rewire.it
Dataset subset

ProteinGym substitution DMS assays (ProteinGym split)

The split of ProteinGym substitution DMS assays that ProteinGym evaluated on. The upstream dataset release is not catalogued here, so no claim is made that this matches its original splits.

Subset and evaluation context

This record describes a particular subset or cohort used in an evaluation. Its results do not describe the full dataset.

Evaluation results

Release 2026-09-17-134cd1815de8 · 185 evaluations · 185 metric rows. Different protocols are not a single leaderboard. Where several source tables report the same metric, the published comparisons above offer a pooled view that names what it does not hold constant.

Results grouped by the exact reported evaluation
Metric and findingCoverage and uncertaintyEvidence
CARP (640M) on ProteinGym ZS-SUB-AUC: Zero-shot substitutions, AUC

Zero-shot scoring of substitution assays, averaged over assays with the correction the paper describes.

Author-reported evaluation · Evaluation metadata: source checked

0.701 auc

Unit: fraction · Direction: higher

Uncertainty: Not reported

Scored: Not reported · Eligible: Not reported

source checkedProteinGym: Large-Scale Benchmarks for Protein Fitness Prediction and Design · Table 2, row(language CARP (640M)), column(Zero-shot substitutions, AUC)

Source checking is not independent reproduction.

CARP (640M) on ProteinGym ZS-SUB-MCC: Zero-shot substitutions, MCC

Zero-shot scoring of substitution assays, averaged over assays with the correction the paper describes.

Author-reported evaluation · Evaluation metadata: source checked

0.285 mcc

Unit: correlation · Direction: higher

Uncertainty: Not reported

Scored: Not reported · Eligible: Not reported

source checkedProteinGym: Large-Scale Benchmarks for Protein Fitness Prediction and Design · Table 2, row(language CARP (640M)), column(Zero-shot substitutions, MCC)

Source checking is not independent reproduction.

CARP (640M) on ProteinGym ZS-SUB-NDCG: Zero-shot substitutions, NDCG@10%

Zero-shot scoring of substitution assays, averaged over assays with the correction the paper describes.

Author-reported evaluation · Evaluation metadata: source checked

0.748 ndcg

Unit: fraction · Direction: higher

Uncertainty: Not reported

Scored: Not reported · Eligible: Not reported

source checkedProteinGym: Large-Scale Benchmarks for Protein Fitness Prediction and Design · Table 2, row(language CARP (640M)), column(Zero-shot substitutions, NDCG@10%)

Source checking is not independent reproduction.

CARP (640M) on ProteinGym ZS-SUB-RECALL: Zero-shot substitutions, top 10% recall

Zero-shot scoring of substitution assays, averaged over assays with the correction the paper describes.

Author-reported evaluation · Evaluation metadata: source checked

0.208 recall

Unit: fraction · Direction: higher

Uncertainty: Not reported

Scored: Not reported · Eligible: Not reported

source checkedProteinGym: Large-Scale Benchmarks for Protein Fitness Prediction and Design · Table 2, row(language CARP (640M)), column(Zero-shot substitutions, top 10% recall)

Source checking is not independent reproduction.

CARP (640M) on ProteinGym ZS-SUB-SPEARMAN: Zero-shot substitutions, Spearman

Zero-shot scoring of substitution assays, averaged over assays with the correction the paper describes.

Author-reported evaluation · Evaluation metadata: source checked

0.368 spearman

Unit: correlation · Direction: higher

Uncertainty: Not reported

Scored: Not reported · Eligible: Not reported

source checkedProteinGym: Large-Scale Benchmarks for Protein Fitness Prediction and Design · Table 2, row(language CARP (640M)), column(Zero-shot substitutions, Spearman)

Source checking is not independent reproduction.

DeepSequence (ensemble) on ProteinGym ZS-SUB-AUC: Zero-shot substitutions, AUC

Zero-shot scoring of substitution assays, averaged over assays with the correction the paper describes.

Author-reported evaluation · Evaluation metadata: source checked

0.729 auc

Unit: fraction · Direction: higher

Uncertainty: Not reported

Scored: Not reported · Eligible: Not reported

source checkedProteinGym: Large-Scale Benchmarks for Protein Fitness Prediction and Design · Table 2, row(DeepSequence (ensemble)), column(Zero-shot substitutions, AUC)

Source checking is not independent reproduction.

DeepSequence (ensemble) on ProteinGym ZS-SUB-MCC: Zero-shot substitutions, MCC

Zero-shot scoring of substitution assays, averaged over assays with the correction the paper describes.

