Protocol

gnomAD rare vs common enrichment in the deleterious tail, odds ratio (GPN-MSA Supplementary Table 4)

Odds ratio of rare versus common gnomAD variants among the variants a score places in its deleterious tail, the tail threshold set so that each score makes 30 false discoveries.

17 evaluations · 34 results

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.

17 recorded evaluations, 34 metric rows. A comparison chart has not yet been validated for these results. The table retains the individual findings and their sources.

View coverage and remaining gaps across all benchmarks

Results

Results are available, but no reviewed comparison panel is linked in this release.

All evaluations

17 evaluations · 34 results. Different protocols are not a single leaderboard.

Protocol: gnomAD rare vs common enrichment in the deleterious tail, odds ratio (GPN-MSA Supplementary Table 4) · Dataset: gnomAD rare (singletons) vs common variants, ablation set (GPN-MSA paper)

Sorted by 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) (higher 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	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)	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)Evaluation	46.8 (best in column)	41.6
GPN-MSA ablation 'Don’t replace non-conserved' (Supplementary Table 4)Evaluation	45.2	41.9 (best in column)
GPN-MSA ablation '51 mammals' (Supplementary Table 4)Evaluation	43.4	41.4
GPN-MSA ablation 'Default' (Supplementary Table 4)Evaluation	43.1	41.4
GPN-MSA ablation 'Window size = 256' (Supplementary Table 4)Evaluation	43	37.9
GPN-MSA ablation 'Don’t upweight conserved' (Supplementary Table 4)Evaluation	40	37.1
GPN-MSA ablation 'Window size = 64' (Supplementary Table 4)Evaluation	39.5	37.5
GPN-MSA ablation 'Window size = 32' (Supplementary Table 4)Evaluation	38.9	36.7
GPN-MSA ablation 'Train on 50% most conserved' (Supplementary Table 4)Evaluation	35.3	32.1
GPN-MSA ablation 'Window size = 16' (Supplementary Table 4)Evaluation	31.6	29.2
GPN-MSA ablation 'Train on 100% of genome' (Supplementary Table 4)Evaluation	30.7	28
GPN-MSA ablation 'Include closest primates' (Supplementary Table 4)Evaluation	26.7	24.7
GPN-MSA ablation 'Window size = 8' (Supplementary Table 4)Evaluation	21.9	21.2
MSA frequencyEvaluation	15.1	15.1
GPN-MSA ablation 'Window size = 4' (Supplementary Table 4)Evaluation	14.8	13.5
Combined phyloP and phastConsEvaluation	13.9	13.9
GPN-MSA ablation 'w/o MSA' (Supplementary Table 4)Evaluation	3.09	2.77

