Source

AlphaGenome Atlas: in silico mutagenesis of the entire human genome (Cheng, Taylor, Nicolaisen et al., medRxiv 2026)

medRxiv preprint version 1, posted 2026-09-20; full-text PDF with supplementary figures and tables (83 pages) · Preprint

DOI: 10.64898/2026.09.16.26363192

What this source contributed

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

1 method
24 configurations
1 pipeline
13 protocols
6 datasets
7 dataset subsets
109 evaluations
109 results
5 evidence claims

Results

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

Model or configuration	Metric	Value	Dataset	Evaluation
AlphaGenome merged splicing score, distilled model, Atlas precomputed (AlphaGenome Atlas v1)	Average precision (DNM1 minigene screen, active vs inactive variants)	0.943 unitless	DNM1 intron saturation mutagenesis minigene splicing screen (AlphaGenome Atlas Data S1)	AG Splicing on the DNM1 minigene splicing screen (AlphaGenome Atlas fig. S9E)
AVI ablation '+ No Indel Type' (Features Set to Zero, AlphaGenome Atlas Table S2)	AUPRC relative difference (ablation '+ No Indel Type (second occurrence, below 'No AlphaMissense Indel Features')' (Features Set to Zero) relative to the AVI baseline row, ClinVar column)	-0.40% percent	ClinVar 2025-06-15 pathogenic versus benign variants, AVI evaluation set	AVI + No Indel Type (Features Set to Zero) on ClinVar (AlphaGenome Atlas Table S2)
AVI ablation '+ No Indel Type' (Features Set to Zero, AlphaGenome Atlas Table S2)	AUPRC relative difference (ablation '+ No Indel Type' (Features Set to Zero) relative to the AVI baseline row, ClinVar column)	0.00% percent	ClinVar 2025-06-15 pathogenic versus benign variants, AVI evaluation set	AVI + No Indel Type (Features Set to Zero) on ClinVar (AlphaGenome Atlas Table S2)
AVI ablation '+ No Indel Type' (Features Set to Zero, AlphaGenome Atlas Table S2)	AUPRC relative difference (ablation '+ No Indel Type (second occurrence, below 'No AlphaMissense Indel Features')' (Features Set to Zero) relative to the AVI baseline row, ncVar column)	0.06% percent	ncVarDB non-coding pathogenic and benign variants (as used in AlphaGenome Atlas)	AVI + No Indel Type (Features Set to Zero) on ncVar (AlphaGenome Atlas Table S2)
AVI ablation '+ No Indel Type' (Features Set to Zero, AlphaGenome Atlas Table S2)	AUPRC relative difference (ablation '+ No Indel Type' (Features Set to Zero) relative to the AVI baseline row, ncVar column)	0.06% percent	ncVarDB non-coding pathogenic and benign variants (as used in AlphaGenome Atlas)	AVI + No Indel Type (Features Set to Zero) on ncVar (AlphaGenome Atlas Table S2)
AVI ablation '+ No Indel Type' (Features Set to Zero, AlphaGenome Atlas Table S2)	AUPRC relative difference (ablation '+ No Indel Type (second occurrence, below 'No AlphaMissense Indel Features')' (Features Set to Zero) relative to the AVI baseline row, SGE column)	-5.14% percent	Saturation genome editing screens of 13 genes from 14 studies (AlphaGenome Atlas)	AVI + No Indel Type (Features Set to Zero) on SGE (AlphaGenome Atlas Table S2)
AVI ablation '+ No Indel Type' (Features Set to Zero, AlphaGenome Atlas Table S2)	AUPRC relative difference (ablation '+ No Indel Type' (Features Set to Zero) relative to the AVI baseline row, SGE column)	-0.55% percent	Saturation genome editing screens of 13 genes from 14 studies (AlphaGenome Atlas)	AVI + No Indel Type (Features Set to Zero) on SGE (AlphaGenome Atlas Table S2)
AVI ablation '+ No Indel Type' (Features Set to Zero, AlphaGenome Atlas Table S2)	AUPRC relative difference (ablation '+ No Indel Type (second occurrence, below 'No AlphaMissense Indel Features')' (Features Set to Zero) relative to the AVI baseline row, TraitGym Complex column)	0.00% percent	TraitGym Complex traits after AVI positional filtering (AlphaGenome Atlas)	AVI + No Indel Type (Features Set to Zero) on TraitGym Complex (AlphaGenome Atlas Table S2)
AVI ablation '+ No Indel Type' (Features Set to Zero, AlphaGenome Atlas Table S2)	AUPRC relative difference (ablation '+ No Indel Type' (Features Set to Zero) relative to the AVI baseline row, TraitGym Complex column)	0.00% percent	TraitGym Complex traits after AVI positional filtering (AlphaGenome Atlas)	AVI + No Indel Type (Features Set to Zero) on TraitGym Complex (AlphaGenome Atlas Table S2)
AVI ablation '+ No Indel Type' (Features Set to Zero, AlphaGenome Atlas Table S2)	AUPRC relative difference (ablation '+ No Indel Type (second occurrence, below 'No AlphaMissense Indel Features')' (Features Set to Zero) relative to the AVI baseline row, TraitGym Mendelian column)	0.00% percent	TraitGym Mendelian traits after AVI positional filtering (AlphaGenome Atlas)	AVI + No Indel Type (Features Set to Zero) on TraitGym Mendelian (AlphaGenome Atlas Table S2)
