Dataset subset

Lu et al. 2025 ClinVar benchmark: Stop gain (Figure 1)

ClinVar SNVs in the 'Stop gain' group of the Lu et al. 2025 benchmark, restricted to variants scored by every model in the panel.

Evaluation results

12 evaluations · 12 results. Different protocols are not a single leaderboard.

Filter evaluations

Tested model or methodProtocolDatasetEvaluation setupMetricEvidence origin

Applied filters: All linked evaluations

Exact evaluated configurations and original reported results
Tested configuration	Protocol and dataset	Finding	Evidence and details
Configuration: AlphaGenome all outputs, quantile, maximum (Lu et al. 2025)	Protocol: ClinVar pathogenic versus benign SNVs within variant type, AUROC (Lu et al. 2025) Dataset subset: Lu et al. 2025 ClinVar benchmark: Stop gain (Figure 1)	0.389 AUROC (ClinVar pathogenic vs benign SNVs, Stop gain) unitless · higher Uncertainty: Not reported by the source Coverage: Not reported scored / Not reported eligible	Independent external evaluation · Source checked Methods, coverage and source AlphaGenome on ClinVar Stop gain SNVs (Lu et al. 2025 Figure 1) lu-hetero-20261019-protocol-clinvar-variant-type-auroc Aggregation: Not reported Genomic heterogeneity inflates the performance of variant pathogenicity predictions (Lu, Liu, Lin and Brandes, bioRxiv 2025, v2) · Figure 1 (page 8), column 3, panel 4 'Stop gain (N=42,408, %P=98.13%)', row 'AlphaGenome', bar label
Configuration: AlphaMissense precomputed scores (Lu et al. 2025)	Protocol: ClinVar pathogenic versus benign SNVs within variant type, AUROC (Lu et al. 2025) Dataset subset: Lu et al. 2025 ClinVar benchmark: Stop gain (Figure 1)	0.849 AUROC (ClinVar pathogenic vs benign SNVs, Stop gain) unitless · higher Uncertainty: Not reported by the source Coverage: Not reported scored / Not reported eligible	Independent external evaluation · Source checked Methods, coverage and source AlphaMissense on ClinVar Stop gain SNVs (Lu et al. 2025 Figure 1) lu-hetero-20261019-protocol-clinvar-variant-type-auroc Aggregation: Not reported Genomic heterogeneity inflates the performance of variant pathogenicity predictions (Lu, Liu, Lin and Brandes, bioRxiv 2025, v2) · Figure 1 (page 8), column 3, panel 4 'Stop gain (N=42,408, %P=98.13%)', row 'AlphaMissense', bar label
Configuration: DNABERT-2 117M, CLS embedding L2 distance (Lu et al. 2025)	Protocol: ClinVar pathogenic versus benign SNVs within variant type, AUROC (Lu et al. 2025) Dataset subset: Lu et al. 2025 ClinVar benchmark: Stop gain (Figure 1)	0.483 AUROC (ClinVar pathogenic vs benign SNVs, Stop gain) unitless · higher Uncertainty: Not reported by the source Coverage: Not reported scored / Not reported eligible	Independent external evaluation · Source checked Methods, coverage and source DNABERT2 on ClinVar Stop gain SNVs (Lu et al. 2025 Figure 1) lu-hetero-20261019-protocol-clinvar-variant-type-auroc Aggregation: Not reported Genomic heterogeneity inflates the performance of variant pathogenicity predictions (Lu, Liu, Lin and Brandes, bioRxiv 2025, v2) · Figure 1 (page 8), column 3, panel 4 'Stop gain (N=42,408, %P=98.13%)', row 'DNABERT2', bar label
Configuration: ESM1b 650M log-likelihood ratio, Brandes et al. pipeline (Lu et al. 2025)	Protocol: ClinVar pathogenic versus benign SNVs within variant type, AUROC (Lu et al. 2025) Dataset subset: Lu et al. 2025 ClinVar benchmark: Stop gain (Figure 1)	0.788 AUROC (ClinVar pathogenic vs benign SNVs, Stop gain) unitless · higher Uncertainty: Not reported by the source Coverage: Not reported scored / Not reported eligible	Independent external evaluation · Source checked Methods, coverage and source ESM1b on ClinVar Stop gain SNVs (Lu et al. 2025 Figure 1) lu-hetero-20261019-protocol-clinvar-variant-type-auroc Aggregation: Not reported Genomic heterogeneity inflates the performance of variant pathogenicity predictions (Lu, Liu, Lin and Brandes, bioRxiv 2025, v2) · Figure 1 (page 8), column 3, panel 4 'Stop gain (N=42,408, %P=98.13%)', row 'ESM1b', bar label
