353619+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/scverse/scvi-tools/blob/73b28e44223621470e582a81a102c107bb22678b/README.md","version":"73b28e44223621470e582a81a102c107bb22678b"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-scverse-scvi-tools","kind":"source","links":[],"name":"scverse/scvi-tools official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"fe78b7b1ede50c673b8f0d91fb2637707685c78b269ee157edb6cbc4d510d61b","licence":null,"locator":"README.md","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:30:26.451955+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/snap-stanford/GEARS/blob/f374e43e197b295016d80395d7a54ddb81cc6769/README.md","version":"f374e43e197b295016d80395d7a54ddb81cc6769"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-snap-stanford-gears","kind":"source","links":[],"name":"snap-stanford/GEARS official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"2f8690d6a4a9767973ba9ea2af019d43d6108555ef0e9a69786033a71547d133","licence":null,"locator":"README.md","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:30:28.898945+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/soedinglab/hh-suite/blob/43095e46ada4ec2a8a47d47ef5ad7e38b1429f7b/README.md","version":"43095e46ada4ec2a8a47d47ef5ad7e38b1429f7b"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-soedinglab-hh-suite","kind":"source","links":[],"name":"soedinglab/hh-suite official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"b6c591a763bf99c027857385f0e87ce5aa96caeaa74d71afd1fcec449eadb3d7","licence":null,"locator":"README.md","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:30:28.691617+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/soedinglab/MMseqs2/blob/d401e78c2d18a822cdb1527d7464a043f6035a15/README.md","version":"d401e78c2d18a822cdb1527d7464a043f6035a15"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-soedinglab-mmseqs2","kind":"source","links":[],"name":"soedinglab/MMseqs2 official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"a384ec62c86d568aeb0fe3e8a3b111f071fbcfacb937b407eb6da24fc70c823f","licence":null,"locator":"README.md","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:30:25.578919+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/sokrypton/ColabFold/blob/84c27d9cc500489fd9b97545d2325b9d00f251d5/README.md","version":"84c27d9cc500489fd9b97545d2325b9d00f251d5"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-sokrypton-colabfold","kind":"source","links":[],"name":"sokrypton/ColabFold official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"b28c74fe3cd6b69a8ba6d84891d0539e54dfef882abd5ed4d11ed0b029bb477a","licence":null,"locator":"README.md","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:31:32.443187+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/songlab-cal/tape/blob/6d345c2b2bbf52cd32cf179325c222afd92aec7e/README.md","version":"6d345c2b2bbf52cd32cf179325c222afd92aec7e"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-songlab-cal-tape","kind":"source","links":[],"name":"songlab-cal/tape official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"0403f84453aace301c7d02895d94a977ccbbec77c3a94a49a23d0d529dd48d31","licence":null,"locator":"README.md","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:30:21.025462+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/terry-r123/RNABenchmark/blob/da7f9c7ac3f39605af27e1dfcdf879adba963d79/README.md","version":"da7f9c7ac3f39605af27e1dfcdf879adba963d79"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-terry-r123-rnabenchmark","kind":"source","links":[],"name":"terry-r123/RNABenchmark official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"db7aa3a701778d541bf47d8214b50e1ce9f73e92dc3e810bfd740270e9e353aa","licence":null,"locator":"README.md","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:31:32.426513+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/theislab/scib/blob/cd67913396b4c0430710b3d90f1d1841f5fa4468/README.md","version":"cd67913396b4c0430710b3d90f1d1841f5fa4468"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-theislab-scib","kind":"source","links":[],"name":"theislab/scib official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"9117f9d255d6b6e810d224a600d381192bccccd3f0cd417fed9559d49ca8fffd","licence":null,"locator":"README.md","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:30:24.697521+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/tkzeng/Pangolin/blob/5cf94b8db938c658391b4305cd7ce33297d44ff7/README.md","version":"5cf94b8db938c658391b4305cd7ce33297d44ff7"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-tkzeng-pangolin","kind":"source","links":[],"name":"tkzeng/Pangolin official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"57dd9cbd5e6b62ed665eebd909d82ffec580a55c3bd2f48a928f807a8c983b1e","licence":null,"locator":"HTML main