AVI ablation '+ No Indel Type' (Features Set to Zero, AlphaGenome Atlas Table S2)	AUPRC relative difference (ablation '+ No Indel Type' (Features Set to Zero) relative to the AVI baseline row, TraitGym Mendelian column)	0.00% percent	TraitGym Mendelian traits after AVI positional filtering (AlphaGenome Atlas)	AVI + No Indel Type (Features Set to Zero) on TraitGym Mendelian (AlphaGenome Atlas Table S2)
AVI ablation 'All except AlphaGenome' (Retrained, AlphaGenome Atlas Table S2)	AUPRC relative difference (ablation 'All except AlphaGenome' (Retrained) relative to the AVI baseline row, ClinVar column)	-54.88% percent	ClinVar 2025-06-15 pathogenic versus benign variants, AVI evaluation set	AVI All except AlphaGenome (Retrained) on ClinVar (AlphaGenome Atlas Table S2)
AVI ablation 'All except AlphaGenome' (Retrained, AlphaGenome Atlas Table S2)	AUPRC relative difference (ablation 'All except AlphaGenome' (Retrained) relative to the AVI baseline row, ncVar column)	-3.82% percent	ncVarDB non-coding pathogenic and benign variants (as used in AlphaGenome Atlas)	AVI All except AlphaGenome (Retrained) on ncVar (AlphaGenome Atlas Table S2)
AVI ablation 'All except AlphaGenome' (Retrained, AlphaGenome Atlas Table S2)	AUPRC relative difference (ablation 'All except AlphaGenome' (Retrained) relative to the AVI baseline row, SGE column)	-47.15% percent	Saturation genome editing screens of 13 genes from 14 studies (AlphaGenome Atlas)	AVI All except AlphaGenome (Retrained) on SGE (AlphaGenome Atlas Table S2)
AVI ablation 'All except AlphaGenome' (Retrained, AlphaGenome Atlas Table S2)	AUPRC relative difference (ablation 'All except AlphaGenome' (Retrained) relative to the AVI baseline row, TraitGym Complex column)	-27.73% percent	TraitGym Complex traits after AVI positional filtering (AlphaGenome Atlas)	AVI All except AlphaGenome (Retrained) on TraitGym Complex (AlphaGenome Atlas Table S2)
AVI ablation 'All except AlphaGenome' (Retrained, AlphaGenome Atlas Table S2)	AUPRC relative difference (ablation 'All except AlphaGenome' (Retrained) relative to the AVI baseline row, TraitGym Mendelian column)	-38.95% percent	TraitGym Mendelian traits after AVI positional filtering (AlphaGenome Atlas)	AVI All except AlphaGenome (Retrained) on TraitGym Mendelian (AlphaGenome Atlas Table S2)
AVI ablation 'AlphaGenome' (Retrained, AlphaGenome Atlas Table S2)	AUPRC relative difference (ablation 'AlphaGenome' (Retrained) relative to the AVI baseline row, ClinVar column)	-2.56% percent	ClinVar 2025-06-15 pathogenic versus benign variants, AVI evaluation set	AVI AlphaGenome (Retrained) on ClinVar (AlphaGenome Atlas Table S2)
AVI ablation 'AlphaGenome' (Retrained, AlphaGenome Atlas Table S2)	AUPRC relative difference (ablation 'AlphaGenome' (Retrained) relative to the AVI baseline row, ncVar column)	-17.61% percent	ncVarDB non-coding pathogenic and benign variants (as used in AlphaGenome Atlas)	AVI AlphaGenome (Retrained) on ncVar (AlphaGenome Atlas Table S2)
AVI ablation 'AlphaGenome' (Retrained, AlphaGenome Atlas Table S2)	AUPRC relative difference (ablation 'AlphaGenome' (Retrained) relative to the AVI baseline row, SGE column)	-8.29% percent	Saturation genome editing screens of 13 genes from 14 studies (AlphaGenome Atlas)	AVI AlphaGenome (Retrained) on SGE (AlphaGenome Atlas Table S2)
AVI ablation 'AlphaGenome' (Retrained, AlphaGenome Atlas Table S2)	AUPRC relative difference (ablation 'AlphaGenome' (Retrained) relative to the AVI baseline row, TraitGym Complex column)	-17.27% percent	TraitGym Complex traits after AVI positional filtering (AlphaGenome Atlas)	AVI AlphaGenome (Retrained) on TraitGym Complex (AlphaGenome Atlas Table S2)
AVI ablation 'AlphaGenome' (Retrained, AlphaGenome Atlas Table S2)	AUPRC relative difference (ablation 'AlphaGenome' (Retrained) relative to the AVI baseline row, TraitGym Mendelian column)	-2.83% percent	TraitGym Mendelian traits after AVI positional filtering (AlphaGenome Atlas)	AVI AlphaGenome (Retrained) on TraitGym Mendelian (AlphaGenome Atlas Table S2)
AVI ablation 'AlphaMissense' (Retrained, AlphaGenome Atlas Table S2)	AUPRC relative difference (ablation 'AlphaMissense' (Retrained) relative to the AVI baseline row, ClinVar column)	-2.51% percent	ClinVar 2025-06-15 pathogenic versus benign variants, AVI evaluation set	AVI AlphaMissense (Retrained) on ClinVar (AlphaGenome Atlas Table S2)
