Source

GPN-MSA: an alignment-based DNA language model for genome-wide variant effect prediction (bioRxiv v2)

bioRxiv · 2024 · bioRxiv v2, posted 2024-04-06 · Preprint

DOI: 10.1101/2023.10.10.561776 Original source

What this source contributed

Records in this release that cite this source. A record can cite several sources.

1 model
2 methods
17 configurations
1 benchmark
5 protocols
5 datasets
70 evaluations
138 results
11 evidence claims

Results

Values as the source printed them. Compare results only under compatible protocols, datasets and metrics.

Value	Metric	Model or configuration	Dataset	Evaluation
0.965 unitless	AUROC (maximum over three random seeds; zero-shot)	GPN-MSA ablation '51 mammals' (Supplementary Table 4)	ClinVar Pathogenic vs gnomAD common missense variants (GPN-MSA paper)	51 mammals on ClinVar (GPN-MSA Supplementary Table 4)
0.964 unitless	AUROC (mean over three random seeds; zero-shot)	GPN-MSA ablation '51 mammals' (Supplementary Table 4)	ClinVar Pathogenic vs gnomAD common missense variants (GPN-MSA paper)	51 mammals on ClinVar (GPN-MSA Supplementary Table 4)
0.325 unitless	AUPRC (maximum over three random seeds; zero-shot)	GPN-MSA ablation '51 mammals' (Supplementary Table 4)	COSMIC frequent vs gnomAD common missense variants (GPN-MSA paper)	51 mammals on COSMIC (GPN-MSA Supplementary Table 4)
0.323 unitless	AUPRC (mean over three random seeds; zero-shot)	GPN-MSA ablation '51 mammals' (Supplementary Table 4)	COSMIC frequent vs gnomAD common missense variants (GPN-MSA paper)	51 mammals on COSMIC (GPN-MSA Supplementary Table 4)
43.408 unitless	Odds ratio (rare (singletons) versus common (MAF > 5%) gnomAD variants in the tail of deleterious scores, threshold at 30 false discoveries, one-sided Fisher's exact test; all variant types, rare subsampled to the common count; maximum over three random seeds; zero-shot)	GPN-MSA ablation '51 mammals' (Supplementary Table 4)	gnomAD rare (singletons) vs common variants, ablation set (GPN-MSA paper)	51 mammals on gnomAD (GPN-MSA Supplementary Table 4)
41.381 unitless	Odds ratio (rare (singletons) versus common (MAF > 5%) gnomAD variants in the tail of deleterious scores, threshold at 30 false discoveries, one-sided Fisher's exact test; all variant types, rare subsampled to the common count; mean over three random seeds; zero-shot)	GPN-MSA ablation '51 mammals' (Supplementary Table 4)	gnomAD rare (singletons) vs common variants, ablation set (GPN-MSA paper)	51 mammals on gnomAD (GPN-MSA Supplementary Table 4)
0.076 unitless	AUPRC (maximum over three random seeds; zero-shot)	GPN-MSA ablation '51 mammals' (Supplementary Table 4)	OMIM pathogenic vs gnomAD common regulatory variants (GPN-MSA paper)	51 mammals on OMIM (GPN-MSA Supplementary Table 4)
0.073 unitless	AUPRC (mean over three random seeds; zero-shot)	GPN-MSA ablation '51 mammals' (Supplementary Table 4)	OMIM pathogenic vs gnomAD common regulatory variants (GPN-MSA paper)	51 mammals on OMIM (GPN-MSA Supplementary Table 4)
0.967 unitless	AUROC (maximum over three random seeds; zero-shot)	GPN-MSA ablation '51 vertebrates' (Supplementary Table 4)	ClinVar Pathogenic vs gnomAD common missense variants (GPN-MSA paper)	51 vertebrates on ClinVar (GPN-MSA Supplementary Table 4)
0.967 unitless	AUROC (mean over three random seeds; zero-shot)	GPN-MSA ablation '51 vertebrates' (Supplementary Table 4)	ClinVar Pathogenic vs gnomAD common missense variants (GPN-MSA paper)	51 vertebrates on ClinVar (GPN-MSA Supplementary Table 4)
0.347 unitless	AUPRC (maximum over three random seeds; zero-shot)	GPN-MSA ablation '51 vertebrates' (Supplementary Table 4)	COSMIC frequent vs gnomAD common missense variants (GPN-MSA paper)	51 vertebrates on COSMIC (GPN-MSA Supplementary Table 4)
0.342 unitless	AUPRC (mean over three random seeds; zero-shot)	GPN-MSA ablation '51 vertebrates' (Supplementary Table 4)	COSMIC frequent vs gnomAD common missense variants (GPN-MSA paper)	51 vertebrates on COSMIC (GPN-MSA Supplementary Table 4)