Configuration: ESM1v model 1 log-likelihood ratio, Brandes et al. pipeline (Lu et al. 2025)	Protocol: ClinVar pathogenic versus benign SNVs within variant type, AUROC (Lu et al. 2025) Dataset subset: Lu et al. 2025 ClinVar benchmark: Stop gain (Figure 1)	0.656 AUROC (ClinVar pathogenic vs benign SNVs, Stop gain) unitless · higher Uncertainty: Not reported by the source Coverage: Not reported scored / Not reported eligible	Independent external evaluation · Source checked Methods, coverage and source ESM1v on ClinVar Stop gain SNVs (Lu et al. 2025 Figure 1) lu-hetero-20261019-protocol-clinvar-variant-type-auroc Aggregation: Not reported Genomic heterogeneity inflates the performance of variant pathogenicity predictions (Lu, Liu, Lin and Brandes, bioRxiv 2025, v2) · Figure 1 (page 8), column 3, panel 4 'Stop gain (N=42,408, %P=98.13%)', row 'ESM1v', bar label
Configuration: ESM2 650M log-likelihood ratio, Brandes et al. pipeline (Lu et al. 2025)	Protocol: ClinVar pathogenic versus benign SNVs within variant type, AUROC (Lu et al. 2025) Dataset subset: Lu et al. 2025 ClinVar benchmark: Stop gain (Figure 1)	0.684 AUROC (ClinVar pathogenic vs benign SNVs, Stop gain) unitless · higher Uncertainty: Not reported by the source Coverage: Not reported scored / Not reported eligible	Independent external evaluation · Source checked Methods, coverage and source ESM2 on ClinVar Stop gain SNVs (Lu et al. 2025 Figure 1) lu-hetero-20261019-protocol-clinvar-variant-type-auroc Aggregation: Not reported Genomic heterogeneity inflates the performance of variant pathogenicity predictions (Lu, Liu, Lin and Brandes, bioRxiv 2025, v2) · Figure 1 (page 8), column 3, panel 4 'Stop gain (N=42,408, %P=98.13%)', row 'ESM2', bar label
Configuration: Evo 2 7B base, 8,192 bp log-likelihood ratio (Lu et al. 2025)	Protocol: ClinVar pathogenic versus benign SNVs within variant type, AUROC (Lu et al. 2025) Dataset subset: Lu et al. 2025 ClinVar benchmark: Stop gain (Figure 1)	0.663 AUROC (ClinVar pathogenic vs benign SNVs, Stop gain) unitless · higher Uncertainty: Not reported by the source Coverage: Not reported scored / Not reported eligible	Independent external evaluation · Source checked Methods, coverage and source Evo2 on ClinVar Stop gain SNVs (Lu et al. 2025 Figure 1) lu-hetero-20261019-protocol-clinvar-variant-type-auroc Aggregation: Not reported Genomic heterogeneity inflates the performance of variant pathogenicity predictions (Lu, Liu, Lin and Brandes, bioRxiv 2025, v2) · Figure 1 (page 8), column 3, panel 4 'Stop gain (N=42,408, %P=98.13%)', row 'Evo2', bar label
Configuration: GPN-MSA precomputed hg38 scores (Lu et al. 2025)	Protocol: ClinVar pathogenic versus benign SNVs within variant type, AUROC (Lu et al. 2025) Dataset subset: Lu et al. 2025 ClinVar benchmark: Stop gain (Figure 1)	0.783 AUROC (ClinVar pathogenic vs benign SNVs, Stop gain) unitless · higher Uncertainty: Not reported by the source Coverage: Not reported scored / Not reported eligible	Independent external evaluation · Source checked Methods, coverage and source GPN-MSA on ClinVar Stop gain SNVs (Lu et al. 2025 Figure 1) lu-hetero-20261019-protocol-clinvar-variant-type-auroc Aggregation: Not reported Genomic heterogeneity inflates the performance of variant pathogenicity predictions (Lu, Liu, Lin and Brandes, bioRxiv 2025, v2) · Figure 1 (page 8), column 3, panel 4 'Stop gain (N=42,408, %P=98.13%)', row 'GPN-MSA', bar label