AVI ablation 'AlphaMissense' (Retrained, AlphaGenome Atlas Table S2)	AUPRC relative difference (ablation 'AlphaMissense' (Retrained) relative to the AVI baseline row, ncVar column)	0.18% percent	ncVarDB non-coding pathogenic and benign variants (as used in AlphaGenome Atlas)	AVI AlphaMissense (Retrained) on ncVar (AlphaGenome Atlas Table S2)
AVI ablation 'AlphaMissense' (Retrained, AlphaGenome Atlas Table S2)	AUPRC relative difference (ablation 'AlphaMissense' (Retrained) relative to the AVI baseline row, SGE column)	-0.45% percent	Saturation genome editing screens of 13 genes from 14 studies (AlphaGenome Atlas)	AVI AlphaMissense (Retrained) on SGE (AlphaGenome Atlas Table S2)
AVI ablation 'AlphaMissense' (Retrained, AlphaGenome Atlas Table S2)	AUPRC relative difference (ablation 'AlphaMissense' (Retrained) relative to the AVI baseline row, TraitGym Complex column)	0.67% percent	TraitGym Complex traits after AVI positional filtering (AlphaGenome Atlas)	AVI AlphaMissense (Retrained) on TraitGym Complex (AlphaGenome Atlas Table S2)
AVI ablation 'AlphaMissense' (Retrained, AlphaGenome Atlas Table S2)	AUPRC relative difference (ablation 'AlphaMissense' (Retrained) relative to the AVI baseline row, TraitGym Mendelian column)	0.75% percent	TraitGym Mendelian traits after AVI positional filtering (AlphaGenome Atlas)	AVI AlphaMissense (Retrained) on TraitGym Mendelian (AlphaGenome Atlas Table S2)
AVI six-model ensemble (AlphaGenome Atlas v1)	AUPRC (AVI baseline row, ClinVar column)	0.978 unitless	ClinVar 2025-06-15 pathogenic versus benign variants, AVI evaluation set	AVI AVI Model (Baseline) on ClinVar (AlphaGenome Atlas Table S2)
AVI six-model ensemble (AlphaGenome Atlas v1)	AUPRC (AVI baseline row, ncVar column)	0.936 unitless	ncVarDB non-coding pathogenic and benign variants (as used in AlphaGenome Atlas)	AVI AVI Model (Baseline) on ncVar (AlphaGenome Atlas Table S2)
AVI six-model ensemble (AlphaGenome Atlas v1)	AUPRC (AVI baseline row, SGE column)	0.731 unitless	Saturation genome editing screens of 13 genes from 14 studies (AlphaGenome Atlas)	AVI AVI Model (Baseline) on SGE (AlphaGenome Atlas Table S2)
AVI six-model ensemble (AlphaGenome Atlas v1)	AUPRC (AVI baseline row, TraitGym Complex column)	0.283 unitless	TraitGym Complex traits after AVI positional filtering (AlphaGenome Atlas)	AVI AVI Model (Baseline) on TraitGym Complex (AlphaGenome Atlas Table S2)
AVI six-model ensemble (AlphaGenome Atlas v1)	AUPRC (AVI baseline row, TraitGym Mendelian column)	0.764 unitless	TraitGym Mendelian traits after AVI positional filtering (AlphaGenome Atlas)	AVI AVI Model (Baseline) on TraitGym Mendelian (AlphaGenome Atlas Table S2)
AVI ablation 'Cactus' (Retrained, AlphaGenome Atlas Table S2)	AUPRC relative difference (ablation 'Cactus' (Retrained) relative to the AVI baseline row, ClinVar column)	-0.13% percent	ClinVar 2025-06-15 pathogenic versus benign variants, AVI evaluation set	AVI Cactus (Retrained) on ClinVar (AlphaGenome Atlas Table S2)
AVI ablation 'Cactus' (Retrained, AlphaGenome Atlas Table S2)	AUPRC relative difference (ablation 'Cactus' (Retrained) relative to the AVI baseline row, ncVar column)	-0.91% percent	ncVarDB non-coding pathogenic and benign variants (as used in AlphaGenome Atlas)	AVI Cactus (Retrained) on ncVar (AlphaGenome Atlas Table S2)
AVI ablation 'Cactus' (Retrained, AlphaGenome Atlas Table S2)	AUPRC relative difference (ablation 'Cactus' (Retrained) relative to the AVI baseline row, SGE column)	-1.32% percent	Saturation genome editing screens of 13 genes from 14 studies (AlphaGenome Atlas)	AVI Cactus (Retrained) on SGE (AlphaGenome Atlas Table S2)
AVI ablation 'Cactus' (Retrained, AlphaGenome Atlas Table S2)	AUPRC relative difference (ablation 'Cactus' (Retrained) relative to the AVI baseline row, TraitGym Complex column)	-6.64% percent	TraitGym Complex traits after AVI positional filtering (AlphaGenome Atlas)	AVI Cactus (Retrained) on TraitGym Complex (AlphaGenome Atlas Table S2)
AVI ablation 'Cactus' (Retrained, AlphaGenome Atlas Table S2)	AUPRC relative difference (ablation 'Cactus' (Retrained) relative to the AVI baseline row, TraitGym Mendelian column)	-16.47% percent	TraitGym Mendelian traits after AVI positional filtering (AlphaGenome Atlas)	AVI Cactus (Retrained) on TraitGym Mendelian (AlphaGenome Atlas Table S2)
AVI ablation 'Consequence' (Retrained, AlphaGenome Atlas Table S2)	AUPRC relative difference (ablation 'Consequence' (Retrained) relative to the AVI baseline row, ClinVar column)	-13.71% percent	ClinVar 2025-06-15 pathogenic versus benign variants, AVI evaluation set	AVI Consequence (Retrained) on ClinVar (AlphaGenome Atlas Table S2)