Author-reported evaluation · Evaluation metadata: source checked

0.328 mcc

Unit: correlation · Direction: higher

Uncertainty: Not reported

Scored: Not reported · Eligible: Not reported

source checkedProteinGym: Large-Scale Benchmarks for Protein Fitness Prediction and Design · Table 2, row(DeepSequence (ensemble)), column(Zero-shot substitutions, MCC)

Source checking is not independent reproduction.

DeepSequence (ensemble) on ProteinGym ZS-SUB-NDCG: Zero-shot substitutions, NDCG@10%

Zero-shot scoring of substitution assays, averaged over assays with the correction the paper describes.

Author-reported evaluation · Evaluation metadata: source checked

0.776 ndcg

Unit: fraction · Direction: higher

Uncertainty: Not reported

Scored: Not reported · Eligible: Not reported

source checkedProteinGym: Large-Scale Benchmarks for Protein Fitness Prediction and Design · Table 2, row(DeepSequence (ensemble)), column(Zero-shot substitutions, NDCG@10%)

Source checking is not independent reproduction.

DeepSequence (ensemble) on ProteinGym ZS-SUB-RECALL: Zero-shot substitutions, top 10% recall

Zero-shot scoring of substitution assays, averaged over assays with the correction the paper describes.

Author-reported evaluation · Evaluation metadata: source checked

0.226 recall

Unit: fraction · Direction: higher

Uncertainty: Not reported

Scored: Not reported · Eligible: Not reported

source checkedProteinGym: Large-Scale Benchmarks for Protein Fitness Prediction and Design · Table 2, row(DeepSequence (ensemble)), column(Zero-shot substitutions, top 10% recall)

Source checking is not independent reproduction.

DeepSequence (ensemble) on ProteinGym ZS-SUB-SPEARMAN: Zero-shot substitutions, Spearman

Zero-shot scoring of substitution assays, averaged over assays with the correction the paper describes.

Author-reported evaluation · Evaluation metadata: source checked

0.419 spearman

Unit: correlation · Direction: higher

Uncertainty: Not reported

Scored: Not reported · Eligible: Not reported

source checkedProteinGym: Large-Scale Benchmarks for Protein Fitness Prediction and Design · Table 2, row(DeepSequence (ensemble)), column(Zero-shot substitutions, Spearman)

Source checking is not independent reproduction.

Embed. ESM-1v on ProteinGym SUP-SUB-MSE-AVG: Supervised substitutions, average over splits, MSE

Supervised fitness prediction under three cross-validation schemes: contiguous, modulo and random splits.

Author-reported evaluation · Evaluation metadata: source checked

0.787 mse

Unit: error · Direction: lower

Uncertainty: Not reported

Scored: Not reported · Eligible: Not reported

source checkedProteinGym: Large-Scale Benchmarks for Protein Fitness Prediction and Design · Table 3, row(Embed. ESM-1v), column(Supervised substitutions, average over splits, MSE)

Source checking is not independent reproduction.

Embed. ESM-1v on ProteinGym SUP-SUB-MSE-CONTIG: Supervised substitutions, contiguous split, MSE

Supervised fitness prediction under three cross-validation schemes: contiguous, modulo and random splits.

Author-reported evaluation · Evaluation metadata: source checked

0.937 mse

Unit: error · Direction: lower

Uncertainty: Not reported

Scored: Not reported · Eligible: Not reported

source checkedProteinGym: Large-Scale Benchmarks for Protein Fitness Prediction and Design · Table 3, row(Embed. ESM-1v), column(Supervised substitutions, contiguous split, MSE)

Source checking is not independent reproduction.

Embed. ESM-1v on ProteinGym SUP-SUB-MSE-MOD: Supervised substitutions, modulo split, MSE

Supervised fitness prediction under three cross-validation schemes: contiguous, modulo and random splits.

Author-reported evaluation · Evaluation metadata: source checked

0.861 mse

Unit: error · Direction: lower

Uncertainty: Not reported

Scored: Not reported · Eligible: Not reported

source checkedProteinGym: Large-Scale Benchmarks for Protein Fitness Prediction and Design · Table 3, row(Embed. ESM-1v), column(Supervised substitutions, modulo split, MSE)

Source checking is not independent reproduction.

Embed. ESM-1v on ProteinGym SUP-SUB-MSE-RAND: Supervised substitutions, random split, MSE

Supervised fitness prediction under three cross-validation schemes: contiguous, modulo and random splits.

