Source

Interpretable variant effect prediction from genomic foundation model representations (EVEE), bioRxiv v4

Evidence concern: excluded from comparisons

The column labelled 'NTv3' in Figure 1b and 1c is not the method the Methods describe. The manuscript repository's plotting code (scripts/figure1/_heatmap_common.py at e5e4a43) renames 'NTv3 cosine dist' to 'NTv3' and drops 'NTv3 subref probe', whereas Methods 'Baseline methods' describe NTv3 650M layer-11 embeddings with L2-regularised logistic regression under 5-fold gene-holdout cross-validation. Figure 1d's NTv3 column is the supervised subref probe. The SNV and deconfounded NTv3 values are recorded as a zero-shot embedding distance.Interpretable variant effect prediction from genomic foundation model representations (EVEE), bioRxiv v4 · Figure 1b and 1c, column 'NTv3' (page 3); Methods 'Baseline methods', NTv3 (page 8); repository scripts/figure1/_heatmap_common.py RENAME_MAP and DROP_METHODS

AlphaMissense's 'Overall' cell in Figure 1b and 1c is its missense AUROC. The plotting code blanks AlphaMissense outside missense and copies the Missense value into the Overall row (scripts/figure1/_heatmap_common.py, am_ov_mask), so it is computed on a different, missense-only population from the other methods' Overall cells.Interpretable variant effect prediction from genomic foundation model representations (EVEE), bioRxiv v4 · Figure 1b and 1c, row 'Overall', column 'AlphaMissense' (page 3); repository scripts/figure1/_heatmap_common.py prepare()

bioRxiv v4, posted 2026-07-11 (v1 posted 2026-04-11) · Preprint

DOI: 10.64898/2026.04.10.717844

What this source contributed

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

1 method
9 configurations
3 protocols
3 datasets
13 evaluations
13 results
2 evidence claims