AVI ablation 'Consequence' (Retrained, AlphaGenome Atlas Table S2)	AUPRC relative difference (ablation 'Consequence' (Retrained) relative to the AVI baseline row, ncVar column)	0.10% percent	ncVarDB non-coding pathogenic and benign variants (as used in AlphaGenome Atlas)	AVI Consequence (Retrained) on ncVar (AlphaGenome Atlas Table S2)
AVI ablation 'Consequence' (Retrained, AlphaGenome Atlas Table S2)	AUPRC relative difference (ablation 'Consequence' (Retrained) relative to the AVI baseline row, SGE column)	-27.42% percent	Saturation genome editing screens of 13 genes from 14 studies (AlphaGenome Atlas)	AVI Consequence (Retrained) on SGE (AlphaGenome Atlas Table S2)
AVI ablation 'Consequence' (Retrained, AlphaGenome Atlas Table S2)	AUPRC relative difference (ablation 'Consequence' (Retrained) relative to the AVI baseline row, TraitGym Complex column)	0.28% percent	TraitGym Complex traits after AVI positional filtering (AlphaGenome Atlas)	AVI Consequence (Retrained) on TraitGym Complex (AlphaGenome Atlas Table S2)
AVI ablation 'Consequence' (Retrained, AlphaGenome Atlas Table S2)	AUPRC relative difference (ablation 'Consequence' (Retrained) relative to the AVI baseline row, TraitGym Mendelian column)	0.26% percent	TraitGym Mendelian traits after AVI positional filtering (AlphaGenome Atlas)	AVI Consequence (Retrained) on TraitGym Mendelian (AlphaGenome Atlas Table S2)
AVI ablation 'Conservation' (Retrained, AlphaGenome Atlas Table S2)	AUPRC relative difference (ablation 'Conservation' (Retrained) relative to the AVI baseline row, ClinVar column)	-0.58% percent	ClinVar 2025-06-15 pathogenic versus benign variants, AVI evaluation set	AVI Conservation (Retrained) on ClinVar (AlphaGenome Atlas Table S2)
AVI ablation 'Conservation' (Retrained, AlphaGenome Atlas Table S2)	AUPRC relative difference (ablation 'Conservation' (Retrained) relative to the AVI baseline row, ncVar column)	-4.26% percent	ncVarDB non-coding pathogenic and benign variants (as used in AlphaGenome Atlas)	AVI Conservation (Retrained) on ncVar (AlphaGenome Atlas Table S2)
AVI ablation 'Conservation' (Retrained, AlphaGenome Atlas Table S2)	AUPRC relative difference (ablation 'Conservation' (Retrained) relative to the AVI baseline row, SGE column)	-2.36% percent	Saturation genome editing screens of 13 genes from 14 studies (AlphaGenome Atlas)	AVI Conservation (Retrained) on SGE (AlphaGenome Atlas Table S2)
AVI ablation 'Conservation' (Retrained, AlphaGenome Atlas Table S2)	AUPRC relative difference (ablation 'Conservation' (Retrained) relative to the AVI baseline row, TraitGym Complex column)	-26.25% percent	TraitGym Complex traits after AVI positional filtering (AlphaGenome Atlas)	AVI Conservation (Retrained) on TraitGym Complex (AlphaGenome Atlas Table S2)
AVI ablation 'Conservation' (Retrained, AlphaGenome Atlas Table S2)	AUPRC relative difference (ablation 'Conservation' (Retrained) relative to the AVI baseline row, TraitGym Mendelian column)	-37.96% percent	TraitGym Mendelian traits after AVI positional filtering (AlphaGenome Atlas)	AVI Conservation (Retrained) on TraitGym Mendelian (AlphaGenome Atlas Table S2)
AVI ablation 'Indel Type' (Retrained, AlphaGenome Atlas Table S2)	AUPRC relative difference (ablation 'Indel Type' (Retrained) relative to the AVI baseline row, ClinVar column)	-0.05% percent	ClinVar 2025-06-15 pathogenic versus benign variants, AVI evaluation set	AVI Indel Type (Retrained) on ClinVar (AlphaGenome Atlas Table S2)
AVI ablation 'Indel Type' (Retrained, AlphaGenome Atlas Table S2)	AUPRC relative difference (ablation 'Indel Type' (Retrained) relative to the AVI baseline row, ncVar column)	-0.44% percent	ncVarDB non-coding pathogenic and benign variants (as used in AlphaGenome Atlas)	AVI Indel Type (Retrained) on ncVar (AlphaGenome Atlas Table S2)
AVI ablation 'Indel Type' (Retrained, AlphaGenome Atlas Table S2)	AUPRC relative difference (ablation 'Indel Type' (Retrained) relative to the AVI baseline row, SGE column)	0.08% percent	Saturation genome editing screens of 13 genes from 14 studies (AlphaGenome Atlas)	AVI Indel Type (Retrained) on SGE (AlphaGenome Atlas Table S2)
AVI ablation 'Indel Type' (Retrained, AlphaGenome Atlas Table S2)	AUPRC relative difference (ablation 'Indel Type' (Retrained) relative to the AVI baseline row, TraitGym Complex column)	-2.19% percent	TraitGym Complex traits after AVI positional filtering (AlphaGenome Atlas)	AVI Indel Type (Retrained) on TraitGym Complex (AlphaGenome Atlas Table S2)