46.792 unitless	Odds ratio (rare (singletons) versus common (MAF > 5%) gnomAD variants in the tail of deleterious scores, threshold at 30 false discoveries, one-sided Fisher's exact test; all variant types, rare subsampled to the common count; maximum over three random seeds; zero-shot)	GPN-MSA ablation '51 vertebrates' (Supplementary Table 4)	gnomAD rare (singletons) vs common variants, ablation set (GPN-MSA paper)	51 vertebrates on gnomAD (GPN-MSA Supplementary Table 4)
41.624 unitless	Odds ratio (rare (singletons) versus common (MAF > 5%) gnomAD variants in the tail of deleterious scores, threshold at 30 false discoveries, one-sided Fisher's exact test; all variant types, rare subsampled to the common count; mean over three random seeds; zero-shot)	GPN-MSA ablation '51 vertebrates' (Supplementary Table 4)	gnomAD rare (singletons) vs common variants, ablation set (GPN-MSA paper)	51 vertebrates on gnomAD (GPN-MSA Supplementary Table 4)
0.108 unitless	AUPRC (maximum over three random seeds; zero-shot)	GPN-MSA ablation '51 vertebrates' (Supplementary Table 4)	OMIM pathogenic vs gnomAD common regulatory variants (GPN-MSA paper)	51 vertebrates on OMIM (GPN-MSA Supplementary Table 4)
0.101 unitless	AUPRC (mean over three random seeds; zero-shot)	GPN-MSA ablation '51 vertebrates' (Supplementary Table 4)	OMIM pathogenic vs gnomAD common regulatory variants (GPN-MSA paper)	51 vertebrates on OMIM (GPN-MSA Supplementary Table 4)
0.928 unitless	AUROC (maximum over three random seeds; zero-shot)	Combined phyloP and phastCons	ClinVar Pathogenic vs gnomAD common missense variants (GPN-MSA paper)	Combined phyloP and phastCons on ClinVar (GPN-MSA Supplementary Table 4)
0.928 unitless	AUROC (mean over three random seeds; zero-shot)	Combined phyloP and phastCons	ClinVar Pathogenic vs gnomAD common missense variants (GPN-MSA paper)	Combined phyloP and phastCons on ClinVar (GPN-MSA Supplementary Table 4)
0.141 unitless	AUPRC (maximum over three random seeds; zero-shot)	Combined phyloP and phastCons	COSMIC frequent vs gnomAD common missense variants (GPN-MSA paper)	Combined phyloP and phastCons on COSMIC (GPN-MSA Supplementary Table 4)
0.141 unitless	AUPRC (mean over three random seeds; zero-shot)	Combined phyloP and phastCons	COSMIC frequent vs gnomAD common missense variants (GPN-MSA paper)	Combined phyloP and phastCons on COSMIC (GPN-MSA Supplementary Table 4)
13.919 unitless	Odds ratio (rare (singletons) versus common (MAF > 5%) gnomAD variants in the tail of deleterious scores, threshold at 30 false discoveries, one-sided Fisher's exact test; all variant types, rare subsampled to the common count; maximum over three random seeds; zero-shot)	Combined phyloP and phastCons	gnomAD rare (singletons) vs common variants, ablation set (GPN-MSA paper)	Combined phyloP and phastCons on gnomAD (GPN-MSA Supplementary Table 4)
13.919 unitless	Odds ratio (rare (singletons) versus common (MAF > 5%) gnomAD variants in the tail of deleterious scores, threshold at 30 false discoveries, one-sided Fisher's exact test; all variant types, rare subsampled to the common count; mean over three random seeds; zero-shot)	Combined phyloP and phastCons	gnomAD rare (singletons) vs common variants, ablation set (GPN-MSA paper)	Combined phyloP and phastCons on gnomAD (GPN-MSA Supplementary Table 4)
0.039 unitless	AUPRC (maximum over three random seeds; zero-shot)	Combined phyloP and phastCons	OMIM pathogenic vs gnomAD common regulatory variants (GPN-MSA paper)	Combined phyloP and phastCons on OMIM (GPN-MSA Supplementary Table 4)
0.039 unitless	AUPRC (mean over three random seeds; zero-shot)	Combined phyloP and phastCons	OMIM pathogenic vs gnomAD common regulatory variants (GPN-MSA paper)	Combined phyloP and phastCons on OMIM (GPN-MSA Supplementary Table 4)
0.969 unitless	AUROC (maximum over three random seeds; zero-shot)	GPN-MSA ablation 'Default' (Supplementary Table 4)	ClinVar Pathogenic vs gnomAD common missense variants (GPN-MSA paper)	Default on ClinVar (GPN-MSA Supplementary Table 4)