All 34 result rows with coverage, uncertainty and sources

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: GPN-MSA ablation '51 mammals' (Supplementary Table 4)	Protocol: gnomAD rare vs common enrichment in the deleterious tail, odds ratio (GPN-MSA Supplementary Table 4) Dataset: gnomAD rare (singletons) vs common variants, ablation set (GPN-MSA paper)	43.4 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) unitless · higher Uncertainty: Not reported by the source: Only mean and max over seeds are printed Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source 51 mammals on gnomAD (GPN-MSA Supplementary Table 4) gfm-guides-20261014-gpnmsa-protocol-ablation-gnomad-odds-ratio Aggregation: Not reported GPN-MSA: an alignment-based DNA language model for genome-wide variant effect prediction (bioRxiv v2) · Supplementary Table 4 (PDF page 23), row '51 mammals', column 'gnomAD' / 'max'
Configuration: GPN-MSA ablation '51 mammals' (Supplementary Table 4)	Protocol: gnomAD rare vs common enrichment in the deleterious tail, odds ratio (GPN-MSA Supplementary Table 4) Dataset: gnomAD rare (singletons) vs common variants, ablation set (GPN-MSA paper)	41.4 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) unitless · higher Uncertainty: Not reported by the source: Only mean and max over seeds are printed Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source 51 mammals on gnomAD (GPN-MSA Supplementary Table 4) gfm-guides-20261014-gpnmsa-protocol-ablation-gnomad-odds-ratio Aggregation: Not reported GPN-MSA: an alignment-based DNA language model for genome-wide variant effect prediction (bioRxiv v2) · Supplementary Table 4 (PDF page 23), row '51 mammals', column 'gnomAD' / 'mean'
Configuration: GPN-MSA ablation '51 vertebrates' (Supplementary Table 4)	Protocol: gnomAD rare vs common enrichment in the deleterious tail, odds ratio (GPN-MSA Supplementary Table 4) Dataset: gnomAD rare (singletons) vs common variants, ablation set (GPN-MSA paper)	46.8 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) unitless · higher Uncertainty: Not reported by the source: Only mean and max over seeds are printed Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source 51 vertebrates on gnomAD (GPN-MSA Supplementary Table 4) gfm-guides-20261014-gpnmsa-protocol-ablation-gnomad-odds-ratio Aggregation: Not reported GPN-MSA: an alignment-based DNA language model for genome-wide variant effect prediction (bioRxiv v2) · Supplementary Table 4 (PDF page 23), row '51 vertebrates', column 'gnomAD' / 'max'
Configuration: GPN-MSA ablation '51 vertebrates' (Supplementary Table 4)	Protocol: gnomAD rare vs common enrichment in the deleterious tail, odds ratio (GPN-MSA Supplementary Table 4) Dataset: gnomAD rare (singletons) vs common variants, ablation set (GPN-MSA paper)	41.6 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) unitless · higher Uncertainty: Not reported by the source: Only mean and max over seeds are printed Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source 51 vertebrates on gnomAD (GPN-MSA Supplementary Table 4) gfm-guides-20261014-gpnmsa-protocol-ablation-gnomad-odds-ratio Aggregation: Not reported GPN-MSA: an alignment-based DNA language model for genome-wide variant effect prediction (bioRxiv v2) · Supplementary Table 4 (PDF page 23), row '51 vertebrates', column 'gnomAD' / 'mean'
Configuration: GPN-MSA ablation 'Default' (Supplementary Table 4)	Protocol: gnomAD rare vs common enrichment in the deleterious tail, odds ratio (GPN-MSA Supplementary Table 4) Dataset: gnomAD rare (singletons) vs common variants, ablation set (GPN-MSA paper)	43.1 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) unitless · higher Uncertainty: Not reported by the source: Only mean and max over seeds are printed Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source Default on gnomAD (GPN-MSA Supplementary Table 4) gfm-guides-20261014-gpnmsa-protocol-ablation-gnomad-odds-ratio Aggregation: Not reported GPN-MSA: an alignment-based DNA language model for genome-wide variant effect prediction (bioRxiv v2) · Supplementary Table 4 (PDF page 23), row 'Default', column 'gnomAD' / 'max'
