Enformer CAGE gene-expression comparison
Complete two-model across-genes CAGE expression comparison reported numerically in the Enformer paper text for Figure 1b left. This bounded panel is not the entire paper evaluation suite.
Overview
Complete two-model across-genes CAGE expression comparison reported numerically in the Enformer paper text for Figure 1b left. This bounded panel is not the entire paper evaluation suite.
Consult the linked sources for architecture or protocol details. Missing evidence is not evidence of a missing capability.
Results
Each comparison retains its reviewed evaluation scope, dataset and metric. Results are shown without a pooled ranking.
Enformer CAGE gene-expression comparison Across genes CAGE Pearson: Mean across-experiment Pearson correlation of human test-gene CAGE expression
mean_across_experiment_gene_pearson (dimensionless) · Higher values are better.
Every method Enformer CAGE gene-expression comparison reports on Mean across-experiment Pearson correlation of human test-gene CAGE expression, scored with Mean across-experiment Pearson correlation across genes on Enformer/Basenji2 human held-out test genes and CAGE experiments.
Enformer CAGE gene-expression comparison Across genes CAGE Pearson: Mean across-experiment Pearson correlation of human test-gene CAGE expression · Enformer/Basenji2 human held-out test genes and CAGE experiments (Enformer CAGE gene-expression comparison split)
Evidence origin: Author-reported evaluation. Numerical source review does not establish independent reproduction.
Enformer primary article: exact Figure 1b across-genes comparison reported in text · Results / Enformer improves gene expression prediction, Par7; Fig1b left caption; Methods Par32–36Only the complete two-model Figure 1b-left across-genes comparison is extracted; no claim to cover the full paper.
All comparison limitations (5)
- Only the complete two-model Figure 1b-left across-genes comparison is extracted; no claim to cover the full paper.
- Mean Pearson values 0.81/0.85 are printed in Par7; separate ExPecto Spearman values 0.812/0.850 later in that paragraph are a different comparison.
- Figure 1b caption states bootstrap SD 0.004 for across-genes estimates; retained as shared source context, not converted into confidence intervals or assumed per-model training-run variability.
- Basenji2 is the pretrained main-comparison configuration, distinct from original Basenji1 and from retrained ablation models.
- Experimental replicate accuracy 0.94 is contextual, not a third model score. Input lengths, architectures and training procedures differ.
Automated source review: 2026-09-23.
No unavailable values; missing scores remain labelled and are never plotted as zero.
Showing 2 of 2 matching rows.
Dots show point estimates. Whiskers show only explicitly defined uncertainty (standard deviation, standard error or a labelled interval); their definitions remain in Table. Unresolved uncertainty is not plotted. Differences do not establish statistical significance.
Methods and evaluation design
Procedure, tasks and evaluated configurations
Evaluation design
Benchmarks bring together tasks and protocols. A task describes the biological question; a protocol defines a particular test.
These source-backed links do not make different protocols or scores interchangeable.
Baseline coverage
Reference methods help show what a model adds beyond simple controls. We track a null control and a conventional method for each protocol.
0 of 2 active baseline roles have published Rewire measurements in this release. Measurements on a selected protocol do not establish coverage of an entire suite.
No execution recipe linked to this protocol. Recipe availability does not establish a completed evaluation.
No reviewed evaluations with results linked in this release.
Literature evidence is not a Rewire measurement. Executed but unpublished runs and private review status are not included.
Null control
Proposed control: requires review
Select a task-valid null control after reviewing inputs and metric
Protocol-specific applicability, permitted inputs, access, split, evaluator and execution requirements need review before implementation or execution.
This is a suggested selection rule, not a validated method or a measured score.
Conventional reference
Proposed control: requires review
Select an upstream conventional reference after reviewing the full protocol
Protocol-specific applicability, permitted inputs, access, split, evaluator and execution requirements need review before implementation or execution.
This is a suggested selection rule, not a validated method or a measured score.
Protocol coverage CSV · Model evaluation matrix · Source table · Release and checksums
Coverage is derived from release 2026-09-23-2b89723c6dd9. Source citations describe the original records; they do not validate an unreviewed baseline proposal. No results have been generated by this audit.
Run this benchmark
Choose a concrete protocol before running an evaluation. Its inputs, split and scoring rules determine which results can be compared.
Run this benchmark
Human held-out test-set protein-coding genes; CAGE read counts summed over all unique TSS locations, using TSS-overlapping 128-bp bin plus two neighboring bins, log(1+x)-transformed and standardized across genes separately per experiment. Pearson correlation across genes per CAGE experiment, then mean across experiments. Same Basenji2 dataset and genomic intervals; human split 34,021 training/2,213 validation/1,937 test sequences. Cross-species homologous 1 Mb regions partitioned by connected components. Test-time mean of 8 random ≤3 bp shift/reverse-complement augmentations. Enformer input 196,608 bp, Basenji2 131,072 bp; receptive fields differ. Validation used for tuning; main comparison on test set. No independent Rewire execution; exact checkpoint and split artifact hashes remain unextracted.
A maintained rewire runner has not been verified for this benchmark. Check data access, weights, licences, dependencies and hardware in the linked official documentation; requirements have not been fully extracted.
Enformer primary article: exact Figure 1b across-genes comparison reported in text · Methods Par32–36Strengths, limitations and unresolved questions
Evidence
Source checking verifies the cited claim or transcription. It does not establish independent reproduction.
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.
0 evidence rows matching the loaded filters
| Property and statement | Original source and location | Review and provenance |
|---|
No evidence rows match these filters. Choose another scope or clear the search.
Sources and history
View linked audit checks and correction history
Release 2026-09-23-2b89723c6dd9 · Record review: source checked
1 source records and release history
- Enformer primary article: exact Figure 1b across-genes comparison reported in text · Original source · PMC8490152; SHA-256 pinned XML snapshot
Technical metadata and extraction receipts
Stable ID: enformer-cage-gene-expression-2021
- areas
- genomics
- source locator
- Results / Enformer improves gene expression prediction, Par7; Fig1b left caption; Methods Par32–36
- run documentation
- record id: enformer-cage-gene-expression-2021; status: source_reviewed_not_executed; source ids: coverage-enformer-2021-primary; source locator: Methods Par32–36; summary: Human held-out test-set protein-coding genes; CAGE read counts summed over all unique TSS locations, using TSS-overlapping 128-bp bin plus two neighboring bins, log(1+x)-transformed and standardized across genes separately per experiment. Pearson correlation across genes per CAGE experiment, then mean across experiments. Same Basenji2 dataset and genomic intervals; human split 34,021 training/2,213 validation/1,937 test sequences. Cross-species homologous 1 Mb regions partitioned by connected components. Test-time mean of 8 random ≤3 bp shift/reverse-complement augmentations. Enformer input 196,608 bp, Basenji2 131,072 bp; receptive fields differ. Validation used for tuning; main comparison on test set. No independent Rewire execution; exact checkpoint and split artifact hashes remain unextracted.
- entity level
- protocol
- legacy kinds
- benchmark