ProteinMPNN Inverse Folding

ProteinMPNN structure-conditioned protein sequence design

License: ProteinMPNN is open source and free for academic and commercial use under an MIT license. Please refer to the license for full terms.

This generator is open source. Any third-party models, product names, or trademarks referenced are the property of their respective owners, and Proto is not affiliated with them.

---

Tools Used Tools Used Source Source Cite Cite

Go to Tool Page

proto-bio/proto-language/proto_language/generator/proteinmpnn_generator.py

View source

@article{dauparas2022proteinmpnn,
  title={Robust deep learning--based protein sequence design using ProteinMPNN},
  author={Dauparas, Justas and Anishchenko, Ivan and Bennett, Nathaniel and Bai, Hua and Ragotte, Robert J and Milles, Lukas F and Wicky, Basile IM and Courber, Alexis and de Haas, Rob J and Bethel, Neville and others},
  journal={Science},
  volume={378},
  number={6615},
  pages={49--56},
  year={2022},
  publisher={American Association for the Advancement of Science},
  doi={10.1126/science.add2187}
}

Copy citation

Protein sequence generator using ProteinMPNN inverse folding model.

This generator uses ProteinMPNN to design protein sequences that are predicted to fold into a given 3D backbone structure. Unlike mutation-based generators that refine existing sequences, ProteinMPNN generates sequences directly from structural information.

ProteinMPNN is particularly effective for:

• Redesigning existing proteins while maintaining fold
• Designing sequences for computationally generated backbones
• Creating sequence diversity for experimental screening
• Stabilizing protein structures through sequence optimization

API Reference

ConfigProteinMPNNGeneratorConfig Source

Configuration object for ProteinMPNNGenerator.

This class defines configuration parameters for the ProteinMPNN generator, which uses the ProteinMPNN inverse folding model to design protein sequences conditioned on a given 3D backbone structure.

ProteinMPNN is a message-passing neural network that predicts amino acid sequences likely to fold into a specified protein backbone structure. It excels at redesigning existing proteins while maintaining structural compatibility.

model_choice · enum · default: "proteinmpnn"

ProteinMPNN weights: ‘proteinmpnn’ (general), ‘abmpnn’ (antibody), or ‘soluble’ (soluble proteins).

Options: proteinmpnn, abmpnn, soluble

structure_inputs · array

Structure(s) with optional chains_to_redesign and fixed_positions constraints.

output_chain_id · string

When sampling a multi-chain structure, write only this chain’s sequence to the target segment.

temperature · number · default: "0.1"

Randomness of sampling (0-1). Near 0 is deterministic; near 1 is proportional to model probs.

excluded_amino_acids · array

Single-letter amino-acid codes to forbid in the designed sequence (e.g. ‘C’ to avoid disulfides).

batch_size · integer · default: "1"

Number of sequences to process simultaneously on GPU

device · string · default: "cuda"

GPU device for inference (e.g. ‘cuda’ or ‘cuda:0’).

verbose · boolean · default: "False"

Whether to print status messages during execution.

Usage

python

>>> from proto_language.generator import ProteinMPNNGenerator, ProteinMPNNGeneratorConfig
>>> from proto_language.core import Segment
>>> config = ProteinMPNNGeneratorConfig(
...     structure_inputs="/path/to/backbone.pdb",
...     temperature=0.1,
... )
>>> gen = ProteinMPNNGenerator(config)
>>> segment = Segment(length=100, sequence_type="protein")
>>> gen.assign(segment)
>>> gen.sample()  # Generates num_proposals sequences from the backbone

Metadata

Property	Value
Key	proteinmpnn
Class	ProteinMPNNGenerator
Category	inverse_folding
Input Type	structure
Uses GPU	True
Supported Sequence Types	protein
Allows Empty Start	False