Results

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

Value	Metric	Model or configuration	Dataset	Evaluation
0.574 unitless	AUROC (Deconfounded ClinVar SNVs, all consequence types (Overall row))	AlphaGenome all-folds composite score (Pearce et al. 2026)	EVEE deconfounded ClinVar SNV benchmark (158,616 variants)	AlphaGenome on the EVEE deconfounded ClinVar SNV benchmark (Pearce et al. 2026)
0.972 unitless	AUROC (Deconfounded ClinVar SNVs, missense only (value plotted in the Overall row))	AlphaMissense hg38, pre-computed (Pearce et al. 2026)	EVEE deconfounded ClinVar SNV benchmark (158,616 variants)	AlphaMissense on the EVEE deconfounded ClinVar SNV benchmark (Pearce et al. 2026)
0.980 unitless	AUROC (ClinVar indels, all strata (Overall row); pre-computed CADD v1.7 InDel scores)	CADD v1.7 PHRED, pre-computed (Pearce et al. 2026)	EVEE ClinVar indel benchmark (73,961 indels)	CADD v1.7 on the EVEE ClinVar indel benchmark (Pearce et al. 2026)
0.963 unitless	AUROC (Deconfounded ClinVar SNVs, all consequence types (Overall row))	CADD v1.7 PHRED, pre-computed (Pearce et al. 2026)	EVEE deconfounded ClinVar SNV benchmark (158,616 variants)	CADD v1.7 on the EVEE deconfounded ClinVar SNV benchmark (Pearce et al. 2026)
∼20,000 hour	Accelerator time (Evo 2 7B embedding extraction for 4,252,870 ClinVar variants)	EVEE covariance probe on Evo 2 7B layer 27 (Pearce et al. 2026)	EVEE full ClinVar variant set (4,252,870 variants)	EVEE Evo 2 7B embedding extraction compute (Pearce et al. 2026)
0.991 unitless	AUROC (ClinVar indels, all strata (Overall row); zero-shot transfer from SNV training)	EVEE covariance probe on Evo 2 7B layer 27 (Pearce et al. 2026)	EVEE ClinVar indel benchmark (73,961 indels)	Evo2 Covariance on the EVEE ClinVar indel benchmark (Pearce et al. 2026)
0.977 unitless	AUROC (Deconfounded ClinVar SNVs, all consequence types (Overall row))	EVEE covariance probe on Evo 2 7B layer 27 (Pearce et al. 2026)	EVEE deconfounded ClinVar SNV benchmark (158,616 variants)	Evo2 Covariance on the EVEE deconfounded ClinVar SNV benchmark (Pearce et al. 2026)
0.929 unitless	AUROC (Deconfounded ClinVar SNVs, all consequence types (Overall row))	Evo 2 7B loss-based zero-shot score (Pearce et al. 2026)	EVEE deconfounded ClinVar SNV benchmark (158,616 variants)	Evo2 Loss on the EVEE deconfounded ClinVar SNV benchmark (Pearce et al. 2026)
0.984 unitless	AUROC (ClinVar indels, all strata (Overall row); supervised on indels)	Evo 2 7B mean-pooled embedding probe (Pearce et al. 2026)	EVEE ClinVar indel benchmark (73,961 indels)	Evo2 Mean on the EVEE ClinVar indel benchmark (Pearce et al. 2026)
0.967 unitless	AUROC (Deconfounded ClinVar SNVs, all consequence types (Overall row))	Evo 2 7B mean-pooled embedding probe (Pearce et al. 2026)	EVEE deconfounded ClinVar SNV benchmark (158,616 variants)	Evo2 Mean on the EVEE deconfounded ClinVar SNV benchmark (Pearce et al. 2026)
0.952 unitless	AUROC (Deconfounded ClinVar SNVs, all consequence types (Overall row))	GPN-MSA hg38 scores, pre-computed (Pearce et al. 2026)	EVEE deconfounded ClinVar SNV benchmark (158,616 variants)	GPN-MSA on the EVEE deconfounded ClinVar SNV benchmark (Pearce et al. 2026)
0.828 unitless	AUROC (ClinVar indels, all strata (Overall row); supervised on indels)	NTv3 650M layer-11 embedding probe (Pearce et al. 2026 Figure 1d)	EVEE ClinVar indel benchmark (73,961 indels)	NTv3 on the EVEE ClinVar indel benchmark (Pearce et al. 2026)
0.710 unitless	AUROC (Deconfounded ClinVar SNVs, all consequence types (Overall row))	NTv3 cosine distance (Pearce et al. 2026 Figure 1b and 1c)	EVEE deconfounded ClinVar SNV benchmark (158,616 variants)	NTv3 on the EVEE deconfounded ClinVar SNV benchmark (Pearce et al. 2026)

Claims

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

Developer overlap: EVEE authors and Evo 2: Developer overlap EVEE is reported by its developers: 11 of its 23 authors, including the first and last authors, are Goodfire employees (Competing interests); 11 are at Mayo Clinic. Three EVEE authors, N. Nguyen, D. Balsam and N. K. Wang, also appear by name and initials in the author list of the Evo 2 preprint (bioRxiv 10.1101/2025.02.18.638918), so the Evo 2 based configurations are treated as author_reported. The match is by name only; affiliations were not compared. (EVEE v4 title page author list and 'Competing interests' (page 15); bioRxiv API details for 10.1101/2025.02.18.638918, field 'authors')
Relevance of EVEE Evo 2 embedding extraction compute (Pearce et al. 2026) to "Select an execution workflow for large-scale diagnostic genomics": Assessed by EVEE Evo 2 embedding extraction compute (Pearce et al. 2026) (gfm-guide-20261014-source-evee-v4: Methods 'Genomic foundation model', paragraphs 'Harvesting' and 'Dataset' (page 8, line 461))

Models, benchmarks and other records

Methods (1)

EVEE

Configurations (9)

Protocols (3)

Datasets (3)

Evaluations (13)

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.