Configuration: GPN-MSA ablation 'Default' (Supplementary Table 4)	Protocol: gnomAD rare vs common enrichment in the deleterious tail, odds ratio (GPN-MSA Supplementary Table 4) Dataset: gnomAD rare (singletons) vs common variants, ablation set (GPN-MSA paper)	41.4 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) unitless · higher Uncertainty: Not reported by the source: Only mean and max over seeds are printed Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source Default on gnomAD (GPN-MSA Supplementary Table 4) gfm-guides-20261014-gpnmsa-protocol-ablation-gnomad-odds-ratio Aggregation: Not reported GPN-MSA: an alignment-based DNA language model for genome-wide variant effect prediction (bioRxiv v2) · Supplementary Table 4 (PDF page 23), row 'Default', column 'gnomAD' / 'mean'
Configuration: GPN-MSA ablation 'Include closest primates' (Supplementary Table 4)	Protocol: gnomAD rare vs common enrichment in the deleterious tail, odds ratio (GPN-MSA Supplementary Table 4) Dataset: gnomAD rare (singletons) vs common variants, ablation set (GPN-MSA paper)	26.7 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) unitless · higher Uncertainty: Not reported by the source: Only mean and max over seeds are printed Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source Include closest primates on gnomAD (GPN-MSA Supplementary Table 4) gfm-guides-20261014-gpnmsa-protocol-ablation-gnomad-odds-ratio Aggregation: Not reported GPN-MSA: an alignment-based DNA language model for genome-wide variant effect prediction (bioRxiv v2) · Supplementary Table 4 (PDF page 23), row 'Include closest primates', column 'gnomAD' / 'max'
Configuration: GPN-MSA ablation 'Include closest primates' (Supplementary Table 4)	Protocol: gnomAD rare vs common enrichment in the deleterious tail, odds ratio (GPN-MSA Supplementary Table 4) Dataset: gnomAD rare (singletons) vs common variants, ablation set (GPN-MSA paper)	24.7 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) unitless · higher Uncertainty: Not reported by the source: Only mean and max over seeds are printed Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source Include closest primates on gnomAD (GPN-MSA Supplementary Table 4) gfm-guides-20261014-gpnmsa-protocol-ablation-gnomad-odds-ratio Aggregation: Not reported GPN-MSA: an alignment-based DNA language model for genome-wide variant effect prediction (bioRxiv v2) · Supplementary Table 4 (PDF page 23), row 'Include closest primates', column 'gnomAD' / 'mean'
Method: MSA frequency	Protocol: gnomAD rare vs common enrichment in the deleterious tail, odds ratio (GPN-MSA Supplementary Table 4) Dataset: gnomAD rare (singletons) vs common variants, ablation set (GPN-MSA paper)	15.1 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) unitless · higher Uncertainty: Not reported by the source: Only mean and max over seeds are printed Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source MSA frequency (no neural net) on gnomAD (GPN-MSA Supplementary Table 4) gfm-guides-20261014-gpnmsa-protocol-ablation-gnomad-odds-ratio Aggregation: Not reported GPN-MSA: an alignment-based DNA language model for genome-wide variant effect prediction (bioRxiv v2) · Supplementary Table 4 (PDF page 23), row 'MSA frequency (no neural net)', column 'gnomAD' / 'max'
Method: MSA frequency	Protocol: gnomAD rare vs common enrichment in the deleterious tail, odds ratio (GPN-MSA Supplementary Table 4) Dataset: gnomAD rare (singletons) vs common variants, ablation set (GPN-MSA paper)	15.1 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) unitless · higher Uncertainty: Not reported by the source: Only mean and max over seeds are printed Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source MSA frequency (no neural net) on gnomAD (GPN-MSA Supplementary Table 4) gfm-guides-20261014-gpnmsa-protocol-ablation-gnomad-odds-ratio Aggregation: Not reported GPN-MSA: an alignment-based DNA language model for genome-wide variant effect prediction (bioRxiv v2) · Supplementary Table 4 (PDF page 23), row 'MSA frequency (no neural net)', column 'gnomAD' / 'mean'