Claims

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

AlphaGenome Atlas evaluations are developer-run: Evaluation actor All AVI and comparator evaluations were run by the AlphaGenome Atlas authors, led by Google DeepMind, which also develops AlphaGenome and AlphaMissense. Benchmark variants overlapping AVI training or validation positions are removed for all methods; comparator training sets were not filtered. (Author list and affiliations (page 1); Methods 'Overlap Filtering during Evaluation' (page 22))
AlphaGenome Atlas values not extracted: Extraction gap Not extracted: Fig. 2A and 2B bar labels other than those the text names (bars sorted by value and identified only by colour, so ClinVar SNV and indel overall AUPRCs and most comparator values are unattributed); fig. S7A ClinVar UTR, intronic and ncVar SNV-plus-indel bars (same problem); Fig. 2D and fig. S7D per-gene Spearman correlations (no printed values); fig. S7E and S7F precision at 90% recall (no printed values); Fig. 3A curve labels at K other than 50 (attributed by colour); TraitGym Complex comparator value 0.27 (the text does not name the method); all 95% bootstrap confidence intervals (drawn, not printed). (Fig. 2 (page 6), Fig. 3A (page 9), fig. S7 (page 57))
AlphaGenome Atlas: indel scoring with shift augmentation: Indel scoring Atlas scores indels with three forward passes: the reference-aligned pass, a pass shifted to align bases after the variant, and the unstitched allele; the base-resolution embeddings are stitched before the heads. Coarse 128 bp and 2,048 bp bins are shifted from the final bin the indel ends in. (Methods 'Indel Scoring', 'Indel shift augmentation', 'Base Resolution', 'Coarse Resolution' (pages 19 to 20))
AlphaGenome Atlas: variant offset from the input centre: Scoring setup Atlas scores use a 1 Mb input centred on each 128 bp window, so variants sit up to 64 bp from the centre. The authors re-ran the AlphaGenome paper evaluations with variants at their windowed positions and report no statistically significant difference in any evaluation metric. (Methods 'Sliding Window Approach' and 'Reference Prediction Caching' (pages 18 to 19))
Random-classifier AUPRC printed for the pooled SGE evaluations: Random baseline auprc fig. S7E legend prints 'Baseline (0.146)' for SNVs; fig. S7F legend prints 'Baseline (0.194)' for SNVs and indels. (fig. S7E and S7F legends (page 57))

Models, benchmarks and other records

Methods (1)

gnomAD allele frequency as a variant score

Configurations (24)

Pipelines (1)

AlphaGenome Variant Impact (AVI)

Protocols (13)

Datasets (6)

Dataset subsets (7)

Evaluations (109)

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.

