rewirebio.iobenchmarks
Configuration

ProteinMPNN

ProteinMPNN designs amino-acid sequences for a supplied protein backbone.

Sources (5)dauparas/ProteinMPNN: README.md; dauparas/ProteinMPNN: training/README.md; dauparas/ProteinMPNN: protein_mpnn_utils.py; proteinmpnn: Journal full-text XML; proteinmpnn-supp: Publisher supplementary archive · ProteinMPNN paper: main-text model development; training/README.md: multi-chain training set, list.csv and cluster manifests; protein_mpnn_utils.py: ProteinMPNN; Supplementary Materials: Methods for training multi-chain models / Training data

1 evaluation · 5 results

How it worksProteinMPNN workflow
ProteinMPNN workflow1. Protein backbone. Then: 2. Structural graph features. Then: 3. Message-passing model. Then: 4. Constrained sequence samplingProteinMPNN workflow1. Protein backbone. Then: 2. Structural graph features. Then: 3. Message-passing model. Then: 4. Constrained sequence samplingProteinMPNN workflow1. Protein backbone. Then: 2. Structural graph features. Then: 3. Message-passing model. Then: 4. Constrained sequence sampling

Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.

Sources (5)dauparas/ProteinMPNN: README.md; dauparas/ProteinMPNN: training/README.md; dauparas/ProteinMPNN: protein_mpnn_utils.py; proteinmpnn: Journal full-text XML; proteinmpnn-supp: Publisher supplementary archive · ProteinMPNN paper: main-text model development; training/README.md: multi-chain training set, list.csv and cluster manifests; protein_mpnn_utils.py: ProteinMPNN; Supplementary Materials: Methods for training multi-chain models / Training data

Overview

Model type

Structure-conditioned message-passing sequence design model

Inputs

Protein backbone coordinates, with optional fixed residues, chain choices, tied positions and amino-acid constraints.

Outputs

Designed sequences, sequence scores and conditional amino-acid probabilities.

Sources (5)dauparas/ProteinMPNN: README.md; dauparas/ProteinMPNN: training/README.md; dauparas/ProteinMPNN: protein_mpnn_utils.py; proteinmpnn: Journal full-text XML; proteinmpnn-supp: Publisher supplementary archive · ProteinMPNN paper: main-text model development; training/README.md: multi-chain training set, list.csv and cluster manifests; protein_mpnn_utils.py: ProteinMPNN; Supplementary Materials: Methods for training multi-chain models / Training data

limited source coverage · Automated source review, 2026-09-16. All specifications and missing details

Model Capabilities & Hardware Requirements

Deployment facts recorded for ProteinMPNN. Fields the record does not carry are marked not recorded.

ParametersNot recorded
Minimum VRAMNot recorded
Inference LatencyNot recorded
Context WindowNot recorded
AccessNot recorded
LicenseNot recorded

Evaluations and results

1 evaluation · 5 results. Different protocols are not a single leaderboard.

Dataset subset: ProteinGym substitution DMS assays (ProteinGym split)

Sorted by AUROC (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 configurationProtocol and datasetAUROCMCCNDCG at 10%SpearmanTop 10% recall
ProteinMPNNEvaluationProteinGym ZS-SUB-AUC: Zero-shot substitutions, AUC0.639Not reportedNot reportedNot reportedNot reported
ProteinMPNNEvaluationProteinGym ZS-SUB-MCC: Zero-shot substitutions, MCCNot reported0.196Not reportedNot reportedNot reported
ProteinMPNNEvaluationProteinGym ZS-SUB-NDCG: Zero-shot substitutions, NDCG@10%Not reportedNot reported0.713Not reportedNot reported
ProteinMPNNEvaluationProteinGym ZS-SUB-RECALL: Zero-shot substitutions, top 10% recallNot reportedNot reportedNot reportedNot reported0.186
ProteinMPNNEvaluationProteinGym ZS-SUB-SPEARMAN: Zero-shot substitutions, SpearmanNot reportedNot reportedNot reported0.258Not reported
All 5 result rows with coverage, uncertainty and sources
Filter evaluations

Applied filters: All linked evaluations

Exact evaluated configurations and original reported results
Tested configurationProtocol and datasetFindingEvidence and details
Configuration: ProteinMPNNTask: ProteinGym ZS-SUB-AUC: Zero-shot substitutions, AUC
Dataset subset: ProteinGym substitution DMS assays (ProteinGym split)
0.639 AUROC
unitless · higher