Configuration: GPN-MSA ablation 'Don’t replace non-conserved' (Supplementary Table 4)	Protocol: gnomAD rare vs common enrichment in the deleterious tail, odds ratio (GPN-MSA Supplementary Table 4) Dataset: gnomAD rare (singletons) vs common variants, ablation set (GPN-MSA paper)	45.2 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) unitless · higher Uncertainty: Not reported by the source: Only mean and max over seeds are printed Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source Don’t replace non-conserved on gnomAD (GPN-MSA Supplementary Table 4) gfm-guides-20261014-gpnmsa-protocol-ablation-gnomad-odds-ratio Aggregation: Not reported GPN-MSA: an alignment-based DNA language model for genome-wide variant effect prediction (bioRxiv v2) · Supplementary Table 4 (PDF page 23), row 'Don’t replace non-conserved', column 'gnomAD' / 'max'
Configuration: GPN-MSA ablation 'Don’t replace non-conserved' (Supplementary Table 4)	Protocol: gnomAD rare vs common enrichment in the deleterious tail, odds ratio (GPN-MSA Supplementary Table 4) Dataset: gnomAD rare (singletons) vs common variants, ablation set (GPN-MSA paper)	41.9 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) unitless · higher Uncertainty: Not reported by the source: Only mean and max over seeds are printed Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source Don’t replace non-conserved on gnomAD (GPN-MSA Supplementary Table 4) gfm-guides-20261014-gpnmsa-protocol-ablation-gnomad-odds-ratio Aggregation: Not reported GPN-MSA: an alignment-based DNA language model for genome-wide variant effect prediction (bioRxiv v2) · Supplementary Table 4 (PDF page 23), row 'Don’t replace non-conserved', column 'gnomAD' / 'mean'
Configuration: GPN-MSA ablation 'Don’t upweight conserved' (Supplementary Table 4)	Protocol: gnomAD rare vs common enrichment in the deleterious tail, odds ratio (GPN-MSA Supplementary Table 4) Dataset: gnomAD rare (singletons) vs common variants, ablation set (GPN-MSA paper)	40 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) unitless · higher Uncertainty: Not reported by the source: Only mean and max over seeds are printed Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source Don’t upweight conserved on gnomAD (GPN-MSA Supplementary Table 4) gfm-guides-20261014-gpnmsa-protocol-ablation-gnomad-odds-ratio Aggregation: Not reported GPN-MSA: an alignment-based DNA language model for genome-wide variant effect prediction (bioRxiv v2) · Supplementary Table 4 (PDF page 23), row 'Don’t upweight conserved', column 'gnomAD' / 'max'
Configuration: GPN-MSA ablation 'Don’t upweight conserved' (Supplementary Table 4)	Protocol: gnomAD rare vs common enrichment in the deleterious tail, odds ratio (GPN-MSA Supplementary Table 4) Dataset: gnomAD rare (singletons) vs common variants, ablation set (GPN-MSA paper)	37.1 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) unitless · higher Uncertainty: Not reported by the source: Only mean and max over seeds are printed Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source Don’t upweight conserved on gnomAD (GPN-MSA Supplementary Table 4) gfm-guides-20261014-gpnmsa-protocol-ablation-gnomad-odds-ratio Aggregation: Not reported GPN-MSA: an alignment-based DNA language model for genome-wide variant effect prediction (bioRxiv v2) · Supplementary Table 4 (PDF page 23), row 'Don’t upweight conserved', column 'gnomAD' / 'mean'
Method: Combined phyloP and phastCons	Protocol: gnomAD rare vs common enrichment in the deleterious tail, odds ratio (GPN-MSA Supplementary Table 4) Dataset: gnomAD rare (singletons) vs common variants, ablation set (GPN-MSA paper)	13.9 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) unitless · higher Uncertainty: Not reported by the source: Only mean and max over seeds are printed Coverage: Not reported scored / Not reported eligible	Independent external evaluation · Source checked Methods, coverage and source Combined phyloP and phastCons on gnomAD (GPN-MSA Supplementary Table 4) gfm-guides-20261014-gpnmsa-protocol-ablation-gnomad-odds-ratio Aggregation: Not reported GPN-MSA: an alignment-based DNA language model for genome-wide variant effect prediction (bioRxiv v2) · Supplementary Table 4 (PDF page 23), row 'Combined phyloP and phastCons', column 'gnomAD' / 'max'