Configuration: Nucleotide Transformer 2.5B multi-species, CLS embedding L2 distance (Lu et al. 2025)	Protocol: ClinVar pathogenic versus benign SNVs within variant type, AUROC (Lu et al. 2025) Dataset subset: Lu et al. 2025 ClinVar benchmark: Stop gain (Figure 1)	0.618 AUROC (ClinVar pathogenic vs benign SNVs, Stop gain) unitless · higher Uncertainty: Not reported by the source Coverage: Not reported scored / Not reported eligible	Independent external evaluation · Source checked Methods, coverage and source Nucleotide Transformer on ClinVar Stop gain SNVs (Lu et al. 2025 Figure 1) lu-hetero-20261019-protocol-clinvar-variant-type-auroc Aggregation: Not reported Genomic heterogeneity inflates the performance of variant pathogenicity predictions (Lu, Liu, Lin and Brandes, bioRxiv 2025, v2) · Figure 1 (page 8), column 3, panel 4 'Stop gain (N=42,408, %P=98.13%)', row 'Nucleotide Transformer', bar label
Configuration: PhyloGPN, official example code (Lu et al. 2025)	Protocol: ClinVar pathogenic versus benign SNVs within variant type, AUROC (Lu et al. 2025) Dataset subset: Lu et al. 2025 ClinVar benchmark: Stop gain (Figure 1)	0.677 AUROC (ClinVar pathogenic vs benign SNVs, Stop gain) unitless · higher Uncertainty: Not reported by the source Coverage: Not reported scored / Not reported eligible	Independent external evaluation · Source checked Methods, coverage and source PhyloGPN on ClinVar Stop gain SNVs (Lu et al. 2025 Figure 1) lu-hetero-20261019-protocol-clinvar-variant-type-auroc Aggregation: Not reported Genomic heterogeneity inflates the performance of variant pathogenicity predictions (Lu, Liu, Lin and Brandes, bioRxiv 2025, v2) · Figure 1 (page 8), column 3, panel 4 'Stop gain (N=42,408, %P=98.13%)', row 'PhyloGPN', bar label
Configuration: PhyloP 100-way vertebrate (Lu et al. 2025)	Protocol: ClinVar pathogenic versus benign SNVs within variant type, AUROC (Lu et al. 2025) Dataset subset: Lu et al. 2025 ClinVar benchmark: Stop gain (Figure 1)	0.704 AUROC (ClinVar pathogenic vs benign SNVs, Stop gain) unitless · higher Uncertainty: Not reported by the source Coverage: Not reported scored / Not reported eligible	Independent external evaluation · Source checked Methods, coverage and source PhyloP on ClinVar Stop gain SNVs (Lu et al. 2025 Figure 1) lu-hetero-20261019-protocol-clinvar-variant-type-auroc Aggregation: Not reported Genomic heterogeneity inflates the performance of variant pathogenicity predictions (Lu, Liu, Lin and Brandes, bioRxiv 2025, v2) · Figure 1 (page 8), column 3, panel 4 'Stop gain (N=42,408, %P=98.13%)', row 'PhyloP', bar label
Configuration: PrimateAI-3D precomputed scores (Lu et al. 2025)	Protocol: ClinVar pathogenic versus benign SNVs within variant type, AUROC (Lu et al. 2025) Dataset subset: Lu et al. 2025 ClinVar benchmark: Stop gain (Figure 1)	0.817 AUROC (ClinVar pathogenic vs benign SNVs, Stop gain) unitless · higher Uncertainty: Not reported by the source Coverage: Not reported scored / Not reported eligible	Independent external evaluation · Source checked Methods, coverage and source PrimateAI-3D* on ClinVar Stop gain SNVs (Lu et al. 2025 Figure 1) lu-hetero-20261019-protocol-clinvar-variant-type-auroc Aggregation: Not reported Genomic heterogeneity inflates the performance of variant pathogenicity predictions (Lu, Liu, Lin and Brandes, bioRxiv 2025, v2) · Figure 1 (page 8), column 3, panel 4 'Stop gain (N=42,408, %P=98.13%)', row 'PrimateAI-3D*', bar label

