Model type
Protein sequence transformer; this record is the paper-specific evaluated configuration.
Configuration
ESM2 650M
This ESM-2 650M configuration is fine-tuned with LoRA to distinguish human from viral protein sequences.
Sources
Protein Language Models Expose Viral Immune Mimicry · Section 2.3 Human-Virus Model Training and Implementation, paragraphs 1–3; Section 2.4 Finding and Analyzing Model Mistakes, paragraphs 1–2; Table 1, ESM2 650M rowHow it worksEvaluated procedure (conceptual)
Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.
Sources
Protein Language Models Expose Viral Immune Mimicry · Section 2.3 Human-Virus Model Training and Implementation, paragraphs 1–3; Section 2.4 Finding and Analyzing Model Mistakes, paragraphs 1–2; Table 1, ESM2 650M rowAt a glance
Biological inputs
Protein sequences represented by pretrained embeddings
Sources
Protein Language Models Expose Viral Immune Mimicry · 2. Materials and Methods/2.2. Pretrained Deep Language Models (ESM, T5) (paragraph 1); 3. Results/3.4. Latent Structure Embeddings Clustering (paragraph 1)Outputs
Viral-versus-human classification scores
Sources
Protein Language Models Expose Viral Immune Mimicry · 3. Results/3.5. Immunogenicity Analysis (paragraph 1); 3. Results (paragraph 1)limited source coverage · Automated source review, 2026-09-17. All specifications and missing details
Evaluations and results
Release 2026-09-17-d277315f7d76 · 1 evaluation · 4 metric rows. Different protocols are not a single leaderboard.
| Metric and finding | Coverage and uncertainty | Evidence |
|---|---|---|
| ESM2 650M: human-versus-viral protein classification Configuration: ESM2 650MProtocol: Held-out human-versus-virus protein classification (human-versus-viral protein classification)Dataset: human and viral proteins Compare protein origin classifiers; T 5 embeddings plus linear/tree models vs ESM 2 fine-tuning and simple controls. UniRef90 deduplication; proteins longer than1,600 residues excluded; UniRef50 clusters assigned80% training and20% test with no cluster shared. Separate four-fold error-analysis experiment not assigned toTable1. Author-reported evaluation · Evaluation metadata: needs review | ||
| 99.67% AUROC Unit: percent · Direction: higher | Uncertainty: not reported in legacy extract Scored: Not reported · Eligible: Not reported | source checkedProtein Language Models Expose Viral Immune Mimicry; Protein Language Models Expose Viral Immune Mimicry · Table 1, ESM2 650M row, AUC (%) column Source checking is not independent reproduction. |
| 96.85% Prec. Unit: percent · Direction: higher | Uncertainty: unreported Scored: Not reported · Eligible: Not reported | source checkedProtein Language Models Expose Viral Immune Mimicry · Table 1, row ESM2 650M, column Prec.; XML row7 column4 Source checking is not independent reproduction. |
| 97.86% Accur. Unit: percent · Direction: higher | Uncertainty: unreported Scored: Not reported · Eligible: Not reported | source checkedProtein Language Models Expose Viral Immune Mimicry · Table 1, row ESM2 650M, column Accur.; XML row7 column3 Source checking is not independent reproduction. |
| 96.68% Recall Unit: percent · Direction: higher | Uncertainty: unreported Scored: Not reported · Eligible: Not reported | source checkedProtein Language Models Expose Viral Immune Mimicry · Table 1, row ESM2 650M, column Recall; XML row7 column5 Source checking is not independent reproduction. |
How it works
How the evaluated method works
The ESM-2 650M model is fine-tuned with LoRA on protein sequences to classify human versus viral origin. The separate embedding-based linear/tree comparisons use T5; the paper’s immune-mimicry error analysis uses Linear-T5, not this ESM-2 configuration.
Sources
Protein Language Models Expose Viral Immune Mimicry · Section 2.3 Human-Virus Model Training and Implementation, paragraphs 1–3; Section 2.4 Finding and Analyzing Model Mistakes, paragraphs 1–2; Table 1, ESM2 650M rowUnderlying method and version boundaries
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.
Sources
facebookresearch/esm README.md · README.md; introduction, model description, pretrained-model and usage sections at pinned revisionWhat was evaluated
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.
Sources
Protein Language Models Expose Viral Immune Mimicry · The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b4-009Strengths and limitations
Strengths and considerations
- Cluster-based splitting reduces close-sequence overlap between train and test.