page","missing_metadata":{"immutable_source_version":"unavailable","licence":"unextracted","version":"unreported"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:31:54.531595+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","scope_note":"Mutable official page: content hash and retrieval time recorded, but the original page is not redistributed.","url":"https://arcinstitute.org/news/virtual-cell-challenge-2026","version":null},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-vcc2026","kind":"source","links":[],"name":"vcc2026 official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"d37146b01e4273062a5230c496a9af8414ee7ef14fcd905cf415959256d59c4e","licence":null,"locator":"README.md","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:30:26.326355+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/ViennaRNA/ViennaRNA/blob/1ffec79f5e258896160f7362ced8263450f371dc/README.md","version":"1ffec79f5e258896160f7362ced8263450f371dc"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-viennarna-viennarna","kind":"source","links":[],"name":"ViennaRNA/ViennaRNA official source","source_ids":[],"status":"source_checked"}
{"attributes":{"artifact_sha256":"d8522e585806ec0008a36558e0dd1f6f0b0bb4deffb3f64a7ee4f09cd087e9f4","licence":null,"locator":"README.md","missing_metadata":{"licence":"unextracted"},"publication_status":"official_project_source","retrieved_at":"2026-09-16T10:30:23.079976+00:00","review_scope":"URL and retrieved artifact identity checked; individual scientific claims need their own review.","url":"https://github.com/Virtual-Cell-Research-Community/scPertEval/blob/4685f11927e887745737600170da7a655b727553/README.md","version":"4685f11927e887745737600170da7a655b727553"},"description":"Primary project documentation or project-maintained evidence.","facets":{},"id":"src-discovery-virtual-cell-research-community-scperteval","kind":"source","links":[],"name":"Virtual-Cell-Research-Community/scPertEval official source","source_ids":[],"status":"source_checked"}
{"id":"structure-informed-plm-2025","kind":"source","name":"Structure-Informed Protein Language Models are Robust Predictors for Variant Effects","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["proteins-complexes"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12068927/","version":"Human Genetics 2025 journal article (online 2024)","retrieved_at":"2026-09-15T23:33:26Z","doi":"10.1007/s00439-024-02695-w","publication_status":"peer_reviewed","year":2025,"artifact_sha256":"76082e1cd992d2c09c38f86d05aba575cc76c5022b53a297123b713bb1ce9267","artifact_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12068927/","artifact_retrieved_at":"2026-09-16T10:45:41.099916+00:00","legacy_paper":{"id":"structure-informed-plm-2025","title":"Structure-Informed Protein Language Models are Robust Predictors for Variant Effects","year":2025,"publication_status":"peer_reviewed","version":"Human Genetics 2025 journal article (online 2024)","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12068927/","primary_domain":"proteins-complexes","retrieved_utc":"2026-09-15T23:33:26Z","doi":"10.1007/s00439-024-02695-w","notes":"Final Human Genetics Table 4, AA+SS+RSA+CM AUROC .803 checked directly; Research Square preprint Table 3 prints the same value."},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"transbind-2026","kind":"source","name":"Integrating protein and DNA embeddings for improving genome-wide transcription factor binding site prediction","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["molecular-interactions"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC13145115/","version":"version of record","retrieved_at":"2026-09-15T23:37:05Z","doi":"10.1093/nargab/lqag047","publication_status":"peer_reviewed","year":2026,"artifact_sha256":"5d777f5925e941b7d087035d5d87e79ef75ae8d6456a770ffe8c527da566fee0","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC13145115/fullTextXML","artifact_retrieved_at":"2026-09-16T10:33:58.585Z","legacy_paper":{"id":"transbind-2026","title":"Integrating protein and DNA embeddings for improving genome-wide transcription factor binding site prediction","year":2026,"publication_status":"peer_reviewed","version":"version of record","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC13145115/","primary_domain":"molecular-interactions","retrieved_utc":"2026-09-15T23:37:05Z","notes":"Primary full text verified via Europe PMC fullTextXML; venue: NAR Genomics and Bioinformatics; PMC ID: PMC13145115. Protein-DNA model; comparator scores in table not copied into this batch.","doi":"10.1093/nargab/lqag047"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"tu-fold-2025","kind":"source","name":"RNA secondary structure prediction by conducting multi-class classifications","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["rna-transcriptomes"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12008525/","version":"version