GPN trained on eight Brassicales genomes, repeat loss weight 0.1 (PNAS 2023)
Masked DNA language model trained on eight unaligned Brassicales reference genomes with the prediction loss on repetitive elements down-weighted to 0.1, which scores a variant by the log-likelihood ratio between its alternate and reference allele.
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.
Model Capabilities & Hardware Requirements
Deployment facts recorded for GPN trained on eight Brassicales genomes, repeat loss weight 0.1 (PNAS 2023). Fields the record does not carry are marked not recorded.
Evaluated Checkpoint: As trained for PNAS 120(44) e2311219120; no released checkpoint is named
Evaluations and results
6 evaluations · 14 results. Different protocols are not a single leaderboard.
Sorted by Accuracy (gene region class CDS, leave-one-chromosome-out logistic regression on averaged 100 bp GPN embeddings) (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 | Protocol and dataset | Accuracy (gene region class CDS, leave-one-chromosome-out logistic regression on averaged 100 bp GPN embeddings) | Accuracy (gene region class ncRNA, leave-one-chromosome-out logistic regression on averaged 100 bp GPN embeddings) | Odds ratio (rare (allele count 1) against common (allele frequency at least 5%) 1001 Genomes variants, putative functional score threshold at the lowest 0.1% quantile) | Perplexity (masked nucleotide prediction, chromosome-wide on the held-out test chromosome) | Perplexity (masked nucleotide prediction, over non-repeat regions on the held-out test chromosome) | Perplexity (median over CDS1 positions in Chr5:3,500,000-4,500,000, repeats excluded) | Perplexity (median over CDS2 positions in Chr5:3,500,000-4,500,000, repeats excluded) | Perplexity (median over CDS3 positions in Chr5:3,500,000-4,500,000, repeats excluded) | Perplexity (median over intergenic and intronic positions in Chr5:3,500,000-4,500,000, repeats excluded) | Perplexity (median over splice acceptor positions in Chr5:3,500,000-4,500,000, repeats excluded) | Perplexity (median over splice donor positions in Chr5:3,500,000-4,500,000, repeats excluded) | Perplexity (median over start codon positions in Chr5:3,500,000-4,500,000, repeats excluded) | Perplexity (median over stop codon positions in Chr5:3,500,000-4,500,000, repeats excluded) | Proportion (simulated repeat variants ranked before the first decile of simulated missense variants, GPN trained with repeat weight 0.1) |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| GPN trained on eight Brassicales genomes, repeat loss weight 0.1 (PNAS 2023)Evaluation | Gene region classification from averaged GPN embeddings, leave-one-chromosome-out, Non-overlapping 100 bp windows of the Arabidopsis thaliana genome, labelled by gene region | 96% | 51% | Not reported | Not reported | Not reported | Not reported | Not reported | Not reported | Not reported | Not reported | Not reported | Not reported | Not reported | Not reported |
| GPN trained on eight Brassicales genomes, repeat loss weight 0.1 (PNAS 2023)Evaluation | Test perplexity on Arabidopsis thaliana chromosome 5, chromosome-wide, Arabidopsis thaliana chromosome 5, held out for testing | Not reported | Not reported | Not reported | 2.9 | Not reported | Not reported | Not reported | Not reported | Not reported | Not reported | Not reported | Not reported | Not reported | Not reported |
| GPN trained on eight Brassicales genomes, repeat loss weight 0.1 (PNAS 2023)Evaluation | Test perplexity on Arabidopsis thaliana chromosome 5, non-repeat regions, Non-repeat regions of Arabidopsis thaliana chromosome 5 | Not reported | Not reported | Not reported | Not reported | 2.92 | Not reported | Not reported | Not reported | Not reported | Not reported | Not reported | Not reported | Not reported | Not reported |
| GPN trained on eight Brassicales genomes, repeat loss weight 0.1 (PNAS 2023)Evaluation | Median held-out perplexity by position class in Chr5:3,500,000-4,500,000, Positions of Arabidopsis thaliana Chr5:3,500,000-4,500,000 outside repeats, by class | Not reported | Not reported | Not reported | Not reported | Not reported | 2.44 | 2.24 | 2.79 | 3.24 | 1.03 | 1.02 | 1.08 | 2.8 | Not reported |