Sources
Protein Language Models Expose Viral Immune Mimicry · 2. Materials and Methods/2.1. Protein Datasets (paragraph 2); 2. Materials and Methods/2.4. Finding and Analyzing Model Mistakes (paragraph 1)
Limitations and conditions
- Classification errors correlated with immune features do not establish that the model measures actual immune recognition or viral pathogenicity.
Sources
Protein Language Models Expose Viral Immune Mimicry · 3. Results (paragraph 1); Abstract (paragraph 1)
Profile review details
Previous profile source review retained. On 2026-09-17, an independent agent and root agent checked Methods 2.3–2.4 and corrected the ESM-2 LoRA mechanism, diagram and parameter wording. This correction does not constitute human review or benchmark reproduction.
Stable record: reported-model-4c73500c39e9d0Specifications
Inputs, training, access and other details
Explanatory profile: limited source coverage · Automated source review, 2026-09-17. Review applies to the cited claims; unresolved fields are listed below. Numerical results retain their own review status.
| Property | Description and evidence |
|---|---|
| Model type | Protein sequence transformer; this record is the paper-specific evaluated configuration.Sourcesfacebookresearch/esm README.md · README.md model description |
| Architecture / procedure | The ESM-2 650M model is fine-tuned with LoRA on protein sequences to classify human versus viral origin. The separate embedding-based linear/tree comparisons use T5; the paper’s immune-mimicry error analysis uses Linear-T5, not this ESM-2 configuration.SourcesProtein Language Models Expose Viral Immune Mimicry · Section 2.3 Human-Virus Model Training and Implementation, paragraphs 1–3; Section 2.4 Finding and Analyzing Model Mistakes, paragraphs 1–2; Table 1, ESM2 650M row |
| Biological inputs | Protein sequences represented by pretrained embeddingsSourcesProtein Language Models Expose Viral Immune Mimicry · 2. Materials and Methods/2.2. Pretrained Deep Language Models (ESM, T5) (paragraph 1); 3. Results/3.4. Latent Structure Embeddings Clustering (paragraph 1) |
| Outputs | Viral-versus-human classification scoresSourcesProtein Language Models Expose Viral Immune Mimicry · 3. Results/3.5. Immunogenicity Analysis (paragraph 1); 3. Results (paragraph 1) |
| Parameters | The ESM-2 backbone is labelled 650M parameters. The paper uses LoRA with rank 8 and scaling factor 8; it does not report an aggregate trained-configuration parameter count.SourcesProtein Language Models Expose Viral Immune Mimicry · Section 2.3 Human-Virus Model Training and Implementation, paragraphs 1–3; Section 2.4 Finding and Analyzing Model Mistakes, paragraphs 1–2; Table 1, ESM2 650M row |
| Known versions / configuration | ESM2 650M is the comparison-table label; that label does not specify an immutable weight revision. · Not reported in inspected sourcesSourcesProtein Language Models Expose Viral Immune Mimicry · Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label. |
| Training data / fitting | UniRef90 duplicate removal and a UniRef50-clustered 80:20 split; sequences above 1,600 residues are excluded.SourcesProtein Language Models Expose Viral Immune Mimicry · 2. Materials and Methods/2.1. Protein Datasets (paragraph 2); 2. Materials and Methods/2.1. Protein Datasets (paragraph 1) |
| Context limits | The study excludes proteins longer than 1,600 residues.SourcesProtein Language Models Expose Viral Immune Mimicry · 2. Materials and Methods/2.1. Protein Datasets (paragraph 2); 3. Results/3.3. Virus Errors Analysis (paragraph 2) |
| Access | 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.Sourcesfacebookresearch/esm README.md · README.md; installation, model download and usage instructions |
| Code licence | MIT (upstream repository code at the cited revision; this does not establish every dependency or historical checkpoint licence).Sourcesfacebookresearch/esm LICENSE · LICENSE; complete licence text |
| Weights licence | The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately. · Not reported in inspected sourcesSourcesfacebookresearch/esm README.md · README.md; checkpoint/access documentation and licence scope |
Evidence table
Inspect claims, sources and review details
Trace each statement to its source and review. A context-only reference supports the record generally; it does not verify an individual field. Source checking does not reproduce an experiment.
One row per statement and cited source. Multiple citations are not independent evaluations. Shared locators are labelled explicitly.
