ProteinMPNN
The field's default inverse-folding model: give it a backbone, get sequences that fold into it.
About
ProteinMPNN (Dauparas et al., Science 2022) is the deep-learning standard for fixed-backbone protein sequence design. It takes a protein backbone structure as input and outputs amino acid sequences likely to fold into it, using a message-passing neural network over the backbone graph. In practice it replaced both physics-based Rosetta design and earlier learned methods as the sequence-design step in de novo pipelines.
It sits at the center of the modern design stack: RFdiffusion generates backbones, ProteinMPNN writes sequences for them, and AlphaFold-class models validate the result. The Baker lab later extended the idea to LigandMPNN, which conditions on small molecules, nucleotides, and metals as well. The code is MIT-licensed and runs on a single GPU, with variants for soluble proteins, membrane proteins, and C-alpha-only inputs.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
Author-reported (Dauparas et al., Science 2022): state-of-the-art native sequence recovery across benchmarks; widely adopted as the sequence-design step in published de novo design pipelines; extensive informal community replication
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 9/10
Scored 9/10 (metrics tier, metrics-v1). Evidence: Science 2022 paper (Dauparas et al., 378:49-56), repo github.com/dauparas/ProteinMPNN, and MIT license metadata confirm the method and its standing; its role as the sequence-design step in RFdiffusion and published binder pipelines is documented across the field. Strengths: landmark inverse-folding method, permissive license, ubiquitous adoption, extended by LigandMPNN to ligands and metals. Docked for: (1) fixed-backbone only, it cannot design backbones itself; (2) designed sequences still need experimental or structure-prediction validation; (3) research tool with no commercial support channel. License: MIT.
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 10.0/10
- Maintenance · recency, releases, issue load
- 7.0/10
- Openness · license + docs
- 9.0/10
- Manual review · sanity read + red-flag check
- 9.0/10
- · Field-standard inverse-folding model; sequence-design step in RFdiffusion and published binder pipelines.
- · Fixed-backbone only; designed sequences need experimental or structure-prediction validation.
- · Extended by LigandMPNN (Nature Methods 2025) to ligands, nucleotides, and metals.