0.969 unitless	AUROC (mean over three random seeds; zero-shot)	GPN-MSA ablation 'Default' (Supplementary Table 4)	ClinVar Pathogenic vs gnomAD common missense variants (GPN-MSA paper)	Default on ClinVar (GPN-MSA Supplementary Table 4)
0.355 unitless	AUPRC (maximum over three random seeds; zero-shot)	GPN-MSA ablation 'Default' (Supplementary Table 4)	COSMIC frequent vs gnomAD common missense variants (GPN-MSA paper)	Default on COSMIC (GPN-MSA Supplementary Table 4)
0.348 unitless	AUPRC (mean over three random seeds; zero-shot)	GPN-MSA ablation 'Default' (Supplementary Table 4)	COSMIC frequent vs gnomAD common missense variants (GPN-MSA paper)	Default on COSMIC (GPN-MSA Supplementary Table 4)
43.118 unitless	Odds ratio (rare (singletons) versus common (MAF > 5%) gnomAD variants in the tail of deleterious scores, threshold at 30 false discoveries, one-sided Fisher's exact test; all variant types, rare subsampled to the common count; maximum over three random seeds; zero-shot)	GPN-MSA ablation 'Default' (Supplementary Table 4)	gnomAD rare (singletons) vs common variants, ablation set (GPN-MSA paper)	Default on gnomAD (GPN-MSA Supplementary Table 4)
41.377 unitless	Odds ratio (rare (singletons) versus common (MAF > 5%) gnomAD variants in the tail of deleterious scores, threshold at 30 false discoveries, one-sided Fisher's exact test; all variant types, rare subsampled to the common count; mean over three random seeds; zero-shot)	GPN-MSA ablation 'Default' (Supplementary Table 4)	gnomAD rare (singletons) vs common variants, ablation set (GPN-MSA paper)	Default on gnomAD (GPN-MSA Supplementary Table 4)
0.133 unitless	AUPRC (maximum over three random seeds; zero-shot)	GPN-MSA ablation 'Default' (Supplementary Table 4)	OMIM pathogenic vs gnomAD common regulatory variants (GPN-MSA paper)	Default on OMIM (GPN-MSA Supplementary Table 4)
0.124 unitless	AUPRC (mean over three random seeds; zero-shot)	GPN-MSA ablation 'Default' (Supplementary Table 4)	OMIM pathogenic vs gnomAD common regulatory variants (GPN-MSA paper)	Default on OMIM (GPN-MSA Supplementary Table 4)
0.967 unitless	AUROC (maximum over three random seeds; zero-shot)	GPN-MSA ablation 'Don’t replace non-conserved' (Supplementary Table 4)	ClinVar Pathogenic vs gnomAD common missense variants (GPN-MSA paper)	Don’t replace non-conserved on ClinVar (GPN-MSA Supplementary Table 4)
0.966 unitless	AUROC (mean over three random seeds; zero-shot)	GPN-MSA ablation 'Don’t replace non-conserved' (Supplementary Table 4)	ClinVar Pathogenic vs gnomAD common missense variants (GPN-MSA paper)	Don’t replace non-conserved on ClinVar (GPN-MSA Supplementary Table 4)
0.340 unitless	AUPRC (maximum over three random seeds; zero-shot)	GPN-MSA ablation 'Don’t replace non-conserved' (Supplementary Table 4)	COSMIC frequent vs gnomAD common missense variants (GPN-MSA paper)	Don’t replace non-conserved on COSMIC (GPN-MSA Supplementary Table 4)
0.333 unitless	AUPRC (mean over three random seeds; zero-shot)	GPN-MSA ablation 'Don’t replace non-conserved' (Supplementary Table 4)	COSMIC frequent vs gnomAD common missense variants (GPN-MSA paper)	Don’t replace non-conserved on COSMIC (GPN-MSA Supplementary Table 4)
45.213 unitless	Odds ratio (rare (singletons) versus common (MAF > 5%) gnomAD variants in the tail of deleterious scores, threshold at 30 false discoveries, one-sided Fisher's exact test; all variant types, rare subsampled to the common count; maximum over three random seeds; zero-shot)	GPN-MSA ablation 'Don’t replace non-conserved' (Supplementary Table 4)	gnomAD rare (singletons) vs common variants, ablation set (GPN-MSA paper)	Don’t replace non-conserved on gnomAD (GPN-MSA Supplementary Table 4)