Source checking is not independent reproduction. Release 2026-11-02-a7f03c9e55a5.

Research readiness

0 of 4 readiness checks met. These checks assess whether the evidence supports a reproducible investigation; a source-checked score alone does not meet them.

Readiness checks, gaps and artifacts

Release 2026-11-02-a7f03c9e55a5 · Evidence verified: Not verified

Evidence incomplete

Replay metrics

Exact outcomes, predictions, identifiers and evaluator are connected.

Missing or unresolved evidence

No verified artifact manifest is linked to this exact record.
File checksums match the recorded files: not yet verified
Predictions are matched to the right samples: not yet verified
Score meaning and direction are confirmed: not yet verified
Metrics are recomputed from the saved predictions: not yet verified

Verified: Not verified

Evidence incomplete

Investigate discrepancies

Replay evidence includes annotations and an assessment of dependence. Unknown independence permits descriptive analysis only.

Missing or unresolved evidence

No verified artifact manifest is linked to this exact record.
File checksums match the recorded files: not yet verified
Predictions are matched to the right samples: not yet verified
Score meaning and direction are confirmed: not yet verified
Metrics are recomputed from the saved predictions: not yet verified
Sample annotations are recorded: not yet verified
Dependence between samples is assessed: not yet verified

Verified: Not verified

Evidence incomplete

Run locally

A pinned recipe describes the inputs, environment and resource requirements.

Missing or unresolved evidence

No verified artifact manifest is linked to this exact record.
File checksums match the recorded files: not yet verified
Predictions are matched to the right samples: not yet verified
Score meaning and direction are confirmed: not yet verified
A pinned run recipe exists: not yet verified
Compute requirements are estimated: not yet verified

Verified: Not verified

Evidence incomplete

Validate independently

Separate data and exposure records support an independent test.

Missing or unresolved evidence

No verified artifact manifest is linked to this exact record.
File checksums match the recorded files: not yet verified
Predictions are matched to the right samples: not yet verified
Score meaning and direction are confirmed: not yet verified
Independent validation data exist: not yet verified
Overlap with training data is checked: not yet verified

Verified: Not verified

Readiness describes the evidence in this release. Availability on your computer is checked separately when an investigation runs. Existing data exposure can prevent independent validation even when files are available.

Artifacts and reproduction

No verified artifact manifest is connected to this record yet. The gaps above identify what is needed before analysis can begin.

Subset and evaluation context

This record describes a particular subset or cohort used in an evaluation. Its results do not describe the full dataset.

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.