20 evidence rows matching the loaded filters

Claims, original sources and review scope · Release 2026-10-22-1b7cf027e4f6
Property and statement	Original source and location	Review and provenance
attributes.artifact_sha256 1d65b07059c50e8e3f88b8fde950f011630886097bf2784a6653282c6424eef8 Source metadata	Interpretable variant effect prediction from genomic foundation model representations (EVEE), bioRxiv v4 Original source ↗ No field-specific location recorded Version: bioRxiv v4, posted 2026-07-11 (v1 posted 2026-04-11) Retrieved: 2026-10-14T12:02:25Z	catalogued No individual claim review recorded Source has a recorded evidence concern. Consult its source page before using the claim. Audit details Field: `attributes.artifact_sha256` Source artifact SHA-256: `1d65b07059c50e8e3f88b8fde950f011630886097bf2784a6653282c6424eef8` Hash scope: SHA-256 of one retrieval of the bioRxiv PDF (2026-10-14T12:02:25Z). bioRxiv stamps the PDF on each download (XMP ModifyDate, MetadataDate, DocumentID and InstanceID, the Info ModDate and the trailer /ID), so later retrievals differ. The extractor's copy was not retained, so the pinned bytes could not be re-verified. The independent reviewer's retrieval on 2026-10-14 (5,466,701 bytes, 26 pages, ModDate D:20261014063735-07'00') had SHA-256 9e2176440dfae6740b9e799796297cdd70b9bd59edb5c0a1e7fcdb5d6978710a; its Figure 1c and 1d match the repository vector PDFs cell for cell in the Overall rows, and the quoted Methods and Competing interests sentences are present. Inspected artifact
Artifact url https://www.biorxiv.org/content/biorxiv/early/2026/07/11/2026.04.10.717844.full.pdf Source metadata	Interpretable variant effect prediction from genomic foundation model representations (EVEE), bioRxiv v4 Original source ↗ No field-specific location recorded Version: bioRxiv v4, posted 2026-07-11 (v1 posted 2026-04-11) Retrieved: 2026-10-14T12:02:25Z	catalogued No individual claim review recorded Source has a recorded evidence concern. Consult its source page before using the claim. Audit details Field: `attributes.artifact_url` Source artifact SHA-256: `1d65b07059c50e8e3f88b8fde950f011630886097bf2784a6653282c6424eef8` Hash scope: SHA-256 of one retrieval of the bioRxiv PDF (2026-10-14T12:02:25Z). bioRxiv stamps the PDF on each download (XMP ModifyDate, MetadataDate, DocumentID and InstanceID, the Info ModDate and the trailer /ID), so later retrievals differ. The extractor's copy was not retained, so the pinned bytes could not be re-verified. The independent reviewer's retrieval on 2026-10-14 (5,466,701 bytes, 26 pages, ModDate D:20261014063735-07'00') had SHA-256 9e2176440dfae6740b9e799796297cdd70b9bd59edb5c0a1e7fcdb5d6978710a; its Figure 1c and 1d match the repository vector PDFs cell for cell in the Overall rows, and the quoted Methods and Competing interests sentences are present. Inspected artifact
Doi 10.64898/2026.04.10.717844 Source metadata	Interpretable variant effect prediction from genomic foundation model representations (EVEE), bioRxiv v4 Original source ↗ No field-specific location recorded Version: bioRxiv v4, posted 2026-07-11 (v1 posted 2026-04-11) Retrieved: 2026-10-14T12:02:25Z	catalogued No individual claim review recorded Source has a recorded evidence concern. Consult its source page before using the claim. Audit details Field: `attributes.doi` Source artifact SHA-256: `1d65b07059c50e8e3f88b8fde950f011630886097bf2784a6653282c6424eef8` Hash scope: SHA-256 of one retrieval of the bioRxiv PDF (2026-10-14T12:02:25Z). bioRxiv stamps the PDF on each download (XMP ModifyDate, MetadataDate, DocumentID and InstanceID, the Info ModDate and the trailer /ID), so later retrievals differ. The extractor's copy was not retained, so the pinned bytes could not be re-verified. The independent reviewer's retrieval on 2026-10-14 (5,466,701 bytes, 26 pages, ModDate D:20261014063735-07'00') had SHA-256 9e2176440dfae6740b9e799796297cdd70b9bd59edb5c0a1e7fcdb5d6978710a; its Figure 1c and 1d match the repository vector PDFs cell for cell in the Overall rows, and the quoted Methods and Competing interests sentences are present. Inspected artifact