Method: Combined phyloP and phastCons	Protocol: gnomAD rare vs common enrichment in the deleterious tail, odds ratio (GPN-MSA Supplementary Table 4) Dataset: gnomAD rare (singletons) vs common variants, ablation set (GPN-MSA paper)	13.9 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) unitless · higher Uncertainty: Not reported by the source: Only mean and max over seeds are printed Coverage: Not reported scored / Not reported eligible	Independent external evaluation · Source checked Methods, coverage and source Combined phyloP and phastCons on gnomAD (GPN-MSA Supplementary Table 4) gfm-guides-20261014-gpnmsa-protocol-ablation-gnomad-odds-ratio Aggregation: Not reported GPN-MSA: an alignment-based DNA language model for genome-wide variant effect prediction (bioRxiv v2) · Supplementary Table 4 (PDF page 23), row 'Combined phyloP and phastCons', column 'gnomAD' / 'mean'
Configuration: GPN-MSA ablation 'Train on 100% of genome' (Supplementary Table 4)	Protocol: gnomAD rare vs common enrichment in the deleterious tail, odds ratio (GPN-MSA Supplementary Table 4) Dataset: gnomAD rare (singletons) vs common variants, ablation set (GPN-MSA paper)	30.7 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) unitless · higher Uncertainty: Not reported by the source: Only mean and max over seeds are printed Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source Train on 100% of genome on gnomAD (GPN-MSA Supplementary Table 4) gfm-guides-20261014-gpnmsa-protocol-ablation-gnomad-odds-ratio Aggregation: Not reported GPN-MSA: an alignment-based DNA language model for genome-wide variant effect prediction (bioRxiv v2) · Supplementary Table 4 (PDF page 23), row 'Train on 100% of genome', column 'gnomAD' / 'max'
Configuration: GPN-MSA ablation 'Train on 100% of genome' (Supplementary Table 4)	Protocol: gnomAD rare vs common enrichment in the deleterious tail, odds ratio (GPN-MSA Supplementary Table 4) Dataset: gnomAD rare (singletons) vs common variants, ablation set (GPN-MSA paper)	28 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) unitless · higher Uncertainty: Not reported by the source: Only mean and max over seeds are printed Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source Train on 100% of genome on gnomAD (GPN-MSA Supplementary Table 4) gfm-guides-20261014-gpnmsa-protocol-ablation-gnomad-odds-ratio Aggregation: Not reported GPN-MSA: an alignment-based DNA language model for genome-wide variant effect prediction (bioRxiv v2) · Supplementary Table 4 (PDF page 23), row 'Train on 100% of genome', column 'gnomAD' / 'mean'
Configuration: GPN-MSA ablation 'Train on 50% most conserved' (Supplementary Table 4)	Protocol: gnomAD rare vs common enrichment in the deleterious tail, odds ratio (GPN-MSA Supplementary Table 4) Dataset: gnomAD rare (singletons) vs common variants, ablation set (GPN-MSA paper)	35.3 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) unitless · higher Uncertainty: Not reported by the source: Only mean and max over seeds are printed Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source Train on 50% most conserved on gnomAD (GPN-MSA Supplementary Table 4) gfm-guides-20261014-gpnmsa-protocol-ablation-gnomad-odds-ratio Aggregation: Not reported GPN-MSA: an alignment-based DNA language model for genome-wide variant effect prediction (bioRxiv v2) · Supplementary Table 4 (PDF page 23), row 'Train on 50% most conserved', column 'gnomAD' / 'max'
Configuration: GPN-MSA ablation 'Train on 50% most conserved' (Supplementary Table 4)	Protocol: gnomAD rare vs common enrichment in the deleterious tail, odds ratio (GPN-MSA Supplementary Table 4) Dataset: gnomAD rare (singletons) vs common variants, ablation set (GPN-MSA paper)	32.1 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) unitless · higher Uncertainty: Not reported by the source: Only mean and max over seeds are printed Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source Train on 50% most conserved on gnomAD (GPN-MSA Supplementary Table 4) gfm-guides-20261014-gpnmsa-protocol-ablation-gnomad-odds-ratio Aggregation: Not reported GPN-MSA: an alignment-based DNA language model for genome-wide variant effect prediction (bioRxiv v2) · Supplementary Table 4 (PDF page 23), row 'Train on 50% most conserved', column 'gnomAD' / 'mean'
