rewirebio.iobenchmarks
Result

AUPRC (best of zero-shot and linear probing; single trait; chromosome aggregation unreported) of Borzoi on TraitGym Complex traits: Platelet count

0.426 AUPRC (best of zero-shot and linear probing; single trait; chromosome aggregation unreported)

Methods

Tested model
Borzoi
Protocol
TraitGym Complex traits per trait, AUPRC, best of zero-shot and linear probing
Dataset subset
TraitGym Complex traits: Platelet count
Procedure
TraitGym Complex traits per trait, AUPRC, best of zero-shot and linear probing
Evaluation
Borzoi on TraitGym Complex traits, trait Platelet count
Coverage
Not reported scored / Not reported eligible
Uncertainty
Not reported by the source: The table prints no standard error; the bootstrap errors of Appendix B.4 are reported only in the figures.
Evidence
Author-reported evaluation · source checkedBenegas, Eraslan and Song 2025, TraitGym benchmark preprint (bioRxiv v2) · Table S11, row "Platelet count", column "Borzoi" (page 28)

A source-checked result verifies the numerical transcription, not every model or protocol detail. Evaluation metadata: source checked. Source checked does not mean independently reproduced.

Reproduction

Split
Leave-one-chromosome-out for linear probing; all variants scored for zero-shot
Adaptation
Best of zero-shot and linear probing; which one is not stated per cell
Scoring implementation
Not reported

No execution recipe has been verified for this exact configuration and evaluation. A benchmark's general instructions may use different inputs, splits or model settings.

Reproducing this published result requires matching its model configuration, data, split and scorer. Source checking or a successful smoke test does not establish score reproduction.

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.

1 evidence row matching the loaded filters

Claims, original sources and review scope · Release 2026-11-21-c8bea25e5abb
Property and statementOriginal source and locationReview and provenance
Reported result
0.426
Individual claims
Benegas, Eraslan and Song 2025, TraitGym benchmark preprint (bioRxiv v2)

Original source ↗

Table S11, row "Platelet count", column "Borzoi" (page 28)

Version: bioRxiv version 2, posted 2025-03-04 (37 pages, main text and Supplementary Material)
Retrieved: 2026-11-16T01:09:36Z

source checked

["source-hash-verification","deterministic-table-parse","pdf-text-parse","independent-cell-check"] · 2026-11-16T02:10:00Z

author reported

Audit details

npm run batch -- check-values reproduced the extractor's report byte for byte: 96 in_cell, 36 in_row, no not_in_row, no wrong_column, no overclaimed_in_row. All 132 results were then re-read by data/omics/traitgym-20261116/review/recheck_values.py, written for this pass and importing nothing from extract/. It fixes a character-column band per table column from the start offset of each column header token (Table S8 at offsets 59, 71, 88, 102 and 122; Table S11 at 79, 92, 111 and 124; Table 4 at 85; Table 5 at 113; Table S4 at 98) and accepts a numeric token only when its own start offset falls inside exactly one band, so column assignment is checked directly for all 132 results, including the 36 that check-values could place only in a row. Result: 132 of 132 agree on printed_value and numeric_value, 0 mismatches (review/recheck-values.tsv). Three negative controls confirm it is a cell check and not a row check: swapping two values inside one Table S11 row rejects exactly those two results, copying one value into a neighbouring column band rejects exactly that one, and shifting a row's values one column left is rejected at parse with the offending offset named. Independently of both scripts, all 100 cells of Table S8 were transcribed by hand into review/s8-reviewer-reading.tsv before extract/comparison.tsv was opened, and all 36 in_row cells of Tables 4, 5, S4 and S11 were read by eye from the text layer; the hand reading of Table S8 agrees with the band parse in all 100 cells.

Field: attributes.printed_value

Source artifact SHA-256: 80426c499e0b498e07c07ea598a2f68049adea7569b30fda58cdc0ebafa6c7aa

Hash scope: SHA-256 of the PDF bytes of one download. bioRxiv stamps each download, so this hash pins this retrieval and not the document: a second download in the same session gave 5f424797ed5c92a3852a52e8ecf29f0c3a3416740a47fe9f9fefe9fd9fb0c876. The stable check is the pdftotext -layout text layer, whose SHA-256 is 23ae4d67751f1092e5cc2da66b52091fa9bbc3ec59310f06bc65be28f93c5097 and was byte-identical across both downloads; every value in this batch was read from that text layer.