Evidence concerns 2 values source id gfm-guide-20261014-source-evee-v4 message The column labelled 'NTv3' in Figure 1b and 1c is not the method the Methods describe. The manuscript repository's plotting code (scripts/figure1/_heatmap_common.py at e5e4a43) renames 'NTv3 cosine dist' to 'NTv3' and drops 'NTv3 subref probe', whereas Methods 'Baseline methods' describe NTv3 650M layer-11 embeddings with L2-regularised logistic regression under 5-fold gene-holdout cross-validation. Figure 1d's NTv3 column is the supervised subref probe. The SNV and deconfounded NTv3 values are recorded as a zero-shot embedding distance. source locator Figure 1b and 1c, column 'NTv3' (page 3); Methods 'Baseline methods', NTv3 (page 8); repository scripts/figure1/_heatmap_common.py RENAME_MAP and DROP_METHODS artifact sha256 1d65b07059c50e8e3f88b8fde950f011630886097bf2784a6653282c6424eef8 reviewed at 2026-10-14T12:58:25Z review method automated-source-review source id gfm-guide-20261014-source-evee-v4 message AlphaMissense's 'Overall' cell in Figure 1b and 1c is its missense AUROC. The plotting code blanks AlphaMissense outside missense and copies the Missense value into the Overall row (scripts/figure1/_heatmap_common.py, am_ov_mask), so it is computed on a different, missense-only population from the other methods' Overall cells. source locator Figure 1b and 1c, row 'Overall', column 'AlphaMissense' (page 3); repository scripts/figure1/_heatmap_common.py prepare() artifact sha256 1d65b07059c50e8e3f88b8fde950f011630886097bf2784a6653282c6424eef8 reviewed at 2026-10-14T12:58:25Z review method automated-source-review Source metadata	Interpretable variant effect prediction from genomic foundation model representations (EVEE), bioRxiv v4 Original source ↗ No field-specific location recorded Version: bioRxiv v4, posted 2026-07-11 (v1 posted 2026-04-11) Retrieved: 2026-10-14T12:02:25Z	catalogued No individual claim review recorded Source has a recorded evidence concern. Consult its source page before using the claim. Audit details Field: `attributes.evidence_concerns` Source artifact SHA-256: `1d65b07059c50e8e3f88b8fde950f011630886097bf2784a6653282c6424eef8` Hash scope: SHA-256 of one retrieval of the bioRxiv PDF (2026-10-14T12:02:25Z). bioRxiv stamps the PDF on each download (XMP ModifyDate, MetadataDate, DocumentID and InstanceID, the Info ModDate and the trailer /ID), so later retrievals differ. The extractor's copy was not retained, so the pinned bytes could not be re-verified. The independent reviewer's retrieval on 2026-10-14 (5,466,701 bytes, 26 pages, ModDate D:20261014063735-07'00') had SHA-256 9e2176440dfae6740b9e799796297cdd70b9bd59edb5c0a1e7fcdb5d6978710a; its Figure 1c and 1d match the repository vector PDFs cell for cell in the Overall rows, and the quoted Methods and Competing interests sentences are present. Inspected artifact
Hash scope SHA-256 of one retrieval of the bioRxiv PDF (2026-10-14T12:02:25Z). bioRxiv stamps the PDF on each download (XMP ModifyDate, MetadataDate, DocumentID and InstanceID, the Info ModDate and the trailer /ID), so later retrievals differ. The extractor's copy was not retained, so the pinned bytes could not be re-verified. The independent reviewer's retrieval on 2026-10-14 (5,466,701 bytes, 26 pages, ModDate D:20261014063735-07'00') had SHA-256 9e2176440dfae6740b9e799796297cdd70b9bd59edb5c0a1e7fcdb5d6978710a; its Figure 1c and 1d match the repository vector PDFs cell for cell in the Overall rows, and the quoted Methods and Competing interests sentences are present. Source metadata	Interpretable variant