Configuration: GPN-MSA ablation 'Window size = 16' (Supplementary Table 4)	Protocol: gnomAD rare vs common enrichment in the deleterious tail, odds ratio (GPN-MSA Supplementary Table 4) Dataset: gnomAD rare (singletons) vs common variants, ablation set (GPN-MSA paper)	31.6 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) unitless · higher Uncertainty: Not reported by the source: Only mean and max over seeds are printed Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source Window size = 16 on gnomAD (GPN-MSA Supplementary Table 4) gfm-guides-20261014-gpnmsa-protocol-ablation-gnomad-odds-ratio Aggregation: Not reported GPN-MSA: an alignment-based DNA language model for genome-wide variant effect prediction (bioRxiv v2) · Supplementary Table 4 (PDF page 23), row 'Window size = 16', column 'gnomAD' / 'max'
Configuration: GPN-MSA ablation 'Window size = 16' (Supplementary Table 4)	Protocol: gnomAD rare vs common enrichment in the deleterious tail, odds ratio (GPN-MSA Supplementary Table 4) Dataset: gnomAD rare (singletons) vs common variants, ablation set (GPN-MSA paper)	29.2 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) unitless · higher Uncertainty: Not reported by the source: Only mean and max over seeds are printed Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source Window size = 16 on gnomAD (GPN-MSA Supplementary Table 4) gfm-guides-20261014-gpnmsa-protocol-ablation-gnomad-odds-ratio Aggregation: Not reported GPN-MSA: an alignment-based DNA language model for genome-wide variant effect prediction (bioRxiv v2) · Supplementary Table 4 (PDF page 23), row 'Window size = 16', column 'gnomAD' / 'mean'
Configuration: GPN-MSA ablation 'Window size = 256' (Supplementary Table 4)	Protocol: gnomAD rare vs common enrichment in the deleterious tail, odds ratio (GPN-MSA Supplementary Table 4) Dataset: gnomAD rare (singletons) vs common variants, ablation set (GPN-MSA paper)	43 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) unitless · higher Uncertainty: Not reported by the source: Only mean and max over seeds are printed Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source Window size = 256 on gnomAD (GPN-MSA Supplementary Table 4) gfm-guides-20261014-gpnmsa-protocol-ablation-gnomad-odds-ratio Aggregation: Not reported GPN-MSA: an alignment-based DNA language model for genome-wide variant effect prediction (bioRxiv v2) · Supplementary Table 4 (PDF page 23), row 'Window size = 256', column 'gnomAD' / 'max'
Configuration: GPN-MSA ablation 'Window size = 256' (Supplementary Table 4)	Protocol: gnomAD rare vs common enrichment in the deleterious tail, odds ratio (GPN-MSA Supplementary Table 4) Dataset: gnomAD rare (singletons) vs common variants, ablation set (GPN-MSA paper)	37.9 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) unitless · higher Uncertainty: Not reported by the source: Only mean and max over seeds are printed Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source Window size = 256 on gnomAD (GPN-MSA Supplementary Table 4) gfm-guides-20261014-gpnmsa-protocol-ablation-gnomad-odds-ratio Aggregation: Not reported GPN-MSA: an alignment-based DNA language model for genome-wide variant effect prediction (bioRxiv v2) · Supplementary Table 4 (PDF page 23), row 'Window size = 256', column 'gnomAD' / 'mean'
Configuration: GPN-MSA ablation 'Window size = 32' (Supplementary Table 4)	Protocol: gnomAD rare vs common enrichment in the deleterious tail, odds ratio (GPN-MSA Supplementary Table 4) Dataset: gnomAD rare (singletons) vs common variants, ablation set (GPN-MSA paper)	38.9 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) unitless · higher Uncertainty: Not reported by the source: Only mean and max over seeds are printed Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source Window size = 32 on gnomAD (GPN-MSA Supplementary Table 4) gfm-guides-20261014-gpnmsa-protocol-ablation-gnomad-odds-ratio Aggregation: Not reported GPN-MSA: an alignment-based DNA language model for genome-wide variant effect prediction (bioRxiv v2) · Supplementary Table 4 (PDF page 23), row 'Window size = 32', column 'gnomAD' / 'max'