of record","retrieved_at":"2026-09-15T23:25:00Z","doi":"10.1016/j.csbj.2025.04.001","publication_status":"peer_reviewed","year":2025,"artifact_sha256":"5aa376d6466daee83fc307baa39fd48da0f185ff30a178624025032d4cbe597d","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12008525/fullTextXML","artifact_retrieved_at":"2026-09-16T10:41:16.505799+00:00","legacy_paper":{"id":"tu-fold-2025","title":"RNA secondary structure prediction by conducting multi-class classifications","year":2025,"publication_status":"peer_reviewed","version":"version of record","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12008525/","primary_domain":"rna-transcriptomes","retrieved_utc":"2026-09-15T23:25:00Z","notes":"Primary full text via Europe PMC XML; venue: Computational and Structural Biotechnology Journal; PMC ID: PMC12008525.","doi":"10.1016/j.csbj.2025.04.001"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"vaxign-esm-2024","kind":"source","name":"Enhancing Vaxign-DL for Vaccine Candidate Prediction with added ESM-Generated Features","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["proteins-complexes"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11398487/","version":"preprint version in PMC","retrieved_at":"2026-09-15T23:37:05Z","doi":"10.1101/2024.09.04.611295","publication_status":"preprint","year":2024,"artifact_sha256":"b76fff917addd0e9ff8a2fc843496132ecf832d3ceef248e3edf2dbc78baab5f","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11398487/fullTextXML","artifact_retrieved_at":"2026-09-16T10:41:06Z","legacy_paper":{"id":"vaxign-esm-2024","title":"Enhancing Vaxign-DL for Vaccine Candidate Prediction with added ESM-Generated Features","year":2024,"publication_status":"preprint","version":"preprint version in PMC","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11398487/","primary_domain":"proteins-complexes","retrieved_utc":"2026-09-15T23:37:05Z","notes":"Primary full text verified via Europe PMC fullTextXML; venue: bioRxiv; PMC ID: PMC11398487. Preprint; combined classifier uses ESM features rather than ESM-only predictions.","doi":"10.1101/2024.09.04.611295"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"viral-contig-simulation-2021","kind":"source","name":"Simulation study and comparative evaluation of viral contiguous sequence identification tools","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["microbes-communities"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC8207588/","version":"version of record","retrieved_at":"2026-09-15T23:29:32Z","doi":"10.1186/s12859-021-04242-0","publication_status":"peer_reviewed","year":2021,"artifact_sha256":"93a24652edfa6d9f686862479df3addf50a2d5d432d2d5a31882075fa59dfdfd","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC8207588/fullTextXML","artifact_retrieved_at":"2026-09-16T10:33:50.134Z","legacy_paper":{"id":"viral-contig-simulation-2021","title":"Simulation study and comparative evaluation of viral contiguous sequence identification tools","year":2021,"publication_status":"peer_reviewed","version":"version of record","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC8207588/","primary_domain":"microbes-communities","retrieved_utc":"2026-09-15T23:29:32Z","notes":"Primary full text verified using Europe PMC XML; venue: BMC Bioinformatics; PMC ID: PMC8207588.","doi":"10.1186/s12859-021-04242-0"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}
{"id":"viral-immune-mimicry-2025","kind":"source","name":"Protein Language Models Expose Viral Immune Mimicry","description":"Primary paper retained with its original identifier. Metadata inherited from the literature collection; individual result checks are separate.","status":"discovered","facets":{"areas":["proteins-complexes"]},"source_ids":[],"links":[],"attributes":{"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12474240/","version":"version of record","retrieved_at":"2026-09-15T23:37:05Z","doi":"10.3390/v17091199","publication_status":"peer_reviewed","year":2025,"artifact_sha256":"15250af2f75f70e2b6a3725d00bc7e276ed9d2d4a54f0ae9d6eaabf6be13e4a1","artifact_url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12474240/fullTextXML","artifact_retrieved_at":"2026-09-16T10:33:57.274Z","legacy_paper":{"id":"viral-immune-mimicry-2025","title":"Protein Language Models Expose Viral Immune Mimicry","year":2025,"publication_status":"peer_reviewed","version":"version of record","source_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12474240/","primary_domain":"proteins-complexes","retrieved_utc":"2026-09-15T23:37:05Z","notes":"Primary full text verified via Europe PMC fullTextXML; venue: Viruses; PMC ID: PMC12474240. Downstream classifier uses ESM2 representations; table does not report a pure zero-shot language-model score.","doi":"10.3390/v17091199"},"scope_decision":"included","missing_metadata":{"licence":"not_reported_in_legacy_extract"}}}