22 evidence rows matching the loaded filters

Claims, original sources and review scope · Release 2026-11-02-a7f03c9e55a5
Property and statement	Original source and location	Review and provenance
Artifact SHA-256 b32d1539005ca4ea31c3dcef49b834424b1d8b5da4ce6b9fc3e1f942d39fc453 Source metadata	AlphaGenome Atlas: in silico mutagenesis of the entire human genome (Cheng, Taylor, Nicolaisen et al., medRxiv 2026) Original source ↗ No field-specific location recorded Version: medRxiv preprint version 1, posted 2026-09-20; full-text PDF with supplementary figures and tables (83 pages) Retrieved: 2026-10-19T12:19:03Z	catalogued No individual claim review recorded Audit details Field: `attributes.artifact_sha256` Source artifact SHA-256: `b32d1539005ca4ea31c3dcef49b834424b1d8b5da4ce6b9fc3e1f942d39fc453` Hash scope: SHA-256 of one retrieval of the medRxiv PDF. A second retrieval at 2026-10-19T12:30:54Z had the same length (15,415,633 bytes) and the same SHA-256. Inspected artifact
Artifact url https://www.medrxiv.org/content/10.64898/2026.09.16.26363192v1.full.pdf Source metadata	AlphaGenome Atlas: in silico mutagenesis of the entire human genome (Cheng, Taylor, Nicolaisen et al., medRxiv 2026) Original source ↗ No field-specific location recorded Version: medRxiv preprint version 1, posted 2026-09-20; full-text PDF with supplementary figures and tables (83 pages) Retrieved: 2026-10-19T12:19:03Z	catalogued No individual claim review recorded Audit details Field: `attributes.artifact_url` Source artifact SHA-256: `b32d1539005ca4ea31c3dcef49b834424b1d8b5da4ce6b9fc3e1f942d39fc453` Hash scope: SHA-256 of one retrieval of the medRxiv PDF. A second retrieval at 2026-10-19T12:30:54Z had the same length (15,415,633 bytes) and the same SHA-256. Inspected artifact
Doi 10.64898/2026.09.16.26363192 Source metadata	AlphaGenome Atlas: in silico mutagenesis of the entire human genome (Cheng, Taylor, Nicolaisen et al., medRxiv 2026) Original source ↗ No field-specific location recorded Version: medRxiv preprint version 1, posted 2026-09-20; full-text PDF with supplementary figures and tables (83 pages) Retrieved: 2026-10-19T12:19:03Z	catalogued No individual claim review recorded Audit details Field: `attributes.doi` Source artifact SHA-256: `b32d1539005ca4ea31c3dcef49b834424b1d8b5da4ce6b9fc3e1f942d39fc453` Hash scope: SHA-256 of one retrieval of the medRxiv PDF. A second retrieval at 2026-10-19T12:30:54Z had the same length (15,415,633 bytes) and the same SHA-256. Inspected artifact
Hash scope SHA-256 of one retrieval of the medRxiv PDF. A second retrieval at 2026-10-19T12:30:54Z had the same length (15,415,633 bytes) and the same SHA-256. Source metadata	AlphaGenome Atlas: in silico mutagenesis of the entire human genome (Cheng, Taylor, Nicolaisen et al., medRxiv 2026) Original source ↗ No field-specific location recorded Version: medRxiv preprint version 1, posted 2026-09-20; full-text PDF with supplementary figures and tables (83 pages) Retrieved: 2026-10-19T12:19:03Z	catalogued No individual claim review recorded Audit details Field: `attributes.hash_scope` Source artifact SHA-256: `b32d1539005ca4ea31c3dcef49b834424b1d8b5da4ce6b9fc3e1f942d39fc453` Hash scope: SHA-256 of one retrieval of the medRxiv PDF. A second retrieval at 2026-10-19T12:30:54Z had the same length (15,415,633 bytes) and the same SHA-256. Inspected artifact
Licence CC-BY-NC-4.0 Source metadata	AlphaGenome Atlas: in silico mutagenesis of the entire human genome (Cheng, Taylor, Nicolaisen et al., medRxiv 2026) Original source ↗ No field-specific location recorded Version: medRxiv preprint version 1, posted 2026-09-20; full-text PDF with supplementary figures and tables (83 pages) Retrieved: 2026-10-19T12:19:03Z	catalogued No individual claim review recorded Audit details Field: `attributes.licence` Source artifact SHA-256: `b32d1539005ca4ea31c3dcef49b834424b1d8b5da4ce6b9fc3e1f942d39fc453` Hash scope: SHA-256 of one retrieval of the medRxiv PDF. A second retrieval at 2026-10-19T12:30:54Z had the same length (15,415,633 bytes) and the same SHA-256. Inspected artifact