7 evidence rows matching the loaded filters

Claims, original sources and review scope · Release 2026-11-02-a7f03c9e55a5
Property and statement	Original source and location	Review and provenance
Limitations 1 values PrimateAI-3D is scored on 19,795 of these variants (Figure 1 footnote and Methods). Context-only references	Genomic heterogeneity inflates the performance of variant pathogenicity predictions (Lu, Liu, Lin and Brandes, bioRxiv 2025, v2) Original source ↗ Figure 1 (page 8), column 3, panel 4, title 'Stop gain (N=42,408, %P=98.13%)' Version: bioRxiv preprint version 2, posted 2025-09-10 (version 1 posted 2025-09-08); full-text PDF (21 pages) Retrieved: 2026-10-19T12:23:08Z	not individually reviewed No individual claim review recorded Audit details Field: `attributes.limitations` Source artifact SHA-256: `5b8e939257000811a6eaddd90d9de3e03bf1ac0e9015a38769df887ee01f9bf2` Hash scope: SHA-256 of one retrieval of the bioRxiv PDF. bioRxiv stamps the PDF on each download: a second retrieval at 2026-10-19T12:31:50Z had the same length (1,100,277 bytes), SHA-256 1a1aa1135ad8aa59bdb949bf8df75262eb6feba51b47f49439dbb819c426fbdd, differed in 32 bytes (ModDate) and had an identical pdftotext -layout text layer. Inspected artifact
Original printed counts N=42,408, %P=98.13% Context-only references	Genomic heterogeneity inflates the performance of variant pathogenicity predictions (Lu, Liu, Lin and Brandes, bioRxiv 2025, v2) Original source ↗ Figure 1 (page 8), column 3, panel 4, title 'Stop gain (N=42,408, %P=98.13%)' Version: bioRxiv preprint version 2, posted 2025-09-10 (version 1 posted 2025-09-08); full-text PDF (21 pages) Retrieved: 2026-10-19T12:23:08Z	not individually reviewed No individual claim review recorded Audit details Field: `attributes.original_printed_counts` Source artifact SHA-256: `5b8e939257000811a6eaddd90d9de3e03bf1ac0e9015a38769df887ee01f9bf2` Hash scope: SHA-256 of one retrieval of the bioRxiv PDF. bioRxiv stamps the PDF on each download: a second retrieval at 2026-10-19T12:31:50Z had the same length (1,100,277 bytes), SHA-256 1a1aa1135ad8aa59bdb949bf8df75262eb6feba51b47f49439dbb819c426fbdd, differed in 32 bytes (ModDate) and had an identical pdftotext -layout text layer. Inspected artifact
Source location Figure 1 (page 8), column 3, panel 4, title 'Stop gain (N=42,408, %P=98.13%)' Context-only references	Genomic heterogeneity inflates the performance of variant pathogenicity predictions (Lu, Liu, Lin and Brandes, bioRxiv 2025, v2) Original source ↗ Figure 1 (page 8), column 3, panel 4, title 'Stop gain (N=42,408, %P=98.13%)' Version: bioRxiv preprint version 2, posted 2025-09-10 (version 1 posted 2025-09-08); full-text PDF (21 pages) Retrieved: 2026-10-19T12:23:08Z	not individually reviewed No individual claim review recorded Audit details Field: `attributes.source_locator` Source artifact SHA-256: `5b8e939257000811a6eaddd90d9de3e03bf1ac0e9015a38769df887ee01f9bf2` Hash scope: SHA-256 of one retrieval of the bioRxiv PDF. bioRxiv stamps the PDF on each download: a second retrieval at 2026-10-19T12:31:50Z had the same length (1,100,277 bytes), SHA-256 1a1aa1135ad8aa59bdb949bf8df75262eb6feba51b47f49439dbb819c426fbdd, differed in 32 bytes (ModDate) and had an identical pdftotext -layout text layer. Inspected artifact
Test count 42408 Context-only references	Genomic heterogeneity inflates the performance of variant pathogenicity predictions (Lu, Liu, Lin and Brandes, bioRxiv 2025, v2) Original source ↗ Figure 1 (page 8), column 3, panel 4, title 'Stop gain (N=42,408, %P=98.13%)' Version: bioRxiv preprint version 2, posted 2025-09-10 (version 1 posted 2025-09-08); full-text PDF (21 pages) Retrieved: 2026-10-19T12:23:08Z	not individually reviewed No individual claim review recorded Audit details Field: `attributes.test_count` Source artifact SHA-256: `5b8e939257000811a6eaddd90d9de3e03bf1ac0e9015a38769df887ee01f9bf2` Hash scope: SHA-256 of one retrieval of the bioRxiv PDF. bioRxiv stamps the PDF on each download: a second retrieval at 2026-10-19T12:31:50Z had the same length (1,100,277 bytes), SHA-256 1a1aa1135ad8aa59bdb949bf8df75262eb6feba51b47f49439dbb819c426fbdd, differed in 32 bytes (ModDate) and had an identical pdftotext -layout text layer. Inspected artifact
