Feng GTEx whole-blood expression prediction, 768 genes, long input
Table 4 gene-expression prediction on the 768-gene long-context subset; Feng et al. 2025 source-reported protocol.
Overview
An explanatory profile has not yet been reviewed for this record.
Consult the linked sources for architecture or protocol details. Missing evidence is not evidence of a missing capability.
3 recorded evaluations, 6 metric rows. A comparison chart has not yet been validated for these results. The table retains the individual findings and their sources.
Results
Results are available, but no reviewed comparison panel is linked in this release.
All evaluations
3 evaluations · 6 results. Different protocols are not a single leaderboard.
Protocol: Feng GTEx whole-blood expression prediction, 768 genes, long input · Dataset: Feng GTEx v8 whole-blood expression, TSS +/- 98 kb, 768 genes
Sorted by MSE (mean across genes of per-gene MSE of covariate-corrected expression) (lower is better). The best value in each column is highlighted. Decimals are rounded for display; each value links to the printed value and its source.
| Tested configuration | MSE (mean across genes of per-gene MSE of covariate-corrected expression) | Pearson (mean across genes of per-gene Pearson correlation between predicted and observed expression) |
|---|---|---|
| HyenaDNA-450K (GTEx whole-blood expression prediction)Evaluation | 0.226 (best in column) | 0.137 (best in column) |
| Caduceus-Ph Long Sequence Input (GTEx whole-blood expression prediction)Evaluation | 0.227 | 0.127 |
| Enformer (GTEx whole-blood expression prediction)Evaluation | 0.227 | 0.129 |
All 6 result rows with coverage, uncertainty and sources
Filter evaluations
Applied filters: All linked evaluations
| Tested configuration | Protocol and dataset | Finding | Evidence and details |
|---|---|---|---|
| Configuration: Caduceus-Ph Long Sequence Input (GTEx whole-blood expression prediction) | Protocol: Feng GTEx whole-blood expression prediction, 768 genes, long input Dataset: Feng GTEx v8 whole-blood expression, TSS +/- 98 kb, 768 genes | 0.227 MSE (mean across genes of per-gene MSE of covariate-corrected expression) unit-unreported · lower Uncertainty: Not reported by the source: Table 4 prints averages only; per-gene spread is shown only as a histogram (Fig. 3). Coverage: Not reported scored / Not reported eligible | Independent external evaluation · Source checkedMethods, coverage and sourceCaduceus-Ph Long Sequence Input GTEx whole-blood expression evaluation Table 4, frozen embeddings + per-gene random forest regression Aggregation: Not reported Benchmarking DNA foundation models for genomic and genetic tasks · Table 4, row 'Caduceus-Ph Long Sequence Input*', column 'Average Prediction MSE' |
| Configuration: Caduceus-Ph Long Sequence Input (GTEx whole-blood expression prediction) | Protocol: Feng GTEx whole-blood expression prediction, 768 genes, long input Dataset: Feng GTEx v8 whole-blood expression, TSS +/- 98 kb, 768 genes | 0.127 Pearson (mean across genes of per-gene Pearson correlation between predicted and observed expression) unitless · higher Uncertainty: Not reported by the source: Table 4 prints averages only; per-gene spread is shown only as a histogram (Fig. 3). Coverage: Not reported scored / Not reported eligible | Independent external evaluation · Source checkedMethods, coverage and sourceCaduceus-Ph Long Sequence Input GTEx whole-blood expression evaluation Table 4, frozen embeddings + per-gene random forest regression Aggregation: Not reported Benchmarking DNA foundation models for genomic and genetic tasks · Table 4, row 'Caduceus-Ph Long Sequence Input*', column 'Average Prediction Correlation' |
| Configuration: Enformer (GTEx whole-blood expression prediction) | Protocol: Feng GTEx whole-blood expression prediction, 768 genes, long input Dataset: Feng GTEx v8 whole-blood expression, TSS +/- 98 kb, 768 genes | 0.227 MSE (mean across genes of per-gene MSE of covariate-corrected expression) unit-unreported · lower Uncertainty: Not reported by the source: Table 4 prints averages only; per-gene spread is shown only as a histogram (Fig. 3). Coverage: Not reported scored / Not reported eligible | Independent external evaluation · Source checkedMethods, coverage and sourceEnformer GTEx whole-blood expression evaluation Table 4, frozen embeddings + per-gene random forest regression Aggregation: Not reported Benchmarking DNA foundation models for genomic and genetic tasks · Table 4, row 'Enformer*', column 'Average Prediction MSE' |
| Configuration: Enformer (GTEx whole-blood expression prediction) | Protocol: Feng GTEx whole-blood expression prediction, 768 genes, long input Dataset: Feng GTEx v8 whole-blood expression, TSS +/- 98 kb, 768 genes | 0.129 Pearson (mean across genes of per-gene Pearson correlation between predicted and observed expression) unitless · higher Uncertainty: Not reported by the source: Table 4 prints averages only; per-gene spread is shown only as a histogram (Fig. 3). Coverage: Not reported scored / Not reported eligible | Independent external evaluation · Source checkedMethods, coverage and sourceEnformer GTEx whole-blood expression evaluation Table 4, frozen embeddings + per-gene random forest regression Aggregation: Not reported Benchmarking DNA foundation models for genomic and genetic tasks · Table 4, row 'Enformer*', column 'Average Prediction Correlation' |