Uncertainty: Not yet extracted

Coverage: Not reported scored / Not reported eligible

Independent external evaluation · Source checked
Methods, coverage and source

ProteinMPNN on ProteinGym ZS-SUB-AUC: Zero-shot substitutions, AUC

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

Aggregation: Not reported

ProteinGym: Large-Scale Benchmarks for Protein Fitness Prediction and Design · Table 2, row(ProteinMPNN), column(Zero-shot substitutions, AUC)
Configuration: ProteinMPNNTask: ProteinGym ZS-SUB-MCC: Zero-shot substitutions, MCC
Dataset subset: ProteinGym substitution DMS assays (ProteinGym split)
0.196 MCC
unitless · higher

Uncertainty: Not yet extracted

Coverage: Not reported scored / Not reported eligible

Independent external evaluation · Source checked
Methods, coverage and source

ProteinMPNN on ProteinGym ZS-SUB-MCC: Zero-shot substitutions, MCC

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

Aggregation: Not reported

ProteinGym: Large-Scale Benchmarks for Protein Fitness Prediction and Design · Table 2, row(ProteinMPNN), column(Zero-shot substitutions, MCC)
Configuration: ProteinMPNNTask: ProteinGym ZS-SUB-NDCG: Zero-shot substitutions, NDCG@10%
Dataset subset: ProteinGym substitution DMS assays (ProteinGym split)
0.713 NDCG at 10%
unitless · higher

Uncertainty: Not yet extracted

Coverage: Not reported scored / Not reported eligible

Independent external evaluation · Source checked
Methods, coverage and source

ProteinMPNN 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.

Aggregation: Not reported

ProteinGym: Large-Scale Benchmarks for Protein Fitness Prediction and Design · Table 2, row(ProteinMPNN), column(Zero-shot substitutions, NDCG@10%)
Configuration: ProteinMPNNTask: ProteinGym ZS-SUB-RECALL: Zero-shot substitutions, top 10% recall
Dataset subset: ProteinGym substitution DMS assays (ProteinGym split)
0.186 Top 10% recall
fraction · higher

Uncertainty: Not yet extracted

Coverage: Not reported scored / Not reported eligible

Independent external evaluation · Source checked
Methods, coverage and source

ProteinMPNN 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.

Aggregation: Not reported

ProteinGym: Large-Scale Benchmarks for Protein Fitness Prediction and Design · Table 2, row(ProteinMPNN), column(Zero-shot substitutions, top 10% recall)
Configuration: ProteinMPNNTask: ProteinGym ZS-SUB-SPEARMAN: Zero-shot substitutions, Spearman
Dataset subset: ProteinGym substitution DMS assays (ProteinGym split)
0.258 Spearman
unitless · higher

Uncertainty: Not yet extracted

Coverage: Not reported scored / Not reported eligible

Independent external evaluation · Source checked
Methods, coverage and source

ProteinMPNN on ProteinGym ZS-SUB-SPEARMAN: Zero-shot substitutions, Spearman

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

Aggregation: Not reported

ProteinGym: Large-Scale Benchmarks for Protein Fitness Prediction and Design · Table 2, row(ProteinMPNN), column(Zero-shot substitutions, Spearman)

Source checking is not independent reproduction. Release 2026-11-21-adc15377fab2.

Use this configuration

How it works, versions and access

Related profile: ProteinMPNN. This page retains the exact record and its evaluation context.

This configuration

Configuration of ProteinMPNN: Fitness prediction model.

record
ProteinMPNN
configuration
Not reported
entity type
Configuration

How it works

How it works

ProteinMPNN converts a supplied backbone into a graph whose edges encode interatomic distances. Message-passing layers update node and edge features, and an autoregressive decoder samples amino acids while conditioning on the backbone and previously assigned residues. Fixed residues, tied positions and chain choices change the design task and must accompany its result.

Sources (5)dauparas/ProteinMPNN: README.md; dauparas/ProteinMPNN: training/README.md; dauparas/ProteinMPNN: protein_mpnn_utils.py; proteinmpnn: Journal full-text XML; proteinmpnn-supp: Publisher supplementary archive · ProteinMPNN paper: main-text model development; training/README.md: multi-chain training set, list.csv and cluster manifests; protein_mpnn_utils.py: ProteinMPNN; Supplementary Materials: Methods for training multi-chain models / Training data
Versions and reproducibility

v_48_002, v_48_010, v_48_020 and v_48_030; distinct soluble and C-alpha-only weights. Structure-size and memory dependent. README --max_length is an implementation guard, not a validated scientific context limit.