41.853 unitless	Odds ratio (rare (singletons) versus common (MAF > 5%) gnomAD variants in the tail of deleterious scores, threshold at 30 false discoveries, one-sided Fisher's exact test; all variant types, rare subsampled to the common count; mean over three random seeds; zero-shot)	GPN-MSA ablation 'Don’t replace non-conserved' (Supplementary Table 4)	gnomAD rare (singletons) vs common variants, ablation set (GPN-MSA paper)	Don’t replace non-conserved on gnomAD (GPN-MSA Supplementary Table 4)
0.112 unitless	AUPRC (maximum over three random seeds; zero-shot)	GPN-MSA ablation 'Don’t replace non-conserved' (Supplementary Table 4)	OMIM pathogenic vs gnomAD common regulatory variants (GPN-MSA paper)	Don’t replace non-conserved on OMIM (GPN-MSA Supplementary Table 4)
0.111 unitless	AUPRC (mean over three random seeds; zero-shot)	GPN-MSA ablation 'Don’t replace non-conserved' (Supplementary Table 4)	OMIM pathogenic vs gnomAD common regulatory variants (GPN-MSA paper)	Don’t replace non-conserved on OMIM (GPN-MSA Supplementary Table 4)
0.967 unitless	AUROC (maximum over three random seeds; zero-shot)	GPN-MSA ablation 'Don’t upweight conserved' (Supplementary Table 4)	ClinVar Pathogenic vs gnomAD common missense variants (GPN-MSA paper)	Don’t upweight conserved on ClinVar (GPN-MSA Supplementary Table 4)
0.967 unitless	AUROC (mean over three random seeds; zero-shot)	GPN-MSA ablation 'Don’t upweight conserved' (Supplementary Table 4)	ClinVar Pathogenic vs gnomAD common missense variants (GPN-MSA paper)	Don’t upweight conserved on ClinVar (GPN-MSA Supplementary Table 4)
0.309 unitless	AUPRC (maximum over three random seeds; zero-shot)	GPN-MSA ablation 'Don’t upweight conserved' (Supplementary Table 4)	COSMIC frequent vs gnomAD common missense variants (GPN-MSA paper)	Don’t upweight conserved on COSMIC (GPN-MSA Supplementary Table 4)
0.303 unitless	AUPRC (mean over three random seeds; zero-shot)	GPN-MSA ablation 'Don’t upweight conserved' (Supplementary Table 4)	COSMIC frequent vs gnomAD common missense variants (GPN-MSA paper)	Don’t upweight conserved on COSMIC (GPN-MSA Supplementary Table 4)
39.991 unitless	Odds ratio (rare (singletons) versus common (MAF > 5%) gnomAD variants in the tail of deleterious scores, threshold at 30 false discoveries, one-sided Fisher's exact test; all variant types, rare subsampled to the common count; maximum over three random seeds; zero-shot)	GPN-MSA ablation 'Don’t upweight conserved' (Supplementary Table 4)	gnomAD rare (singletons) vs common variants, ablation set (GPN-MSA paper)	Don’t upweight conserved on gnomAD (GPN-MSA Supplementary Table 4)
37.094 unitless	Odds ratio (rare (singletons) versus common (MAF > 5%) gnomAD variants in the tail of deleterious scores, threshold at 30 false discoveries, one-sided Fisher's exact test; all variant types, rare subsampled to the common count; mean over three random seeds; zero-shot)	GPN-MSA ablation 'Don’t upweight conserved' (Supplementary Table 4)	gnomAD rare (singletons) vs common variants, ablation set (GPN-MSA paper)	Don’t upweight conserved on gnomAD (GPN-MSA Supplementary Table 4)
0.152 unitless	AUPRC (maximum over three random seeds; zero-shot)	GPN-MSA ablation 'Don’t upweight conserved' (Supplementary Table 4)	OMIM pathogenic vs gnomAD common regulatory variants (GPN-MSA paper)	Don’t upweight conserved on OMIM (GPN-MSA Supplementary Table 4)
0.133 unitless	AUPRC (mean over three random seeds; zero-shot)	GPN-MSA ablation 'Don’t upweight conserved' (Supplementary Table 4)	OMIM pathogenic vs gnomAD common regulatory variants (GPN-MSA paper)	Don’t upweight conserved on OMIM (GPN-MSA Supplementary Table 4)
0.75 unitless	AUROC (binary labels from absolute relative toxicity > 0.125; zero-shot)	ESM-1b, precomputed log-likelihood ratios (GPN-MSA paper)	ProteinGym v0.1 DMS TADBP_HUMAN Bolognesi 2019, binarised (GPN-MSA paper)	ESM-1b on ProteinGym DMS TADBP binary classification (GPN-MSA paper)
0.83 unitless	AUROC (binary labels from absolute relative toxicity > 0.125; zero-shot)	GPN-MSA, main results model (bioRxiv v2)	ProteinGym v0.1 DMS TADBP_HUMAN Bolognesi 2019, binarised (GPN-MSA paper)	GPN-MSA on ProteinGym DMS TADBP binary classification (GPN-MSA paper)