20 evidence rows matching the loaded filters
| Property and statement | Original source and location | Review and provenance |
|---|---|---|
| Diagram caption Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings. Individual claims | Protein Language Models Expose Viral Immune Mimicry Section 2.3 Human-Virus Model Training and Implementation, paragraphs 1–3; Section 2.4 Finding and Analyzing Model Mistakes, paragraphs 1–2; Table 1, ESM2 650M row Version: version of record | source checked automated source review · 2026-09-17 Audit detailsPrevious profile source review retained. On 2026-09-17, an independent agent and root agent checked Methods 2.3–2.4 and corrected the ESM-2 LoRA mechanism, diagram and parameter wording. This correction does not constitute human review or benchmark reproduction. Field: Source artifact SHA-256: Hash scope: Hash scope not separately documented; inspect source record |
| Diagram steps ["Protein sequences","ESM-2 650M fine-tuned with LoRA","Human-versus-viral classification scores"] Individual claims | Protein Language Models Expose Viral Immune Mimicry Section 2.3 Human-Virus Model Training and Implementation, paragraphs 1–3; Section 2.4 Finding and Analyzing Model Mistakes, paragraphs 1–2; Table 1, ESM2 650M row Version: version of record | source checked automated source review · 2026-09-17 Audit detailsPrevious profile source review retained. On 2026-09-17, an independent agent and root agent checked Methods 2.3–2.4 and corrected the ESM-2 LoRA mechanism, diagram and parameter wording. This correction does not constitute human review or benchmark reproduction. Field: Source artifact SHA-256: Hash scope: Hash scope not separately documented; inspect source record |
| Diagram title Evaluated procedure (conceptual) Individual claims | Protein Language Models Expose Viral Immune Mimicry Section 2.3 Human-Virus Model Training and Implementation, paragraphs 1–3; Section 2.4 Finding and Analyzing Model Mistakes, paragraphs 1–2; Table 1, ESM2 650M row Version: version of record | source checked automated source review · 2026-09-17 Audit detailsPrevious profile source review retained. On 2026-09-17, an independent agent and root agent checked Methods 2.3–2.4 and corrected the ESM-2 LoRA mechanism, diagram and parameter wording. This correction does not constitute human review or benchmark reproduction. Field: Source artifact SHA-256: Hash scope: Hash scope not separately documented; inspect source record |
| Model type Protein sequence transformer; this record is the paper-specific evaluated configuration. Individual claims | facebookresearch/esm README.md README.md model description Version: 2b369911bb5b4b0dda914521b9475cad1656b2ac | source checked automated source review · 2026-09-17 Audit detailsPrevious profile source review retained. On 2026-09-17, an independent agent and root agent checked Methods 2.3–2.4 and corrected the ESM-2 LoRA mechanism, diagram and parameter wording. This correction does not constitute human review or benchmark reproduction. Field: Source artifact SHA-256: Hash scope: Hash scope not separately documented; inspect source record |
| Architecture / procedure The ESM-2 650M model is fine-tuned with LoRA on protein sequences to classify human versus viral origin. The separate embedding-based linear/tree comparisons use T5; the paper’s immune-mimicry error analysis uses Linear-T5, not this ESM-2 configuration. Individual claims | Protein Language Models Expose Viral Immune Mimicry Section 2.3 Human-Virus Model Training and Implementation, paragraphs 1–3; Section 2.4 Finding and Analyzing Model Mistakes, paragraphs 1–2; Table 1, ESM2 650M row Version: version of record | source checked automated source review · 2026-09-17 Audit detailsPrevious profile source review retained. On 2026-09-17, an independent agent and root agent checked Methods 2.3–2.4 and corrected the ESM-2 LoRA mechanism, diagram and parameter wording. This correction does not constitute human review or benchmark reproduction. Field: Source artifact SHA-256: Hash scope: Hash scope not separately documented; inspect source record |
| Weights licence The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately. Individual claims | facebookresearch/esm README.md README.md; checkpoint/access documentation and licence scope Version: 2b369911bb5b4b0dda914521b9475cad1656b2ac | unreported automated source review · 2026-09-17 Audit detailsPrevious profile source review retained. On 2026-09-17, an independent agent and root agent checked Methods 2.3–2.4 and corrected the ESM-2 LoRA mechanism, diagram and parameter wording. This correction does not constitute human review or benchmark reproduction. Field: Source artifact SHA-256: Hash scope: Hash scope not separately documented; inspect source record |