Description ClinVar SNVs in the 'Stop gain' group of the Lu et al. 2025 benchmark, restricted to variants scored by every model in the panel. Context-only references	Genomic heterogeneity inflates the performance of variant pathogenicity predictions (Lu, Liu, Lin and Brandes, bioRxiv 2025, v2) Original source ↗ Figure 1 (page 8), column 3, panel 4, title 'Stop gain (N=42,408, %P=98.13%)' Version: bioRxiv preprint version 2, posted 2025-09-10 (version 1 posted 2025-09-08); full-text PDF (21 pages) Retrieved: 2026-10-19T12:23:08Z	not individually reviewed No individual claim review recorded Audit details Field: `description` Source artifact SHA-256: `5b8e939257000811a6eaddd90d9de3e03bf1ac0e9015a38769df887ee01f9bf2` Hash scope: SHA-256 of one retrieval of the bioRxiv PDF. bioRxiv stamps the PDF on each download: a second retrieval at 2026-10-19T12:31:50Z had the same length (1,100,277 bytes), SHA-256 1a1aa1135ad8aa59bdb949bf8df75262eb6feba51b47f49439dbb819c426fbdd, differed in 32 bytes (ModDate) and had an identical pdftotext -layout text layer. Inspected artifact
Relationship: used in lu-hetero-20261019-protocol-clinvar-variant-type-auroc Context-only references	Genomic heterogeneity inflates the performance of variant pathogenicity predictions (Lu, Liu, Lin and Brandes, bioRxiv 2025, v2) Original source ↗ Figure 1 (page 8), column 3, panel 4, title 'Stop gain (N=42,408, %P=98.13%)' Version: bioRxiv preprint version 2, posted 2025-09-10 (version 1 posted 2025-09-08); full-text PDF (21 pages) Retrieved: 2026-10-19T12:23:08Z	not individually reviewed No individual claim review recorded Audit details Field: `links:used_in:lu-hetero-20261019-protocol-clinvar-variant-type-auroc` Source artifact SHA-256: `5b8e939257000811a6eaddd90d9de3e03bf1ac0e9015a38769df887ee01f9bf2` Hash scope: SHA-256 of one retrieval of the bioRxiv PDF. bioRxiv stamps the PDF on each download: a second retrieval at 2026-10-19T12:31:50Z had the same length (1,100,277 bytes), SHA-256 1a1aa1135ad8aa59bdb949bf8df75262eb6feba51b47f49439dbb819c426fbdd, differed in 32 bytes (ModDate) and had an identical pdftotext -layout text layer. Inspected artifact
Name Lu et al. 2025 ClinVar benchmark: Stop gain (Figure 1) Context-only references	Genomic heterogeneity inflates the performance of variant pathogenicity predictions (Lu, Liu, Lin and Brandes, bioRxiv 2025, v2) Original source ↗ Figure 1 (page 8), column 3, panel 4, title 'Stop gain (N=42,408, %P=98.13%)' Version: bioRxiv preprint version 2, posted 2025-09-10 (version 1 posted 2025-09-08); full-text PDF (21 pages) Retrieved: 2026-10-19T12:23:08Z	not individually reviewed No individual claim review recorded Audit details Field: `name` Source artifact SHA-256: `5b8e939257000811a6eaddd90d9de3e03bf1ac0e9015a38769df887ee01f9bf2` Hash scope: SHA-256 of one retrieval of the bioRxiv PDF. bioRxiv stamps the PDF on each download: a second retrieval at 2026-10-19T12:31:50Z had the same length (1,100,277 bytes), SHA-256 1a1aa1135ad8aa59bdb949bf8df75262eb6feba51b47f49439dbb819c426fbdd, differed in 32 bytes (ModDate) and had an identical pdftotext -layout text layer. Inspected artifact

How to read the evidence

Sources and history

Release 2026-11-02-a7f03c9e55a5 · Record review: source checked

1 source record and release history

Genomic heterogeneity inflates the performance of variant pathogenicity predictions (Lu, Liu, Lin and Brandes, bioRxiv 2025, v2) · Original source · bioRxiv preprint version 2, posted 2025-09-10 (version 1 posted 2025-09-08); full-text PDF (21 pages)

Download this release (gzip)

Technical metadata and extraction receipts

Stable ID: lu-hetero-20261019-subset-stop-gain

areas: dna-genomes
contexts: clinical_research
test count: 42408
original printed counts: N=42,408, %P=98.13%
source locator: Figure 1 (page 8), column 3, panel 4, title 'Stop gain (N=42,408, %P=98.13%)'
limitations: PrimateAI-3D is scored on 19,795 of these variants (Figure 1 footnote and Methods).

Related records

Suggest a correction