Claims

Statements this source supports, each with where in the source it was checked.

GPN-MSA (evaluation mode not printed (Table 4 has no Eval. column), GUANinE v1.1): configuration of GPN-MSA: Configuration of GPN-MSA (Configuration name as printed in its source (GPN-MSA); GPN-MSA identity from gfm-guides-20261014-gpnmsa-source-biorxiv-v2 (Abstract) and gfm-guides-20261014-gpnmsa-source-gpn-msa-doc)
GPN-MSA (Evo 2 paper): configuration of GPN-MSA: Configuration of GPN-MSA (Supplementary Data 1 column 'Model' ('GPN-MSA'); GPN-MSA identity from gfm-guides-20261014-gpnmsa-source-biorxiv-v2 (Abstract))
GPN-MSA (Position) (zero-shot, GUANinE v1.1): configuration of GPN-MSA: Configuration of GPN-MSA (Configuration name as printed in its source (GPN-MSA (Position)); GPN-MSA identity from gfm-guides-20261014-gpnmsa-source-biorxiv-v2 (Abstract) and gfm-guides-20261014-gpnmsa-source-gpn-msa-doc)
GPN-MSA (zero-shot, GUANinE v1.1): configuration of GPN-MSA: Configuration of GPN-MSA (Configuration name as printed in its source (GPN-MSA); GPN-MSA identity from gfm-guides-20261014-gpnmsa-source-biorxiv-v2 (Abstract) and gfm-guides-20261014-gpnmsa-source-gpn-msa-doc)
GPN-MSA_absLLR.plus.score (TraitGym): configuration of GPN-MSA: Configuration of GPN-MSA (Configuration name as printed in its source (GPN-MSA_absLLR.plus.score); GPN-MSA identity from gfm-guides-20261014-gpnmsa-source-biorxiv-v2 (Abstract) and gfm-guides-20261014-gpnmsa-source-gpn-msa-doc)
GPN-MSA_Embeddings.minus.inner_product (TraitGym): configuration of GPN-MSA: Configuration of GPN-MSA (Configuration name as printed in its source (GPN-MSA_Embeddings.minus.inner_product); GPN-MSA identity from gfm-guides-20261014-gpnmsa-source-biorxiv-v2 (Abstract) and gfm-guides-20261014-gpnmsa-source-gpn-msa-doc)
GPN-MSA_Embeddings.plus.cosine_distance (TraitGym): configuration of GPN-MSA: Configuration of GPN-MSA (Configuration name as printed in its source (GPN-MSA_Embeddings.plus.cosine_distance); GPN-MSA identity from gfm-guides-20261014-gpnmsa-source-biorxiv-v2 (Abstract) and gfm-guides-20261014-gpnmsa-source-gpn-msa-doc)
GPN-MSA_Embeddings.plus.euclidean_distance (TraitGym): configuration of GPN-MSA: Configuration of GPN-MSA (Configuration name as printed in its source (GPN-MSA_Embeddings.plus.euclidean_distance); GPN-MSA identity from gfm-guides-20261014-gpnmsa-source-biorxiv-v2 (Abstract) and gfm-guides-20261014-gpnmsa-source-gpn-msa-doc)
GPN-MSA_InnerProduct.minus.score (TraitGym): configuration of GPN-MSA: Configuration of GPN-MSA (Configuration name as printed in its source (GPN-MSA_InnerProduct.minus.score); GPN-MSA identity from gfm-guides-20261014-gpnmsa-source-biorxiv-v2 (Abstract) and gfm-guides-20261014-gpnmsa-source-gpn-msa-doc)
GPN-MSA_LLR.minus.score (TraitGym): configuration of GPN-MSA: Configuration of GPN-MSA (Configuration name as printed in its source (GPN-MSA_LLR.minus.score); GPN-MSA identity from gfm-guides-20261014-gpnmsa-source-biorxiv-v2 (Abstract) and gfm-guides-20261014-gpnmsa-source-gpn-msa-doc)
GPN-MSA.LogisticRegression.chrom (TraitGym): configuration of GPN-MSA: Configuration of GPN-MSA (Configuration name as printed in its source (GPN-MSA.LogisticRegression.chrom); GPN-MSA identity from gfm-guides-20261014-gpnmsa-source-biorxiv-v2 (Abstract) and gfm-guides-20261014-gpnmsa-source-gpn-msa-doc)

Models, benchmarks and other records

Models (1)

GPN-MSA

Methods (2)

Configurations (17)

Benchmarks (1)

GPN-MSA variant effect prediction evaluation (Benegas et al.)

Protocols (5)

Datasets (5)

Evaluations (70)

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.