| GPN trained on eight Brassicales genomes, repeat loss weight 0.1 (PNAS 2023)Evaluation | Rare versus common variant enrichment in the lowest 0.1% of scores, 1001 Genomes, 1001 Genomes Arabidopsis thaliana SNPs with allele frequencies | Not reported | Not reported | 5.5-fold | Not reported | Not reported | Not reported | Not reported | Not reported | Not reported | Not reported | Not reported | Not reported | Not reported | Not reported |
| GPN trained on eight Brassicales genomes, repeat loss weight 0.1 (PNAS 2023)Evaluation | Share of simulated repeat variants scored below the first decile of missense variants, All possible SNPs in Arabidopsis thaliana Chr5:3,500,000-4,500,000, annotated by consequence | Not reported | Not reported | Not reported | Not reported | Not reported | Not reported | Not reported | Not reported | Not reported | Not reported | Not reported | Not reported | Not reported | 9% |
All 14 result rows with coverage, uncertainty and sources
Filter evaluations
Applied filters: All linked evaluations
| Tested configuration | Protocol and dataset | Finding | Evidence and details |
|---|---|---|---|
| Configuration: GPN trained on eight Brassicales genomes, repeat loss weight 0.1 (PNAS 2023) | Protocol: Gene region classification from averaged GPN embeddings, leave-one-chromosome-out Dataset subset: Non-overlapping 100 bp windows of the Arabidopsis thaliana genome, labelled by gene region | 96% Accuracy (gene region class CDS, leave-one-chromosome-out logistic regression on averaged 100 bp GPN embeddings) percent · higher Uncertainty: Not reported by the source: The source prints no interval or spread for this value. Coverage: Not reported scored / Not reported eligible | Author-reported evaluation · Source checkedMethods, coverage and sourceGene region classification from averaged GPN embeddings (PNAS 2023) gpn-paper-20261116-protocol-gene-region-classification Aggregation: Not reported DNA language models are powerful predictors of genome-wide variant effects (PNAS 2023) · Results, 'Unsupervised Clustering of Genomic Regions', sentence beginning "To quantify GPN's ability to distinguish genomic regions", term "CDS (96%)" |
| Configuration: GPN trained on eight Brassicales genomes, repeat loss weight 0.1 (PNAS 2023) | Protocol: Gene region classification from averaged GPN embeddings, leave-one-chromosome-out Dataset subset: Non-overlapping 100 bp windows of the Arabidopsis thaliana genome, labelled by gene region | 51% Accuracy (gene region class ncRNA, leave-one-chromosome-out logistic regression on averaged 100 bp GPN embeddings) percent · higher Uncertainty: Not reported by the source: The source prints no interval or spread for this value. Coverage: Not reported scored / Not reported eligible | Author-reported evaluation · Source checkedMethods, coverage and sourceGene region classification from averaged GPN embeddings (PNAS 2023) gpn-paper-20261116-protocol-gene-region-classification Aggregation: Not reported DNA language models are powerful predictors of genome-wide variant effects (PNAS 2023) · Results, 'Unsupervised Clustering of Genomic Regions', sentence beginning "To quantify GPN's ability to distinguish genomic regions", term "ncRNA (51%)" |
| Configuration: GPN trained on eight Brassicales genomes, repeat loss weight 0.1 (PNAS 2023) | Protocol: Test perplexity on Arabidopsis thaliana chromosome 5, chromosome-wide Dataset subset: Arabidopsis thaliana chromosome 5, held out for testing | 2.9 Perplexity (masked nucleotide prediction, chromosome-wide on the held-out test chromosome) unitless · lower Uncertainty: Not reported by the source: The source prints no interval or spread for this value. Coverage: Not reported scored / Not reported eligible | Author-reported evaluation · Source checkedMethods, coverage and sourceTest perplexity, chromosome-wide, GPN with repeat weight 0.1 (PNAS 2023 Table S2) gpn-paper-20261116-protocol-test-perplexity-chromosome-wide Aggregation: Not reported SI Appendix for PNAS 2023 GPN paper (Tables S1 to S3, Figures S1 to S12); DNA language models are powerful predictors of genome-wide variant effects (PNAS 2023) · Table S2 (SI Appendix), row "Repeat weight 0.1", column "Chromosome-wide" |