Author-reported evaluation · Evaluation metadata: source checked

0.563 mse

Unit: error · Direction: lower

Uncertainty: Not reported

Scored: Not reported · Eligible: Not reported

source checkedProteinGym: Large-Scale Benchmarks for Protein Fitness Prediction and Design · Table 3, row(Embed. ESM-1v), column(Supervised substitutions, random split, MSE)

Source checking is not independent reproduction.

Embed. ESM-1v on ProteinGym SUP-SUB-SPEARMAN-AVG: Supervised substitutions, average over splits, Spearman

Supervised fitness prediction under three cross-validation schemes: contiguous, modulo and random splits.

Author-reported evaluation · Evaluation metadata: source checked

0.542 spearman

Unit: correlation · Direction: higher

Uncertainty: Not reported

Scored: Not reported · Eligible: Not reported

source checkedProteinGym: Large-Scale Benchmarks for Protein Fitness Prediction and Design · Table 3, row(Embed. ESM-1v), column(Supervised substitutions, average over splits, Spearman)

Source checking is not independent reproduction.

Embed. ESM-1v on ProteinGym SUP-SUB-SPEARMAN-CONTIG: Supervised substitutions, contiguous split, Spearman

Supervised fitness prediction under three cross-validation schemes: contiguous, modulo and random splits.

Author-reported evaluation · Evaluation metadata: source checked

0.481 spearman

Unit: correlation · Direction: higher

Uncertainty: Not reported

Scored: Not reported · Eligible: Not reported

source checkedProteinGym: Large-Scale Benchmarks for Protein Fitness Prediction and Design · Table 3, row(Embed. ESM-1v), column(Supervised substitutions, contiguous split, Spearman)

Source checking is not independent reproduction.

Embed. ESM-1v on ProteinGym SUP-SUB-SPEARMAN-MOD: Supervised substitutions, modulo split, Spearman

Supervised fitness prediction under three cross-validation schemes: contiguous, modulo and random splits.

Author-reported evaluation · Evaluation metadata: source checked

0.506 spearman

Unit: correlation · Direction: higher

Uncertainty: Not reported

Scored: Not reported · Eligible: Not reported

source checkedProteinGym: Large-Scale Benchmarks for Protein Fitness Prediction and Design · Table 3, row(Embed. ESM-1v), column(Supervised substitutions, modulo split, Spearman)

Source checking is not independent reproduction.

Embed. ESM-1v on ProteinGym SUP-SUB-SPEARMAN-RAND: Supervised substitutions, random split, Spearman

Supervised fitness prediction under three cross-validation schemes: contiguous, modulo and random splits.

Author-reported evaluation · Evaluation metadata: source checked

0.639 spearman

Unit: correlation · Direction: higher

Uncertainty: Not reported

Scored: Not reported · Eligible: Not reported

source checkedProteinGym: Large-Scale Benchmarks for Protein Fitness Prediction and Design · Table 3, row(Embed. ESM-1v), column(Supervised substitutions, random split, Spearman)

Source checking is not independent reproduction.

Embed. MSAT on ProteinGym SUP-SUB-MSE-AVG: Supervised substitutions, average over splits, MSE

Supervised fitness prediction under three cross-validation schemes: contiguous, modulo and random splits.

Author-reported evaluation · Evaluation metadata: source checked

0.735 mse

Unit: error · Direction: lower

Uncertainty: Not reported

Scored: Not reported · Eligible: Not reported

source checkedProteinGym: Large-Scale Benchmarks for Protein Fitness Prediction and Design · Table 3, row(MSAT), column(Supervised substitutions, average over splits, MSE)

Source checking is not independent reproduction.

Embed. MSAT on ProteinGym SUP-SUB-MSE-CONTIG: Supervised substitutions, contiguous split, MSE

Supervised fitness prediction under three cross-validation schemes: contiguous, modulo and random splits.

Author-reported evaluation · Evaluation metadata: source checked

0.836 mse

Unit: error · Direction: lower

Uncertainty: Not reported

Scored: Not reported · Eligible: Not reported

source checkedProteinGym: Large-Scale Benchmarks for Protein Fitness Prediction and Design · Table 3, row(MSAT), column(Supervised substitutions, contiguous split, MSE)

Source checking is not independent reproduction.

Embed. MSAT on ProteinGym SUP-SUB-MSE-MOD: Supervised substitutions, modulo split, MSE

Supervised fitness prediction under three cross-validation schemes: contiguous, modulo and random splits.

Author-reported evaluation · Evaluation metadata: source checked

0.795 mse

Unit: error · Direction: lower

Uncertainty: Not reported

Scored: Not reported · Eligible: Not reported

source checkedProteinGym: Large-Scale Benchmarks for Protein Fitness Prediction and Design · Table 3, row(MSAT), column(Supervised substitutions, modulo split, MSE)

Source checking is not independent reproduction.