30 evidence rows matching the loaded filters

Claims, original sources and review scope · Release 2026-10-22-1b7cf027e4f6
Property and statement	Original source and location	Review and provenance
attributes.artifact_sha256 10411077df1ae59614f9a9a9f85384103d54127ba67a9b3c469a4a2d6c33cee4 Source metadata	GPN-MSA: an alignment-based DNA language model for genome-wide variant effect prediction (bioRxiv v2) Original source ↗ No field-specific location recorded Version: bioRxiv v2, posted 2024-04-06 Retrieved: 2026-10-14T12:02:16Z	catalogued No individual claim review recorded Audit details Field: `attributes.artifact_sha256` Source artifact SHA-256: `10411077df1ae59614f9a9a9f85384103d54127ba67a9b3c469a4a2d6c33cee4` Hash scope: SHA-256 of the PDF bytes as downloaded 2026-10-14T12:02:16Z. bioRxiv rewrites the XMP dates and document IDs, the Info ModDate and the trailer /ID on download, so a re-download does not reproduce the hash; the review found a fresh copy of the same length differing only in those 51 bytes, with an identical pdftotext -layout text layer. Inspected artifact
Artifact url https://www.biorxiv.org/content/10.1101/2023.10.10.561776v2.full.pdf Source metadata	GPN-MSA: an alignment-based DNA language model for genome-wide variant effect prediction (bioRxiv v2) Original source ↗ No field-specific location recorded Version: bioRxiv v2, posted 2024-04-06 Retrieved: 2026-10-14T12:02:16Z	catalogued No individual claim review recorded Audit details Field: `attributes.artifact_url` Source artifact SHA-256: `10411077df1ae59614f9a9a9f85384103d54127ba67a9b3c469a4a2d6c33cee4` Hash scope: SHA-256 of the PDF bytes as downloaded 2026-10-14T12:02:16Z. bioRxiv rewrites the XMP dates and document IDs, the Info ModDate and the trailer /ID on download, so a re-download does not reproduce the hash; the review found a fresh copy of the same length differing only in those 51 bytes, with an identical pdftotext -layout text layer. Inspected artifact
Doi 10.1101/2023.10.10.561776 Source metadata	GPN-MSA: an alignment-based DNA language model for genome-wide variant effect prediction (bioRxiv v2) Original source ↗ No field-specific location recorded Version: bioRxiv v2, posted 2024-04-06 Retrieved: 2026-10-14T12:02:16Z	catalogued No individual claim review recorded Audit details Field: `attributes.doi` Source artifact SHA-256: `10411077df1ae59614f9a9a9f85384103d54127ba67a9b3c469a4a2d6c33cee4` Hash scope: SHA-256 of the PDF bytes as downloaded 2026-10-14T12:02:16Z. bioRxiv rewrites the XMP dates and document IDs, the Info ModDate and the trailer /ID on download, so a re-download does not reproduce the hash; the review found a fresh copy of the same length differing only in those 51 bytes, with an identical pdftotext -layout text layer. Inspected artifact
Hash scope SHA-256 of the PDF bytes as downloaded 2026-10-14T12:02:16Z. bioRxiv rewrites the XMP dates and document IDs, the Info ModDate and the trailer /ID on download, so a re-download does not reproduce the hash; the review found a fresh copy of the same length differing only in those 51 bytes, with an identical pdftotext -layout text layer. Source metadata	GPN-MSA: an alignment-based DNA language model for genome-wide variant effect prediction (bioRxiv v2) Original source ↗ No field-specific location recorded Version: bioRxiv v2, posted 2024-04-06 Retrieved: 2026-10-14T12:02:16Z	catalogued No individual claim review recorded Audit details Field: `attributes.hash_scope` Source artifact SHA-256: `10411077df1ae59614f9a9a9f85384103d54127ba67a9b3c469a4a2d6c33cee4` Hash scope: SHA-256 of the PDF bytes as downloaded 2026-10-14T12:02:16Z. bioRxiv rewrites the XMP dates and document IDs, the Info ModDate and the trailer /ID on download, so a re-download does not reproduce the hash; the review found a fresh copy of the same length differing only in those 51 bytes, with an identical pdftotext -layout text layer. Inspected artifact
Licence CC BY-NC-ND 4.0 Source metadata	GPN-MSA: an alignment-based DNA language model for genome-wide variant effect prediction (bioRxiv v2) Original source ↗ No field-specific location recorded Version: bioRxiv v2, posted 2024-04-06 Retrieved: 2026-10-14T12:02:16Z	catalogued No individual claim review recorded Audit details Field: `attributes.licence` Source artifact SHA-256: `10411077df1ae59614f9a9a9f85384103d54127ba67a9b3c469a4a2d6c33cee4` Hash scope: SHA-256 of the PDF bytes as downloaded 2026-10-14T12:02:16Z. bioRxiv rewrites the XMP dates and document IDs, the Info ModDate and the trailer /ID on download, so a re-download does not reproduce the hash; the review found a fresh copy of the same length differing only in those 51 bytes, with an identical pdftotext -layout text layer. Inspected artifact