| Configuration: GPN trained on eight Brassicales genomes, repeat loss weight 0.1 (PNAS 2023) | Protocol: Test perplexity on Arabidopsis thaliana chromosome 5, non-repeat regions Dataset subset: Non-repeat regions of Arabidopsis thaliana chromosome 5 | 2.92 Perplexity (masked nucleotide prediction, over non-repeat regions on the held-out test chromosome) unitless · lower Uncertainty: Not reported by the source: The source prints no interval or spread for this value. Coverage: Not reported scored / Not reported eligible | Author-reported evaluation · Source checkedMethods, coverage and sourceTest perplexity, non-repeat, GPN with repeat weight 0.1 (PNAS 2023 Table S2) gpn-paper-20261116-protocol-test-perplexity-non-repeat Aggregation: Not reported SI Appendix for PNAS 2023 GPN paper (Tables S1 to S3, Figures S1 to S12); DNA language models are powerful predictors of genome-wide variant effects (PNAS 2023) · Table S2 (SI Appendix), row "Repeat weight 0.1", column "Non-repeat regions" |
| Configuration: GPN trained on eight Brassicales genomes, repeat loss weight 0.1 (PNAS 2023) | Protocol: Median held-out perplexity by position class in Chr5:3,500,000-4,500,000 Dataset subset: Positions of Arabidopsis thaliana Chr5:3,500,000-4,500,000 outside repeats, by class | 2.44 Perplexity (median over CDS1 positions in Chr5:3,500,000-4,500,000, repeats excluded) unitless · lower Uncertainty: Not reported by the source: The source prints no interval or spread for this value. Coverage: Not reported scored / Not reported eligible | Author-reported evaluation · Source checkedMethods, coverage and sourceMedian held-out perplexity by position class, Chr5:3,500,000-4,500,000 (PNAS 2023) gpn-paper-20261116-protocol-positional-perplexity-chr5-1mb Aggregation: Median over positions of the class DNA language models are powerful predictors of genome-wide variant effects (PNAS 2023) · Results, 'DNA Motifs Revealed by High-Confidence Model Predictions', sentence beginning "Across a 1-Mb region in the test chromosome", term "CDS1 (median = 2.44)" |
| Configuration: GPN trained on eight Brassicales genomes, repeat loss weight 0.1 (PNAS 2023) | Protocol: Median held-out perplexity by position class in Chr5:3,500,000-4,500,000 Dataset subset: Positions of Arabidopsis thaliana Chr5:3,500,000-4,500,000 outside repeats, by class | 2.24 Perplexity (median over CDS2 positions in Chr5:3,500,000-4,500,000, repeats excluded) unitless · lower Uncertainty: Not reported by the source: The source prints no interval or spread for this value. Coverage: Not reported scored / Not reported eligible | Author-reported evaluation · Source checkedMethods, coverage and sourceMedian held-out perplexity by position class, Chr5:3,500,000-4,500,000 (PNAS 2023) gpn-paper-20261116-protocol-positional-perplexity-chr5-1mb Aggregation: Median over positions of the class DNA language models are powerful predictors of genome-wide variant effects (PNAS 2023) · Results, 'DNA Motifs Revealed by High-Confidence Model Predictions', sentence beginning "Across a 1-Mb region in the test chromosome", term "CDS2 (median = 2.24)" |
| Configuration: GPN trained on eight Brassicales genomes, repeat loss weight 0.1 (PNAS 2023) | Protocol: Median held-out perplexity by position class in Chr5:3,500,000-4,500,000 Dataset subset: Positions of Arabidopsis thaliana Chr5:3,500,000-4,500,000 outside repeats, by class | 2.79 Perplexity (median over CDS3 positions in Chr5:3,500,000-4,500,000, repeats excluded) unitless · lower Uncertainty: Not reported by the source: The source prints no interval or spread for this value. Coverage: Not reported scored / Not reported eligible | Author-reported evaluation · Source checkedMethods, coverage and sourceMedian held-out perplexity by position class, Chr5:3,500,000-4,500,000 (PNAS 2023) gpn-paper-20261116-protocol-positional-perplexity-chr5-1mb Aggregation: Median over positions of the class DNA language models are powerful predictors of genome-wide variant effects (PNAS 2023) · Results, 'DNA Motifs Revealed by High-Confidence Model Predictions', sentence beginning "Across a 1-Mb region in the test chromosome", term "CDS3 (median = 2.79)" |