Inspected artifact

Extraction artifact SHA-256: 23ae4d67751f1092e5cc2da66b52091fa9bbc3ec59310f06bc65be28f93c5097

Extraction artifact

Sources and history

Release 2026-11-21-c8bea25e5abb · Record review: source checked

1 source record and release historyDownload this release (gzip)
Technical metadata and extraction receipts

Stable ID: traitgym-20261116-result-s11-complex-platelet-count-borzoi

limitations
Whether this cell is the zero-shot or the linear probing score is not stated.
metric
auprc
metric direction
higher
metric qualifier
best of zero-shot and linear probing; single trait; chromosome aggregation unreported
missing metadata
uncertainty: note: The table prints no standard error; the bootstrap errors of Appendix B.4 are reported only in the figures.; reason: unreported
numeric value
0.426
printed value
0.426
printed value basis
Value as printed in the table, at the paper's rounding
review
artifact sha256: 23ae4d67751f1092e5cc2da66b52091fa9bbc3ec59310f06bc65be28f93c5097; method: source-hash-verification; deterministic-table-parse; pdf-text-parse; independent-cell-check; method note: The pinned artifact was re-downloaded from the v2 full-PDF URL for this pass. Its PDF bytes hash to 961933ed2396ecb0a75737cc27f0655db07ca62be49654f9139ae95815d050f9, not to the 80426c499e0b498e07c07ea598a2f68049adea7569b30fda58cdc0ebafa6c7aa on the source record, and cannot: bioRxiv restamps the container on every request. The reviewer's download differs from the extractor's two in 41 and 44 bytes, every one of them a download-time stamp (the XMP ModifyDate and MetadataDate, the document ModDate and the trailer ID), with the same 37 pages, the same CreationDate of 2025-03-04 and the same pdftotext -layout text layer. That text layer hashes to 23ae4d67751f1092e5cc2da66b52091fa9bbc3ec59310f06bc65be28f93c5097, which is the hash the source record's hash_scope names as the stable check and the hash recorded here; every value in this batch was read from it.; note: npm run batch -- check-values reproduced the extractor's report byte for byte: 96 in_cell, 36 in_row, no not_in_row, no wrong_column, no overclaimed_in_row. All 132 results were then re-read by data/omics/traitgym-20261116/review/recheck_values.py, written for this pass and importing nothing from extract/. It fixes a character-column band per table column from the start offset of each column header token (Table S8 at offsets 59, 71, 88, 102 and 122; Table S11 at 79, 92, 111 and 124; Table 4 at 85; Table 5 at 113; Table S4 at 98) and accepts a numeric token only when its own start offset falls inside exactly one band, so column assignment is checked directly for all 132 results, including the 36 that check-values could place only in a row. Result: 132 of 132 agree on printed_value and numeric_value, 0 mismatches (review/recheck-values.tsv). Three negative controls confirm it is a cell check and not a row check: swapping two values inside one Table S11 row rejects exactly those two results, copying one value into a neighbouring column band rejects exactly that one, and shifting a row's values one column left is rejected at parse with the offending offset named. Independently of both scripts, all 100 cells of Table S8 were transcribed by hand into review/s8-reviewer-reading.tsv before extract/comparison.tsv was opened, and all 36 in_row cells of Tables 4, 5, S4 and S11 were read by eye from the text layer; the hand reading of Table S8 agrees with the band parse in all 100 cells.; retrieval url: https://www.biorxiv.org/content/10.1101/2025.02.11.637758v2.full.pdf; reviewed at: 2026-11-16T02:10:00Z; reviewer: claude; reviewer note: Claude review agent (independent of the extractor); no human review claimed. Every value was read against the pinned artifact; nothing was reproduced and no model was run.
source locator
Table S11, row "Platelet count", column "Borzoi" (page 28)
unit
unitless
Related records

Suggest a correction