Limitations 5 values Preprint. The peer-reviewed version is Benegas et al., Nature Biotechnology 43, 1960-1965 (2025), doi:10.1038/s41587-024-02511-w, which is not open access; its Extended Data Table 1 (Ablation study) is published only as an image and was not read, so the printed values here may differ from the journal version. The issue named bioRxiv 2023.10.10.562811; that DOI does not resolve. bioRxiv and Crossref give 10.1101/2023.10.10.561776 for this preprint, and the bioRxiv API lists the journal DOI 10.1038/s41587-024-02511-w as its published version. The main comparisons with CADD, phyloP, ESM-1b, Nucleotide Transformer and HyenaDNA on ClinVar, COSMIC, OMIM and gnomAD (Figure 2c-f, Supplementary Figures 3-5) are printed only as figures and were not extracted. Supplementary Table 4 caption says ClinVar, COSMIC, OMIM and gnomAD use 'same setting as Figure 2a', '2b', '2c' and '2d', but in v2 those settings are Figure 2c, 2d, 2e and 2f (Figure 2 caption and Results). The settings were taken from the Methods 'Main VEP metrics' and the caption's own description. Licence CC BY-NC-ND 4.0; no archived copy is kept in the batch folder. Source metadata	GPN-MSA: an alignment-based DNA language model for genome-wide variant effect prediction (bioRxiv v2) Original source ↗ No field-specific location recorded Version: bioRxiv v2, posted 2024-04-06 Retrieved: 2026-10-14T12:02:16Z	catalogued No individual claim review recorded Audit details Field: `attributes.limitations` Source artifact SHA-256: `10411077df1ae59614f9a9a9f85384103d54127ba67a9b3c469a4a2d6c33cee4` Hash scope: SHA-256 of the PDF bytes as downloaded 2026-10-14T12:02:16Z. bioRxiv rewrites the XMP dates and document IDs, the Info ModDate and the trailer /ID on download, so a re-download does not reproduce the hash; the review found a fresh copy of the same length differing only in those 51 bytes, with an identical pdftotext -layout text layer. Inspected artifact
Media type application/pdf Source metadata	GPN-MSA: an alignment-based DNA language model for genome-wide variant effect prediction (bioRxiv v2) Original source ↗ No field-specific location recorded Version: bioRxiv v2, posted 2024-04-06 Retrieved: 2026-10-14T12:02:16Z	catalogued No individual claim review recorded Audit details Field: `attributes.media_type` Source artifact SHA-256: `10411077df1ae59614f9a9a9f85384103d54127ba67a9b3c469a4a2d6c33cee4` Hash scope: SHA-256 of the PDF bytes as downloaded 2026-10-14T12:02:16Z. bioRxiv rewrites the XMP dates and document IDs, the Info ModDate and the trailer /ID on download, so a re-download does not reproduce the hash; the review found a fresh copy of the same length differing only in those 51 bytes, with an identical pdftotext -layout text layer. Inspected artifact
Publication status preprint Source metadata	GPN-MSA: an alignment-based DNA language model for genome-wide variant effect prediction (bioRxiv v2) Original source ↗ No field-specific location recorded Version: bioRxiv v2, posted 2024-04-06 Retrieved: 2026-10-14T12:02:16Z	catalogued No individual claim review recorded Audit details Field: `attributes.publication_status` Source artifact SHA-256: `10411077df1ae59614f9a9a9f85384103d54127ba67a9b3c469a4a2d6c33cee4` Hash scope: SHA-256 of the PDF bytes as downloaded 2026-10-14T12:02:16Z. bioRxiv rewrites the XMP dates and document IDs, the Info ModDate and the trailer /ID on download, so a re-download does not reproduce the hash; the review found a fresh copy of the same length differing only in those 51 bytes, with an identical pdftotext -layout text layer. Inspected artifact
Retrieved at 2026-10-14T12:02:16Z Source metadata	GPN-MSA: an alignment-based DNA language model for genome-wide variant effect prediction (bioRxiv v2) Original source ↗ No field-specific location recorded Version: bioRxiv v2, posted 2024-04-06 Retrieved: 2026-10-14T12:02:16Z	catalogued No individual claim review recorded Audit details Field: `attributes.retrieved_at` Source artifact SHA-256: `10411077df1ae59614f9a9a9f85384103d54127ba67a9b3c469a4a2d6c33cee4` Hash scope: SHA-256 of the PDF bytes as downloaded 2026-10-14T12:02:16Z. bioRxiv rewrites the XMP dates and document IDs, the Info ModDate and the trailer /ID on download, so a re-download does not reproduce the hash; the review found a fresh copy of the same length differing only in those 51 bytes, with an identical pdftotext -layout text layer. Inspected artifact
attributes.review.artifact_sha256 10411077df1ae59614f9a9a9f85384103d54127ba67a9b3c469a4a2d6c33cee4 Source metadata	GPN-MSA: an alignment-based DNA language model for genome-wide variant effect prediction (bioRxiv v2) Original source ↗ No field-specific location recorded Version: bioRxiv v2, posted 2024-04-06 Retrieved: 2026-10-14T12:02:16Z	catalogued No individual claim review recorded Audit details Field: `attributes.review.artifact_sha256` Source artifact SHA-256: `10411077df1ae59614f9a9a9f85384103d54127ba67a9b3c469a4a2d6c33cee4` Hash scope: SHA-256 of the PDF bytes as downloaded 2026-10-14T12:02:16Z. bioRxiv rewrites the XMP dates and document IDs, the Info ModDate and the trailer /ID on download, so a re-download does not reproduce the hash; the review found a fresh copy of the same length differing only in those 51 bytes, with an identical pdftotext -layout text layer. Inspected artifact