| Configuration: HyenaDNA-450K (GTEx whole-blood expression prediction) | Protocol: Feng GTEx whole-blood expression prediction, 768 genes, long input Dataset: Feng GTEx v8 whole-blood expression, TSS +/- 98 kb, 768 genes | 0.226 MSE (mean across genes of per-gene MSE of covariate-corrected expression) unit-unreported · lower Uncertainty: Not reported by the source: Table 4 prints averages only; per-gene spread is shown only as a histogram (Fig. 3). Coverage: Not reported scored / Not reported eligible | Independent external evaluation · Source checkedMethods, coverage and sourceHyenaDNA-450K GTEx whole-blood expression evaluation Table 4, frozen embeddings + per-gene random forest regression Aggregation: Not reported Benchmarking DNA foundation models for genomic and genetic tasks · Table 4, row 'HyenaDNA-450K*', column 'Average Prediction MSE' |
| Configuration: HyenaDNA-450K (GTEx whole-blood expression prediction) | Protocol: Feng GTEx whole-blood expression prediction, 768 genes, long input Dataset: Feng GTEx v8 whole-blood expression, TSS +/- 98 kb, 768 genes | 0.137 Pearson (mean across genes of per-gene Pearson correlation between predicted and observed expression) unitless · higher Uncertainty: Not reported by the source: Table 4 prints averages only; per-gene spread is shown only as a histogram (Fig. 3). Coverage: Not reported scored / Not reported eligible | Independent external evaluation · Source checkedMethods, coverage and sourceHyenaDNA-450K GTEx whole-blood expression evaluation Table 4, frozen embeddings + per-gene random forest regression Aggregation: Not reported Benchmarking DNA foundation models for genomic and genetic tasks · Table 4, row 'HyenaDNA-450K*', column 'Average Prediction Correlation' |
Source checking is not independent reproduction. Release 2026-11-02-a7f03c9e55a5.
Methods and evaluation design
Procedure, tasks and evaluated configurations
Recorded evaluations
Each evaluation records what was tested and under which conditions.
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.
- External evaluations
- 3
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 (gzip) · Model evaluation matrix (gzip) · Source table (gzip) · Release and checksums (gzip)
Coverage is derived from release 2026-11-02-a7f03c9e55a5. Source citations describe the original records; they do not validate an unreviewed baseline proposal. No results have been generated by this audit.
Run instructions
No runnable recipe has been reviewed for this protocol. Dataset access, model requirements, licences and compute requirements must be checked against its sources before execution.
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
Release 2026-11-02-a7f03c9e55a5 · Record review: source checked
4 source records and release history
- Benchmarking DNA foundation models for genomic and genetic tasks · Original source · PMC12663285.1
- Benchmarking DNA foundation models for genomic and genetic tasks: Supplementary Information · Original source · Nature Communications 16, 10780 (2025); Supplementary Information PDF (MOESM1, 11 pages)
- ChongWuLab/dna_foundation_benchmark README.md · Original source · 3f4c81ce066f3c47422a83466b085aac1a6be902
- ChongWuLab/dna_foundation_benchmark job_scripts/gtex_hyena.py · Original source · 3f4c81ce066f3c47422a83466b085aac1a6be902
Technical metadata and extraction receipts
Stable ID: feng-dna-fm-20261013-protocol-gtex-wb-expression-long-768
- areas
- dna-genomes
- contexts
- research
- scope note
- Per-gene regression of covariate-corrected GTEx v8 whole-blood expression (residuals after regressing sex, PEER factors and genotype principal components) on frozen zero-shot embeddings of each subject's phased sequence around the transcription start site. One random forest regressor per gene; subjects split 75:25 into training and test; fivefold cross-validation on MSE for hyperparameter tuning. Pearson correlation and MSE between predicted and observed expression are computed per gene and averaged across genes. Random forest was chosen over XGBoost because it was better for every model (Supplementary Table 3). Supervised on expression labels; not label-free prediction.
- task
- Subject-level gene expression prediction from personal sequence with long input (average per-gene Pearson correlation and MSE)
- version
- Table 4, asterisked rows
- access note
- Individual-level GTEx v8 whole-genome sequences and expression are protected human-participant data: per the Data availability statement, raw individual WGS must be requested through the GTEx Protected Data Access portal, and the processed expression benchmark is not deposited with the other processed datasets on Hugging Face. Only aggregated GTEx releases are public.
- reproducibility note
- Reproduction requires approved GTEx protected access, the per-model GPU environments described in the author README (pinned at 3f4c81ce066f3c47422a83466b085aac1a6be902), and scheduler-style job scripts that take a job index and placeholder dataset and checkpoint paths. The pinned job_scripts/gtex_hyena.py reads 'long_seqs_by_gene', centre-crops to 6000 bp and writes to an output folder named 'cadph_long'; which script settings produced each Table 4 row cannot be confirmed from the repository. Not attempted by Rewire.
- comparison limit
- Only the three asterisked rows share this 768-gene population. Short-input models on the same genes appear only in Supplementary Table 4 (HyenaDNA and Caduceus-Ph).
- hyperparameters
- Random forest grid: n_estimators 100, 200, 500; max_depth None, 5, 10 (printed as Supplementary Table 5; the main text cites it as Supplementary Table 6).
Related records
- subject: reported_finding: feng-dna-fm-20261013-protocol-gtex-wb-expression-long-768
- assessment: Caduceus-Ph Long Sequence Input GTEx whole-blood expression evaluation
- assessment: Enformer GTEx whole-blood expression evaluation
- assessment: HyenaDNA-450K GTEx whole-blood expression evaluation
- assessed by: Select regulatory variants and genes for functional follow-up