effect prediction from genomic foundation model representations (EVEE), bioRxiv v4 Original source ↗ No field-specific location recorded Version: bioRxiv v4, posted 2026-07-11 (v1 posted 2026-04-11) Retrieved: 2026-10-14T12:02:25Z	catalogued No individual claim review recorded Source has a recorded evidence concern. Consult its source page before using the claim. Audit details Field: `attributes.hash_scope` Source artifact SHA-256: `1d65b07059c50e8e3f88b8fde950f011630886097bf2784a6653282c6424eef8` Hash scope: SHA-256 of one retrieval of the bioRxiv PDF (2026-10-14T12:02:25Z). bioRxiv stamps the PDF on each download (XMP ModifyDate, MetadataDate, DocumentID and InstanceID, the Info ModDate and the trailer /ID), so later retrievals differ. The extractor's copy was not retained, so the pinned bytes could not be re-verified. The independent reviewer's retrieval on 2026-10-14 (5,466,701 bytes, 26 pages, ModDate D:20261014063735-07'00') had SHA-256 9e2176440dfae6740b9e799796297cdd70b9bd59edb5c0a1e7fcdb5d6978710a; its Figure 1c and 1d match the repository vector PDFs cell for cell in the Overall rows, and the quoted Methods and Competing interests sentences are present. Inspected artifact
Licence CC-BY-4.0 Source metadata	Interpretable variant effect prediction from genomic foundation model representations (EVEE), bioRxiv v4 Original source ↗ No field-specific location recorded Version: bioRxiv v4, posted 2026-07-11 (v1 posted 2026-04-11) Retrieved: 2026-10-14T12:02:25Z	catalogued No individual claim review recorded Source has a recorded evidence concern. Consult its source page before using the claim. Audit details Field: `attributes.licence` Source artifact SHA-256: `1d65b07059c50e8e3f88b8fde950f011630886097bf2784a6653282c6424eef8` Hash scope: SHA-256 of one retrieval of the bioRxiv PDF (2026-10-14T12:02:25Z). bioRxiv stamps the PDF on each download (XMP ModifyDate, MetadataDate, DocumentID and InstanceID, the Info ModDate and the trailer /ID), so later retrievals differ. The extractor's copy was not retained, so the pinned bytes could not be re-verified. The independent reviewer's retrieval on 2026-10-14 (5,466,701 bytes, 26 pages, ModDate D:20261014063735-07'00') had SHA-256 9e2176440dfae6740b9e799796297cdd70b9bd59edb5c0a1e7fcdb5d6978710a; its Figure 1c and 1d match the repository vector PDFs cell for cell in the Overall rows, and the quoted Methods and Competing interests sentences are present. Inspected artifact
Media type application/pdf Source metadata	Interpretable variant effect prediction from genomic foundation model representations (EVEE), bioRxiv v4 Original source ↗ No field-specific location recorded Version: bioRxiv v4, posted 2026-07-11 (v1 posted 2026-04-11) Retrieved: 2026-10-14T12:02:25Z	catalogued No individual claim review recorded Source has a recorded evidence concern. Consult its source page before using the claim. Audit details Field: `attributes.media_type` Source artifact SHA-256: `1d65b07059c50e8e3f88b8fde950f011630886097bf2784a6653282c6424eef8` Hash scope: SHA-256 of one retrieval of the bioRxiv PDF (2026-10-14T12:02:25Z). bioRxiv stamps the PDF on each download (XMP ModifyDate, MetadataDate, DocumentID and InstanceID, the Info ModDate and the trailer /ID), so later retrievals differ. The extractor's copy was not retained, so the pinned bytes could not be re-verified. The independent reviewer's retrieval on 2026-10-14 (5,466,701 bytes, 26 pages, ModDate D:20261014063735-07'00') had SHA-256 9e2176440dfae6740b9e799796297cdd70b9bd59edb5c0a1e7fcdb5d6978710a; its Figure 1c and 1d match the repository vector PDFs cell for cell in the Overall rows, and the quoted Methods and Competing interests sentences are present. Inspected artifact