| Configuration: GPN trained on eight Brassicales genomes, repeat loss weight 0.1 (PNAS 2023) | Protocol: Median held-out perplexity by position class in Chr5:3,500,000-4,500,000 Dataset subset: Positions of Arabidopsis thaliana Chr5:3,500,000-4,500,000 outside repeats, by class | 3.24 Perplexity (median over intergenic and intronic positions in Chr5:3,500,000-4,500,000, repeats excluded) unitless · lower Uncertainty: Not reported by the source: The source prints no interval or spread for this value. Coverage: Not reported scored / Not reported eligible | Author-reported evaluation · Source checkedMethods, coverage and sourceMedian held-out perplexity by position class, Chr5:3,500,000-4,500,000 (PNAS 2023) gpn-paper-20261116-protocol-positional-perplexity-chr5-1mb Aggregation: Median over positions of the class DNA language models are powerful predictors of genome-wide variant effects (PNAS 2023) · Results, 'DNA Motifs Revealed by High-Confidence Model Predictions', sentence beginning "Across a 1-Mb region in the test chromosome", term "intergenic and intronic (median = 3.24)" |
| Configuration: GPN trained on eight Brassicales genomes, repeat loss weight 0.1 (PNAS 2023) | Protocol: Median held-out perplexity by position class in Chr5:3,500,000-4,500,000 Dataset subset: Positions of Arabidopsis thaliana Chr5:3,500,000-4,500,000 outside repeats, by class | 1.03 Perplexity (median over splice acceptor positions in Chr5:3,500,000-4,500,000, repeats excluded) unitless · lower Uncertainty: Not reported by the source: The source prints no interval or spread for this value. Coverage: Not reported scored / Not reported eligible | Author-reported evaluation · Source checkedMethods, coverage and sourceMedian held-out perplexity by position class, Chr5:3,500,000-4,500,000 (PNAS 2023) gpn-paper-20261116-protocol-positional-perplexity-chr5-1mb Aggregation: Median over positions of the class DNA language models are powerful predictors of genome-wide variant effects (PNAS 2023) · Results, 'DNA Motifs Revealed by High-Confidence Model Predictions', sentence beginning "Across a 1-Mb region in the test chromosome", term "splice acceptor (median = 1.03)" |
| Configuration: GPN trained on eight Brassicales genomes, repeat loss weight 0.1 (PNAS 2023) | Protocol: Median held-out perplexity by position class in Chr5:3,500,000-4,500,000 Dataset subset: Positions of Arabidopsis thaliana Chr5:3,500,000-4,500,000 outside repeats, by class | 1.02 Perplexity (median over splice donor positions in Chr5:3,500,000-4,500,000, repeats excluded) unitless · lower Uncertainty: Not reported by the source: The source prints no interval or spread for this value. Coverage: Not reported scored / Not reported eligible | Author-reported evaluation · Source checkedMethods, coverage and sourceMedian held-out perplexity by position class, Chr5:3,500,000-4,500,000 (PNAS 2023) gpn-paper-20261116-protocol-positional-perplexity-chr5-1mb Aggregation: Median over positions of the class DNA language models are powerful predictors of genome-wide variant effects (PNAS 2023) · Results, 'DNA Motifs Revealed by High-Confidence Model Predictions', sentence beginning "Across a 1-Mb region in the test chromosome", term "splice donor (median = 1.02)" |