Source checking is not independent reproduction. Release 2026-10-22-1b7cf027e4f6.

Methods and evaluation design

Procedure, tasks and evaluated configurations

Recorded evaluations

Each evaluation records what was tested and under which conditions.

Explore all linked results

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.

Author-reported evaluations: 16
External evaluations: 1

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 2026-10-22-1b7cf027e4f6. 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.

Strengths, 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

Claims, original sources and review scope · Release 2026-10-22-1b7cf027e4f6
Property and statement	Original source and location	Review and provenance

No evidence rows match these filters. Choose another scope or clear the search.

How to read the evidence

Sources and history

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

1 source record and release history

GPN-MSA: an alignment-based DNA language model for genome-wide variant effect prediction (bioRxiv v2) · Original source · bioRxiv v2, posted 2024-04-06

Download this release (gzip)

Technical metadata and extraction receipts

Stable ID: gfm-guides-20261014-gpnmsa-protocol-ablation-gnomad-odds-ratio

areas: dna-genomes
contexts: research
protocol: Each model or score is applied zero-shot (no training on these labels). Variants are ranked by score; the deleterious tail is cut where the score makes 30 false discoveries (common variants), and the odds ratio of rare versus common variants in the tail is computed with a one-sided Fisher's exact test. All variant types, with rare variants subsampled to match the number of common variants. For each ablation setting three replicate models with different seeds were trained where applicable; the table prints the mean and the maximum over seeds.
version: bioRxiv v2 Supplementary Table 4
metric: odds-ratio
metric direction: higher
unit: unitless
limitations: Benchmark built and run by the GPN-MSA authors (author-reported).; The caption says 'same approach as Figure 2d'; in v2 the gnomAD enrichment panel is Figure 2f (the caption's figure letters are off by two, as for the other columns).; The false-discovery count for the ablation table is not restated; Figure 2f and Methods 'Main VEP metrics' give 30 as the example threshold.; Variant counts and p-values are not printed in Supplementary Table 4.
source locator: Supplementary Table 4 caption and column group 'gnomAD'; Figure 2 caption panel (f); Methods 'Main VEP metrics' (gnomAD) and 'Additional VEP metrics' ('gnomAD (ablation set)')

Related records

part of: GPN-MSA variant effect prediction evaluation (Benegas et al.)
uses data: gnomAD rare (singletons) vs common variants, ablation set (GPN-MSA paper)
assessment: 51 mammals on gnomAD (GPN-MSA Supplementary Table 4)
assessment: 51 vertebrates on gnomAD (GPN-MSA Supplementary Table 4)
assessment: Default on gnomAD (GPN-MSA Supplementary Table 4)
assessment: Include closest primates on gnomAD (GPN-MSA Supplementary Table 4)
assessment: MSA frequency (no neural net) on gnomAD (GPN-MSA Supplementary Table 4)
assessment: Don’t replace non-conserved on gnomAD (GPN-MSA Supplementary Table 4)
assessment: Don’t upweight conserved on gnomAD (GPN-MSA Supplementary Table 4)
assessment: Combined phyloP and phastCons on gnomAD (GPN-MSA Supplementary Table 4)
assessment: Train on 100% of genome on gnomAD (GPN-MSA Supplementary Table 4)
assessment: Train on 50% most conserved on gnomAD (GPN-MSA Supplementary Table 4)
assessment: Window size = 16 on gnomAD (GPN-MSA Supplementary Table 4)
assessment: Window size = 256 on gnomAD (GPN-MSA Supplementary Table 4)
assessment: Window size = 32 on gnomAD (GPN-MSA Supplementary Table 4)
assessment: Window size = 4 on gnomAD (GPN-MSA Supplementary Table 4)
assessment: Window size = 64 on gnomAD (GPN-MSA Supplementary Table 4)
assessment: Window size = 8 on gnomAD (GPN-MSA Supplementary Table 4)
assessment: w/o MSA on gnomAD (GPN-MSA Supplementary Table 4)

Suggest a correction