Publication status preprint Source metadata	Interpretable variant effect prediction from genomic foundation model representations (EVEE), bioRxiv v4 Original source ↗ No field-specific location recorded Version: bioRxiv v4, posted 2026-07-11 (v1 posted 2026-04-11) Retrieved: 2026-10-14T12:02:25Z	catalogued No individual claim review recorded Source has a recorded evidence concern. Consult its source page before using the claim. Audit details Field: `attributes.publication_status` Source artifact SHA-256: `1d65b07059c50e8e3f88b8fde950f011630886097bf2784a6653282c6424eef8` Hash scope: SHA-256 of one retrieval of the bioRxiv PDF (2026-10-14T12:02:25Z). bioRxiv stamps the PDF on each download (XMP ModifyDate, MetadataDate, DocumentID and InstanceID, the Info ModDate and the trailer /ID), so later retrievals differ. The extractor's copy was not retained, so the pinned bytes could not be re-verified. The independent reviewer's retrieval on 2026-10-14 (5,466,701 bytes, 26 pages, ModDate D:20261014063735-07'00') had SHA-256 9e2176440dfae6740b9e799796297cdd70b9bd59edb5c0a1e7fcdb5d6978710a; its Figure 1c and 1d match the repository vector PDFs cell for cell in the Overall rows, and the quoted Methods and Competing interests sentences are present. Inspected artifact
Retrieval note Figure 1 heatmaps are raster images in the PDF; their printed cell values were read from a 300 dpi render of page 3 and matched against the manuscript repository's vector figure PDFs. Source metadata	Interpretable variant effect prediction from genomic foundation model representations (EVEE), bioRxiv v4 Original source ↗ No field-specific location recorded Version: bioRxiv v4, posted 2026-07-11 (v1 posted 2026-04-11) Retrieved: 2026-10-14T12:02:25Z	catalogued No individual claim review recorded Source has a recorded evidence concern. Consult its source page before using the claim. Audit details Field: `attributes.retrieval_note` Source artifact SHA-256: `1d65b07059c50e8e3f88b8fde950f011630886097bf2784a6653282c6424eef8` Hash scope: SHA-256 of one retrieval of the bioRxiv PDF (2026-10-14T12:02:25Z). bioRxiv stamps the PDF on each download (XMP ModifyDate, MetadataDate, DocumentID and InstanceID, the Info ModDate and the trailer /ID), so later retrievals differ. The extractor's copy was not retained, so the pinned bytes could not be re-verified. The independent reviewer's retrieval on 2026-10-14 (5,466,701 bytes, 26 pages, ModDate D:20261014063735-07'00') had SHA-256 9e2176440dfae6740b9e799796297cdd70b9bd59edb5c0a1e7fcdb5d6978710a; its Figure 1c and 1d match the repository vector PDFs cell for cell in the Overall rows, and the quoted Methods and Competing interests sentences are present. Inspected artifact
Retrieved at 2026-10-14T12:02:25Z Source metadata	Interpretable variant effect prediction from genomic foundation model representations (EVEE), bioRxiv v4 Original source ↗ No field-specific location recorded Version: bioRxiv v4, posted 2026-07-11 (v1 posted 2026-04-11) Retrieved: 2026-10-14T12:02:25Z	catalogued No individual claim review recorded Source has a recorded evidence concern. Consult its source page before using the claim. Audit details Field: `attributes.retrieved_at` Source artifact SHA-256: `1d65b07059c50e8e3f88b8fde950f011630886097bf2784a6653282c6424eef8` Hash scope: SHA-256 of one retrieval of the bioRxiv PDF (2026-10-14T12:02:25Z). bioRxiv stamps the PDF on each download (XMP ModifyDate, MetadataDate, DocumentID and InstanceID, the Info ModDate and the trailer /ID), so later retrievals differ. The extractor's copy was not retained, so the pinned bytes could not be re-verified. The independent reviewer's retrieval on 2026-10-14 (5,466,701 bytes, 26 pages, ModDate D:20261014063735-07'00') had SHA-256 9e2176440dfae6740b9e799796297cdd70b9bd59edb5c0a1e7fcdb5d6978710a; its Figure 1c and 1d match the repository vector PDFs cell for cell in the Overall rows, and the quoted Methods and Competing interests sentences are present. Inspected artifact