Limitations 5 values Preprint, not peer reviewed (medRxiv v1; the medRxiv API 'published' field was not checked against a journal version). Developer study: the AlphaGenome Atlas team is led by Google DeepMind, which also develops AlphaGenome and AlphaMissense. All evaluations, including those of comparator methods, were run by the authors. Fig. 2A and 2B print bar labels but identify the predictor of each bar only by colour; those labels are not extracted. Only values the text attributes to a named method, and figure legends that name the method, are recorded. No numeric confidence intervals are printed. Fig. 2 error bars are 95% bootstrap intervals (100 bootstraps) drawn without values. Data S1 (DNM1 screen) and the precomputed Atlas scores were not retrieved. Source metadata	AlphaGenome Atlas: in silico mutagenesis of the entire human genome (Cheng, Taylor, Nicolaisen et al., medRxiv 2026) Original source ↗ No field-specific location recorded Version: medRxiv preprint version 1, posted 2026-09-20; full-text PDF with supplementary figures and tables (83 pages) Retrieved: 2026-10-19T12:19:03Z	catalogued No individual claim review recorded Audit details Field: `attributes.limitations` Source artifact SHA-256: `b32d1539005ca4ea31c3dcef49b834424b1d8b5da4ce6b9fc3e1f942d39fc453` Hash scope: SHA-256 of one retrieval of the medRxiv PDF. A second retrieval at 2026-10-19T12:30:54Z had the same length (15,415,633 bytes) and the same SHA-256. Inspected artifact
Media type application/pdf Source metadata	AlphaGenome Atlas: in silico mutagenesis of the entire human genome (Cheng, Taylor, Nicolaisen et al., medRxiv 2026) Original source ↗ No field-specific location recorded Version: medRxiv preprint version 1, posted 2026-09-20; full-text PDF with supplementary figures and tables (83 pages) Retrieved: 2026-10-19T12:19:03Z	catalogued No individual claim review recorded Audit details Field: `attributes.media_type` Source artifact SHA-256: `b32d1539005ca4ea31c3dcef49b834424b1d8b5da4ce6b9fc3e1f942d39fc453` Hash scope: SHA-256 of one retrieval of the medRxiv PDF. A second retrieval at 2026-10-19T12:30:54Z had the same length (15,415,633 bytes) and the same SHA-256. Inspected artifact
Missing metadata: artifact member: note Single PDF; supplementary material is inside it. Source metadata	AlphaGenome Atlas: in silico mutagenesis of the entire human genome (Cheng, Taylor, Nicolaisen et al., medRxiv 2026) Original source ↗ No field-specific location recorded Version: medRxiv preprint version 1, posted 2026-09-20; full-text PDF with supplementary figures and tables (83 pages) Retrieved: 2026-10-19T12:19:03Z	catalogued No individual claim review recorded Audit details Field: `attributes.missing_metadata.artifact_member.note` Source artifact SHA-256: `b32d1539005ca4ea31c3dcef49b834424b1d8b5da4ce6b9fc3e1f942d39fc453` Hash scope: SHA-256 of one retrieval of the medRxiv PDF. A second retrieval at 2026-10-19T12:30:54Z had the same length (15,415,633 bytes) and the same SHA-256. Inspected artifact
Missing metadata: artifact member: reason inapplicable Source metadata	AlphaGenome Atlas: in silico mutagenesis of the entire human genome (Cheng, Taylor, Nicolaisen et al., medRxiv 2026) Original source ↗ No field-specific location recorded Version: medRxiv preprint version 1, posted 2026-09-20; full-text PDF with supplementary figures and tables (83 pages) Retrieved: 2026-10-19T12:19:03Z	catalogued No individual claim review recorded Audit details Field: `attributes.missing_metadata.artifact_member.reason` Source artifact SHA-256: `b32d1539005ca4ea31c3dcef49b834424b1d8b5da4ce6b9fc3e1f942d39fc453` Hash scope: SHA-256 of one retrieval of the medRxiv PDF. A second retrieval at 2026-10-19T12:30:54Z had the same length (15,415,633 bytes) and the same SHA-256. Inspected artifact
Publication status preprint Source metadata	AlphaGenome Atlas: in silico mutagenesis of the entire human genome (Cheng, Taylor, Nicolaisen et al., medRxiv 2026) Original source ↗ No field-specific location recorded Version: medRxiv preprint version 1, posted 2026-09-20; full-text PDF with supplementary figures and tables (83 pages) Retrieved: 2026-10-19T12:19:03Z	catalogued No individual claim review recorded Audit details Field: `attributes.publication_status` Source artifact SHA-256: `b32d1539005ca4ea31c3dcef49b834424b1d8b5da4ce6b9fc3e1f942d39fc453` Hash scope: SHA-256 of one retrieval of the medRxiv PDF. A second retrieval at 2026-10-19T12:30:54Z had the same length (15,415,633 bytes) and the same SHA-256. Inspected artifact