How to read the evidence

Sources and history

Release 2026-10-22-1b7cf027e4f6 · Record review: source checked

Preprint of the GPN-MSA paper (Benegas, Albors, Aw, Ye and Song), version 2, retrieved as PDF.

Release history

No supporting source is linked yet.

Read original source

Download this release (gzip)

Technical metadata and extraction receipts

Stable ID: gfm-guides-20261014-gpnmsa-source-biorxiv-v2

areas: dna-genomes
contexts: research
url: https://www.biorxiv.org/content/10.1101/2023.10.10.561776v2
artifact url: https://www.biorxiv.org/content/10.1101/2023.10.10.561776v2.full.pdf
doi: 10.1101/2023.10.10.561776
version: bioRxiv v2, posted 2024-04-06
retrieved at: 2026-10-14T12:02:16Z
artifact sha256: 10411077df1ae59614f9a9a9f85384103d54127ba67a9b3c469a4a2d6c33cee4
publication status: preprint
licence: CC BY-NC-ND 4.0
media type: application/pdf
venue: bioRxiv
year: 2024
limitations: Preprint. The peer-reviewed version is Benegas et al., Nature Biotechnology 43, 1960-1965 (2025), doi:10.1038/s41587-024-02511-w, which is not open access; its Extended Data Table 1 (Ablation study) is published only as an image and was not read, so the printed values here may differ from the journal version.; The issue named bioRxiv 2023.10.10.562811; that DOI does not resolve. bioRxiv and Crossref give 10.1101/2023.10.10.561776 for this preprint, and the bioRxiv API lists the journal DOI 10.1038/s41587-024-02511-w as its published version.; The main comparisons with CADD, phyloP, ESM-1b, Nucleotide Transformer and HyenaDNA on ClinVar, COSMIC, OMIM and gnomAD (Figure 2c-f, Supplementary Figures 3-5) are printed only as figures and were not extracted.; Supplementary Table 4 caption says ClinVar, COSMIC, OMIM and gnomAD use 'same setting as Figure 2a', '2b', '2c' and '2d', but in v2 those settings are Figure 2c, 2d, 2e and 2f (Figure 2 caption and Results). The settings were taken from the Methods 'Main VEP metrics' and the caption's own description.; Licence CC BY-NC-ND 4.0; no archived copy is kept in the batch folder.
source locator: Supplementary Table 4 (PDF page 23); Results paragraph on DMS (PDF page 6); Methods 'Ablation Study', 'VEP data sources', 'Main VEP metrics', 'Additional VEP metrics', 'VEP scores'
review: method: source-hash-verification; method note: The extractor's downloaded copy hashes to the pinned value. A fresh download on 2026-10-14 differs in 51 bytes, all in bioRxiv's per-download stamps; both give the same text layer.; reviewer: claude; reviewer note: Claude (Opus 5.5) research agent, independent of the extractor; no human review claimed; date: 2026-10-14; artifact sha256: 10411077df1ae59614f9a9a9f85384103d54127ba67a9b3c469a4a2d6c33cee4; retrieval url: https://www.biorxiv.org/content/10.1101/2023.10.10.561776v2.full.pdf; note: Independent review 2026-10-14. hash_scope added, following the BioML-bench and SpliceAI-2 precedent.
hash scope: SHA-256 of the PDF bytes as downloaded 2026-10-14T12:02:16Z. bioRxiv rewrites the XMP dates and document IDs, the Info ModDate and the trailer /ID on download, so a re-download does not reproduce the hash; the review found a fresh copy of the same length differing only in those 51 bytes, with an identical pdftotext -layout text layer.

Related records

Suggest a correction