Sources (5)dauparas/ProteinMPNN: README.md; dauparas/ProteinMPNN: training/README.md; dauparas/ProteinMPNN: protein_mpnn_utils.py; proteinmpnn: Journal full-text XML; proteinmpnn-supp: Publisher supplementary archive · ProteinMPNN paper: main-text model development; training/README.md: multi-chain training set, list.csv and cluster manifests; protein_mpnn_utils.py: ProteinMPNN; Supplementary Materials: Methods for training multi-chain models / Training data
Strengths, limitations and unresolved questions

Strengths and limitations

Strengths and considerations

Limitations and conditions

Profile review details

Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied.

Stable record: discovery-model-proteinmpnn

Specifications

Inputs, training, access and other details

Explanatory profile: limited source coverage · Automated source review, 2026-09-16. Review applies to the cited claims; unresolved fields are listed below. Numerical results retain their own review status.

Inputs, outputs and configuration
PropertyDescription and evidence
Model typeStructure-conditioned message-passing sequence design model
Sources (5)dauparas/ProteinMPNN: README.md; dauparas/ProteinMPNN: training/README.md; dauparas/ProteinMPNN: protein_mpnn_utils.py; proteinmpnn: Journal full-text XML; proteinmpnn-supp: Publisher supplementary archive · ProteinMPNN paper: main-text model development; training/README.md: multi-chain training set, list.csv and cluster manifests; protein_mpnn_utils.py: ProteinMPNN; Supplementary Materials: Methods for training multi-chain models / Training data
ArchitectureMessage-passing encoder-decoder with structural interatomic-distance features and edge updates; sequences are sampled with the configured autoregressive decoding procedure.
Sources (5)dauparas/ProteinMPNN: README.md; dauparas/ProteinMPNN: training/README.md; dauparas/ProteinMPNN: protein_mpnn_utils.py; proteinmpnn: Journal full-text XML; proteinmpnn-supp: Publisher supplementary archive · ProteinMPNN paper: main-text model development; training/README.md: multi-chain training set, list.csv and cluster manifests; protein_mpnn_utils.py: ProteinMPNN; Supplementary Materials: Methods for training multi-chain models / Training data
InputsProtein backbone coordinates, with optional fixed residues, chain choices, tied positions and amino-acid constraints.
Sources (5)dauparas/ProteinMPNN: README.md; dauparas/ProteinMPNN: training/README.md; dauparas/ProteinMPNN: protein_mpnn_utils.py; proteinmpnn: Journal full-text XML; proteinmpnn-supp: Publisher supplementary archive · ProteinMPNN paper: main-text model development; training/README.md: multi-chain training set, list.csv and cluster manifests; protein_mpnn_utils.py: ProteinMPNN; Supplementary Materials: Methods for training multi-chain models / Training data
OutputsDesigned sequences, sequence scores and conditional amino-acid probabilities.
Sources (5)dauparas/ProteinMPNN: README.md; dauparas/ProteinMPNN: training/README.md; dauparas/ProteinMPNN: protein_mpnn_utils.py; proteinmpnn: Journal full-text XML; proteinmpnn-supp: Publisher supplementary archive · ProteinMPNN paper: main-text model development; training/README.md: multi-chain training set, list.csv and cluster manifests; protein_mpnn_utils.py: ProteinMPNN; Supplementary Materials: Methods for training multi-chain models / Training data
ParametersThe inspected model implementation is configured through encoder/decoder depth and feature width. The paper and training README do not state an exact total for every released checkpoint. · Not reported in inspected sources
Sources (5)dauparas/ProteinMPNN: README.md; dauparas/ProteinMPNN: training/README.md; dauparas/ProteinMPNN: protein_mpnn_utils.py; proteinmpnn: Journal full-text XML; proteinmpnn-supp: Publisher supplementary archive · ProteinMPNN paper: main-text model development; training/README.md: multi-chain training set, list.csv and cluster manifests; protein_mpnn_utils.py: ProteinMPNN; Supplementary Materials: Methods for training multi-chain models / Training data
Known versionsv_48_002, v_48_010, v_48_020 and v_48_030; distinct soluble and C-alpha-only weights.
Sources (5)dauparas/ProteinMPNN: README.md; dauparas/ProteinMPNN: training/README.md; dauparas/ProteinMPNN: protein_mpnn_utils.py; proteinmpnn: Journal full-text XML; proteinmpnn-supp: Publisher supplementary archive · ProteinMPNN paper: main-text model development; training/README.md: multi-chain training set, list.csv and cluster manifests; protein_mpnn_utils.py: ProteinMPNN; Supplementary Materials: Methods for training multi-chain models / Training data
Training dataReleased multi-chain training set of PDB biological units, with chain metadata and validation/test cluster manifests. The documented set is dated 2 August 2021.
Sources (5)dauparas/ProteinMPNN: README.md; dauparas/ProteinMPNN: training/README.md; dauparas/ProteinMPNN: protein_mpnn_utils.py; proteinmpnn: Journal full-text XML; proteinmpnn-supp: Publisher supplementary archive · ProteinMPNN paper: main-text model development; training/README.md: multi-chain training set, list.csv and cluster manifests; protein_mpnn_utils.py: ProteinMPNN; Supplementary Materials: Methods for training multi-chain models / Training data
Training cutoffThe released PDB training-set snapshot is dated 2021-08-02; preserve its chain-level deposition metadata and cluster split for a run.
Sources (5)dauparas/ProteinMPNN: README.md; dauparas/ProteinMPNN: training/README.md; dauparas/ProteinMPNN: protein_mpnn_utils.py; proteinmpnn: Journal full-text XML; proteinmpnn-supp: Publisher supplementary archive · ProteinMPNN paper: main-text model development; training/README.md: multi-chain training set, list.csv and cluster manifests; protein_mpnn_utils.py: ProteinMPNN; Supplementary Materials: Methods for training multi-chain models / Training data
Context limitsStructure-size and memory dependent. README --max_length is an implementation guard, not a validated scientific context limit.
Sources (5)dauparas/ProteinMPNN: README.md; dauparas/ProteinMPNN: training/README.md; dauparas/ProteinMPNN: protein_mpnn_utils.py; proteinmpnn: Journal full-text XML; proteinmpnn-supp: Publisher supplementary archive · ProteinMPNN paper: main-text model development; training/README.md: multi-chain training set, list.csv and cluster manifests; protein_mpnn_utils.py: ProteinMPNN; Supplementary Materials: Methods for training multi-chain models / Training data
Weights licenceSeparate checkpoint-distribution terms are not stated in the inspected release documentation and licence material. The source-code licence alone is not recorded as an explicit weight grant. · Not reported in inspected sources
Sources (6)dauparas/ProteinMPNN: README.md; dauparas/ProteinMPNN: training/README.md; dauparas/ProteinMPNN: protein_mpnn_utils.py; proteinmpnn: Journal full-text XML; proteinmpnn-supp: Publisher supplementary archive; dauparas/ProteinMPNN: LICENSE · ProteinMPNN paper: main-text model development; training/README.md: multi-chain training set, list.csv and cluster manifests; protein_mpnn_utils.py: ProteinMPNN; Supplementary Materials: Methods for training multi-chain models / Training data; LICENSE: licence text
AccessOfficial project documentation and implementation: https://github.com/dauparas/ProteinMPNN
Sources (5)dauparas/ProteinMPNN: README.md; dauparas/ProteinMPNN: training/README.md; dauparas/ProteinMPNN: protein_mpnn_utils.py; proteinmpnn: Journal full-text XML; proteinmpnn-supp: Publisher supplementary archive · ProteinMPNN paper: main-text model development; training/README.md: multi-chain training set, list.csv and cluster manifests; protein_mpnn_utils.py: ProteinMPNN; Supplementary Materials: Methods for training multi-chain models / Training data
Code licenceMIT
Sourcesdauparas/ProteinMPNN: LICENSE · LICENSE: licence text