How to read the evidence

Sources and history

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

Primary source retrieved and hashed for the 2026-10-19 AlphaGenome variant evidence extraction.

Release history

No supporting source is linked yet.

Read original source

Download this release (gzip)

Technical metadata and extraction receipts

Stable ID: ag-atlas-20261019-source-preprint

areas: dna-genomes
contexts: research; clinical_research
url: https://doi.org/10.64898/2026.09.16.26363192
artifact url: https://www.medrxiv.org/content/10.64898/2026.09.16.26363192v1.full.pdf
doi: 10.64898/2026.09.16.26363192
version: medRxiv preprint version 1, posted 2026-09-20; full-text PDF with supplementary figures and tables (83 pages)
retrieved at: 2026-10-19T12:19:03Z
artifact sha256: b32d1539005ca4ea31c3dcef49b834424b1d8b5da4ce6b9fc3e1f942d39fc453
publication status: preprint
licence: CC-BY-NC-4.0
media type: application/pdf
source locator: Main text Fig. 2 and 3 and their Results paragraphs; Materials and methods; fig. S7 and S9; Tables S1 and S2
hash scope: SHA-256 of one retrieval of the medRxiv PDF. A second retrieval at 2026-10-19T12:30:54Z had the same length (15,415,633 bytes) and the same SHA-256.
limitations: Preprint, not peer reviewed (medRxiv v1; the medRxiv API 'published' field was not checked against a journal version).; Developer study: the AlphaGenome Atlas team is led by Google DeepMind, which also develops AlphaGenome and AlphaMissense. All evaluations, including those of comparator methods, were run by the authors.; Fig. 2A and 2B print bar labels but identify the predictor of each bar only by colour; those labels are not extracted. Only values the text attributes to a named method, and figure legends that name the method, are recorded.; No numeric confidence intervals are printed. Fig. 2 error bars are 95% bootstrap intervals (100 bootstraps) drawn without values.; Data S1 (DNM1 screen) and the precomputed Atlas scores were not retrieved.
missing metadata: artifact member: reason: inapplicable; note: Single PDF; supplementary material is inside it.

Related records

subject: AlphaGenome Atlas values not extracted

Suggest a correction