| Configuration: GPN trained on eight Brassicales genomes, repeat loss weight 0.1 (PNAS 2023) | Protocol: Median held-out perplexity by position class in Chr5:3,500,000-4,500,000 Dataset subset: Positions of Arabidopsis thaliana Chr5:3,500,000-4,500,000 outside repeats, by class | 1.08 Perplexity (median over start codon positions in Chr5:3,500,000-4,500,000, repeats excluded) unitless · lower Uncertainty: Not reported by the source: The source prints no interval or spread for this value. Coverage: Not reported scored / Not reported eligible | Author-reported evaluation · Source checkedMethods, coverage and sourceMedian held-out perplexity by position class, Chr5:3,500,000-4,500,000 (PNAS 2023) gpn-paper-20261116-protocol-positional-perplexity-chr5-1mb Aggregation: Median over positions of the class DNA language models are powerful predictors of genome-wide variant effects (PNAS 2023) · Results, 'DNA Motifs Revealed by High-Confidence Model Predictions', sentence beginning "Across a 1-Mb region in the test chromosome", term "start codon (median = 1.08)" |
| Configuration: GPN trained on eight Brassicales genomes, repeat loss weight 0.1 (PNAS 2023) | Protocol: Median held-out perplexity by position class in Chr5:3,500,000-4,500,000 Dataset subset: Positions of Arabidopsis thaliana Chr5:3,500,000-4,500,000 outside repeats, by class | 2.8 Perplexity (median over stop codon positions in Chr5:3,500,000-4,500,000, repeats excluded) unitless · lower Uncertainty: Not reported by the source: The source prints no interval or spread for this value. Coverage: Not reported scored / Not reported eligible | Author-reported evaluation · Source checkedMethods, coverage and sourceMedian held-out perplexity by position class, Chr5:3,500,000-4,500,000 (PNAS 2023) gpn-paper-20261116-protocol-positional-perplexity-chr5-1mb Aggregation: Median over positions of the class DNA language models are powerful predictors of genome-wide variant effects (PNAS 2023) · Results, 'DNA Motifs Revealed by High-Confidence Model Predictions', sentence beginning "Across a 1-Mb region in the test chromosome", term "stop codon (median = 2.8)" |
| Configuration: GPN trained on eight Brassicales genomes, repeat loss weight 0.1 (PNAS 2023) | Protocol: Rare versus common variant enrichment in the lowest 0.1% of scores, 1001 Genomes Dataset: 1001 Genomes Arabidopsis thaliana SNPs with allele frequencies | 5.5-fold Odds ratio (rare (allele count 1) against common (allele frequency at least 5%) 1001 Genomes variants, putative functional score threshold at the lowest 0.1% quantile) unitless · higher Uncertainty: Not reported by the source: The source prints no interval or spread for this value. Coverage: Not reported scored / Not reported eligible | Author-reported evaluation · Source checkedMethods, coverage and sourceRare versus common variant enrichment for GPN, 1001 Genomes (PNAS 2023) gpn-paper-20261116-protocol-rare-vs-common-odds-ratio Aggregation: Not reported DNA language models are powerful predictors of genome-wide variant effects (PNAS 2023) · Results, 'Benchmarking using allele frequencies in 1001 genomes', sentence beginning "Putative functional SNPs, defined as the lowest 0.1% of GPN scores"; the same value is drawn in Figure 5 C |
| Configuration: GPN trained on eight Brassicales genomes, repeat loss weight 0.1 (PNAS 2023) | Protocol: Share of simulated repeat variants scored below the first decile of missense variants Dataset: All possible SNPs in Arabidopsis thaliana Chr5:3,500,000-4,500,000, annotated by consequence | 9% Proportion (simulated repeat variants ranked before the first decile of simulated missense variants, GPN trained with repeat weight 0.1) percent · lower Uncertainty: Not reported by the source: The source prints no interval or spread for this value. Coverage: Not reported scored / Not reported eligible | Author-reported evaluation · Source checkedMethods, coverage and sourceRepeat variants below the missense first decile, GPN with repeat weight 0.1 (PNAS 2023) gpn-paper-20261116-protocol-repeat-variants-below-missense-decile Aggregation: Not reported DNA language models are powerful predictors of genome-wide variant effects (PNAS 2023) · Results, 'In silico mutagenesis', sentence beginning "More precisely, in models with 0.0 and 0.1 down-weighting", the 9% of the 0.1 down-weighting model |
Source checking is not independent reproduction. Release 2026-11-21-c8bea25e5abb.