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-adc15377fab2
Property and statementOriginal source and locationReview and provenance
Relationship: family
discovery-model-proteinmpnn
Individual claims
ProteinGym: Large-Scale Benchmarks for Protein Fitness Prediction and Design

Original source ↗

Section 4.1 Zero-shot benchmarks; Table 2 named model row

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

source checked

["automated-source-review"] · 2026-09-23

Audit details

Source review establishes this relationship only. Exact evaluated configurations and original numerical review status remain unchanged. The primary table and baseline descriptions identify the evaluated family. ESM-1v remains an ensemble; ESM-2 remains 15B; inverse-folding conditions remain distinct.

Field: links:family:discovery-model-proteinmpnn

Claim: model-evaluation-identity-a15c945f534a82d3675f

Source artifact SHA-256: a3b08cc4a6befd64620cf0f287d78d55a36dc2639955f52c5655f83833c50104

Hash scope: Exact retrieved primary paper artifact bytes.

Inspected artifact

Sources and history

Release 2026-11-21-adc15377fab2 · Record review: source checked

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

Technical metadata for this record is tracked under stable ID proteingym-method-proteinmpnn.

Source locator: Table 2, row(ProteinMPNN)

Related records

How to cite

ProteinMPNN. Rewire biological model benchmark database, release 2026-11-21-adc15377fab2. https://benchmarks.rewirebio.io/database/configuration/proteingym-method-proteinmpnn/

Release 2026-11-21-adc15377fab2. 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

Suggest a correction