How to read the evidence

Sources and history

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

Preprint PDF retrieved and hashed for the genomic foundation model guide sources pass (issue #138).

Release history

No supporting source is linked yet.

Read original source

Download this release (gzip)

Technical metadata and extraction receipts

Stable ID: gfm-guide-20261014-source-evee-v4

areas: dna-genomes
contexts: research
artifact url: https://www.biorxiv.org/content/biorxiv/early/2026/07/11/2026.04.10.717844.full.pdf
retrieved at: 2026-10-14T12:02:25Z
artifact sha256: 1d65b07059c50e8e3f88b8fde950f011630886097bf2784a6653282c6424eef8
doi: 10.64898/2026.04.10.717844
publication status: preprint
licence: CC-BY-4.0
url: https://doi.org/10.64898/2026.04.10.717844
version: bioRxiv v4, posted 2026-07-11 (v1 posted 2026-04-11)
media type: application/pdf
evidence concerns: source id: gfm-guide-20261014-source-evee-v4; message: The column labelled 'NTv3' in Figure 1b and 1c is not the method the Methods describe. The manuscript repository's plotting code (scripts/figure1/_heatmap_common.py at e5e4a43) renames 'NTv3 cosine dist' to 'NTv3' and drops 'NTv3 subref probe', whereas Methods 'Baseline methods' describe NTv3 650M layer-11 embeddings with L2-regularised logistic regression under 5-fold gene-holdout cross-validation. Figure 1d's NTv3 column is the supervised subref probe. The SNV and deconfounded NTv3 values are recorded as a zero-shot embedding distance.; source locator: Figure 1b and 1c, column 'NTv3' (page 3); Methods 'Baseline methods', NTv3 (page 8); repository scripts/figure1/_heatmap_common.py RENAME_MAP and DROP_METHODS; artifact sha256: 1d65b07059c50e8e3f88b8fde950f011630886097bf2784a6653282c6424eef8; reviewed at: 2026-10-14T12:58:25Z; review method: automated-source-review; source id: gfm-guide-20261014-source-evee-v4; message: AlphaMissense's 'Overall' cell in Figure 1b and 1c is its missense AUROC. The plotting code blanks AlphaMissense outside missense and copies the Missense value into the Overall row (scripts/figure1/_heatmap_common.py, am_ov_mask), so it is computed on a different, missense-only population from the other methods' Overall cells.; source locator: Figure 1b and 1c, row 'Overall', column 'AlphaMissense' (page 3); repository scripts/figure1/_heatmap_common.py prepare(); artifact sha256: 1d65b07059c50e8e3f88b8fde950f011630886097bf2784a6653282c6424eef8; reviewed at: 2026-10-14T12:58:25Z; review method: automated-source-review
retrieval note: Figure 1 heatmaps are raster images in the PDF; their printed cell values were read from a 300 dpi render of page 3 and matched against the manuscript repository's vector figure PDFs.
hash scope: SHA-256 of one retrieval of the bioRxiv PDF (2026-10-14T12:02:25Z). bioRxiv stamps the PDF on each download (XMP ModifyDate, MetadataDate, DocumentID and InstanceID, the Info ModDate and the trailer /ID), so later retrievals differ. The extractor's copy was not retained, so the pinned bytes could not be re-verified. The independent reviewer's retrieval on 2026-10-14 (5,466,701 bytes, 26 pages, ModDate D:20261014063735-07'00') had SHA-256 9e2176440dfae6740b9e799796297cdd70b9bd59edb5c0a1e7fcdb5d6978710a; its Figure 1c and 1d match the repository vector PDFs cell for cell in the Overall rows, and the quoted Methods and Competing interests sentences are present.

Related records

Suggest a correction