Embed. MSAT on ProteinGym SUP-SUB-MSE-RAND: Supervised substitutions, random split, MSE

Supervised fitness prediction under three cross-validation schemes: contiguous, modulo and random splits.

Author-reported evaluation · Evaluation metadata: source checked

0.573 mse

Unit: error · Direction: lower

Uncertainty: Not reported

Scored: Not reported · Eligible: Not reported

source checkedProteinGym: Large-Scale Benchmarks for Protein Fitness Prediction and Design · Table 3, row(MSAT), column(Supervised substitutions, random split, MSE)

Source checking is not independent reproduction.

Embed. MSAT on ProteinGym SUP-SUB-SPEARMAN-AVG: Supervised substitutions, average over splits, Spearman

Supervised fitness prediction under three cross-validation schemes: contiguous, modulo and random splits.

Author-reported evaluation · Evaluation metadata: source checked

0.568 spearman

Unit: correlation · Direction: higher

Uncertainty: Not reported

Scored: Not reported · Eligible: Not reported

source checkedProteinGym: Large-Scale Benchmarks for Protein Fitness Prediction and Design · Table 3, row(MSAT), column(Supervised substitutions, average over splits, Spearman)

Source checking is not independent reproduction.

Embed. MSAT on ProteinGym SUP-SUB-SPEARMAN-CONTIG: Supervised substitutions, contiguous split, Spearman

Supervised fitness prediction under three cross-validation schemes: contiguous, modulo and random splits.

Author-reported evaluation · Evaluation metadata: source checked

0.525 spearman

Unit: correlation · Direction: higher

Uncertainty: Not reported

Scored: Not reported · Eligible: Not reported

source checkedProteinGym: Large-Scale Benchmarks for Protein Fitness Prediction and Design · Table 3, row(MSAT), column(Supervised substitutions, contiguous split, Spearman)

Source checking is not independent reproduction.

Embed. MSAT on ProteinGym SUP-SUB-SPEARMAN-MOD: Supervised substitutions, modulo split, Spearman

Supervised fitness prediction under three cross-validation schemes: contiguous, modulo and random splits.

Author-reported evaluation · Evaluation metadata: source checked

0.538 spearman

Unit: correlation · Direction: higher

Uncertainty: Not reported

Scored: Not reported · Eligible: Not reported

source checkedProteinGym: Large-Scale Benchmarks for Protein Fitness Prediction and Design · Table 3, row(MSAT), column(Supervised substitutions, modulo split, Spearman)

Source checking is not 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.

2 evidence rows matching the loaded filters

Claims, original sources and review scope · Release 2026-09-17-134cd1815de8
Property and statementOriginal source and locationReview and provenance
description

The split of ProteinGym substitution DMS assays that ProteinGym evaluated on. The upstream dataset release is not catalogued here, so no claim is made that this matches its original splits.

Context-only references
ProteinGym: Large-Scale Benchmarks for Protein Fitness Prediction and Design

Original source ↗

No field-specific location recorded

Version: Primary full-text snapshot retrieved 2026-09-17; exact bytes pinned by SHA-256
Retrieved: 2026-09-17T08:06:28.522924+00:00

not individually reviewed

No individual claim review recorded

Audit details

Field: description

Source artifact SHA-256: a3b08cc4a6befd64620cf0f287d78d55a36dc2639955f52c5655f83833c50104

Hash scope: Exact retrieved primary paper artifact bytes.

Inspected artifact

name

ProteinGym substitution DMS assays (ProteinGym split)

Context-only references
ProteinGym: Large-Scale Benchmarks for Protein Fitness Prediction and Design

Original source ↗

No field-specific location recorded

Version: Primary full-text snapshot retrieved 2026-09-17; exact bytes pinned by SHA-256
Retrieved: 2026-09-17T08:06:28.522924+00:00

not individually reviewed

No individual claim review recorded

Audit details

Field: name

Source artifact SHA-256: a3b08cc4a6befd64620cf0f287d78d55a36dc2639955f52c5655f83833c50104

Hash scope: Exact retrieved primary paper artifact bytes.

Inspected artifact

Sources and history

Release 2026-09-17-134cd1815de8 · Record review: source checked

1 source records and release historyDownload this release
Technical metadata and extraction receipts

Stable ID: proteingym-dataset-proteingym-substitution-dms-assays

areas
proteins-complexes
missing metadata
version: unreported; url: unextracted
Related records

Suggest a correction