Use this configuration
How it works, versions and access
Additional methodological details, versions and access terms have not yet been extracted into a structured profile for this configuration. Consult the linked sources and evaluation records below.
Evidence
Source checking verifies the cited claim or transcription. It does not establish independent reproduction.
Evidence table
Inspect claims, sources and review details
Trace each statement to its source and review. A context-only reference supports the record generally; it does not verify an individual field. Source checking does not reproduce an experiment.
One row per statement and cited source. Multiple citations are not independent evaluations. Shared locators are labelled explicitly.
0 evidence rows matching the loaded filters
| Property and statement | Original source and location | Review and provenance |
|---|
No evidence rows match these filters. Choose another scope or clear the search.
Sources and history
Release 2026-11-21-c8bea25e5abb · Record review: source checked
1 source record and release history
- DNA language models are powerful predictors of genome-wide variant effects (PNAS 2023) · Original source · PNAS 120(44) e2311219120, published 26 October 2023; Europe PMC full text JATS XML of PMC10622914
Technical metadata and record identifiers
Technical metadata for this record is tracked under stable ID gpn-paper-20261116-config-gpn-brassicales-repeat-weight-0-1.
Source locator: Results, 'Training a Multispecies DNA Language Model'; Materials and Methods, 'Pretraining'; SI Appendix Tables S1 to S3
Related records
- configuration of: GPN
- family: GPN
- subject: Share of variants that alignment-based scores cannot reach
- subject: Correlation between GPN scores and phastCons
- subject: Correlation between GPN scores and phyloP
- subject: GPN training compute as published
- subject: GPN pretraining corpus and held-out chromosomes
- system: Gene region classification from averaged GPN embeddings (PNAS 2023)
- system: Test perplexity, chromosome-wide, GPN with repeat weight 0.1 (PNAS 2023 Table S2)
- system: Test perplexity, non-repeat, GPN with repeat weight 0.1 (PNAS 2023 Table S2)
- system: Median held-out perplexity by position class, Chr5:3,500,000-4,500,000 (PNAS 2023)
- system: Rare versus common variant enrichment for GPN, 1001 Genomes (PNAS 2023)
- system: Repeat variants below the missense first decile, GPN with repeat weight 0.1 (PNAS 2023)
How to cite
GPN trained on eight Brassicales genomes, repeat loss weight 0.1 (PNAS 2023). Rewire biological model benchmark database, release 2026-11-21-c8bea25e5abb. https://benchmarks.rewirebio.io/database/configuration/gpn-paper-20261116-config-gpn-brassicales-repeat-weight-0-1/
Release 2026-11-21-c8bea25e5abb. Rankings and values change between releases, so keep the release id in the citation. Cite the original publication as well for any value you reuse: each result links its source.
The compilation and annotations made by Rewire in this release are offered under CC BY 4.0; values and text taken from cited sources stay under the terms of those sources and are not covered by this licence. Data licence