| Biological inputs Protein sequences represented by pretrained embeddings Individual claims | Protein Language Models Expose Viral Immune Mimicry 2. Materials and Methods/2.2. Pretrained Deep Language Models (ESM, T5) (paragraph 1); 3. Results/3.4. Latent Structure Embeddings Clustering (paragraph 1) Version: version of record | source checked automated source review · 2026-09-17 Audit detailsPrevious profile source review retained. On 2026-09-17, an independent agent and root agent checked Methods 2.3–2.4 and corrected the ESM-2 LoRA mechanism, diagram and parameter wording. This correction does not constitute human review or benchmark reproduction. Field: Source artifact SHA-256: Hash scope: Hash scope not separately documented; inspect source record |
| Outputs Viral-versus-human classification scores Individual claims | Protein Language Models Expose Viral Immune Mimicry 3. Results/3.5. Immunogenicity Analysis (paragraph 1); 3. Results (paragraph 1) Version: version of record | source checked automated source review · 2026-09-17 Audit detailsPrevious profile source review retained. On 2026-09-17, an independent agent and root agent checked Methods 2.3–2.4 and corrected the ESM-2 LoRA mechanism, diagram and parameter wording. This correction does not constitute human review or benchmark reproduction. Field: Source artifact SHA-256: Hash scope: Hash scope not separately documented; inspect source record |
| Parameters The ESM-2 backbone is labelled 650M parameters. The paper uses LoRA with rank 8 and scaling factor 8; it does not report an aggregate trained-configuration parameter count. Individual claims | Protein Language Models Expose Viral Immune Mimicry Section 2.3 Human-Virus Model Training and Implementation, paragraphs 1–3; Section 2.4 Finding and Analyzing Model Mistakes, paragraphs 1–2; Table 1, ESM2 650M row Version: version of record | source checked automated source review · 2026-09-17 Audit detailsPrevious profile source review retained. On 2026-09-17, an independent agent and root agent checked Methods 2.3–2.4 and corrected the ESM-2 LoRA mechanism, diagram and parameter wording. This correction does not constitute human review or benchmark reproduction. Field: Source artifact SHA-256: Hash scope: Hash scope not separately documented; inspect source record |
| Known versions / configuration ESM2 650M is the comparison-table label; that label does not specify an immutable weight revision. Individual claims | Protein Language Models Expose Viral Immune Mimicry Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label. Version: version of record | unreported automated source review · 2026-09-17 Audit detailsPrevious profile source review retained. On 2026-09-17, an independent agent and root agent checked Methods 2.3–2.4 and corrected the ESM-2 LoRA mechanism, diagram and parameter wording. This correction does not constitute human review or benchmark reproduction. Field: Source artifact SHA-256: Hash scope: Hash scope not separately documented; inspect source record |
Sources and history
Release 2026-09-17-d277315f7d76 · Record review: needs review
3 source records and release history
- Protein Language Models Expose Viral Immune Mimicry · Original source · version of record
- facebookresearch/esm README.md · Original source · 2b369911bb5b4b0dda914521b9475cad1656b2ac
- facebookresearch/esm LICENSE · Original source · 2b369911bb5b4b0dda914521b9475cad1656b2ac
Technical metadata and extraction receipts
Stable ID: reported-model-4c73500c39e9d0
- areas
- proteins-complexes
- entity level
- method
- version
- Not reported
- 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.
- legacy kinds
- model
- entity classification
- review date: 2026-09-17; rationale: This source-scoped entry preserves the method/configuration actually named in an evaluation. It is neither a global family identity nor proof of an immutable checkpoint; the linked evaluation retains adaptation, fitting and scoring details.; source ids: viral-immune-mimicry-2025; evidence-reported-base-esm-readme-md; source locator: 3. Results/3.2. Error Analysis Models Insights (paragraph 1); Abstract (paragraph 1) | README.md model description | 3. Results (paragraph 1); 3. Results/3.4. Latent Structure Embeddings Clustering (paragraph 1); ambiguities: Configuration means the source-labelled evaluated identity. It does not establish missing checkpoint hashes, default settings or equivalence to same-named records in other papers.
- profile correction
- review date: 2026-09-17; review method: Independent agent inspection of cached complete primary article Methods2.3–2.4 and Table1, cross-checked against preserved evaluation-lit-b4-009; rationale: The current mechanism and diagram conflate the ESM2 Table 1 fine-tuning configuration with T5 static-embedding classifiers and the separate Linear-T5 error analysis. Sections2.3–2.4 explicitly distinguish these. Preserve configuration kind and all numerical results; correct only explanatory content.; source ids: viral-immune-mimicry-2025; source locator: Section 2.3 Human-Virus Model Training and Implementation, paragraphs 1–3; Section 2.4 Finding and Analyzing Model Mistakes, paragraphs 1–2; Table 1, ESM2 650M row