MPNN Perplexity

Score protein sequences by ProteinMPNN perplexity against a fixed backbone; differentiable.

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 constraint 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/constraint/sequence_scoring/mpnn_perplexity_constraint.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

Score proposals by ProteinMPNN perplexity against the configured backbone.

API Reference

ConfigMpnnPerplexityConfig Source

Configuration for ProteinMPNN perplexity scoring.

structure_input · InverseFoldingStructureInput · required

Backbone structure, optional chains_to_redesign, and fixed positions for direct ProteinMPNN scoring.

model_choice · enum · default: "proteinmpnn"

Weights: proteinmpnn (=v_48_020), v_48_{002,010,030} noise variants, abmpnn, soluble.

Options: proteinmpnn, v_48_002, v_48_010, v_48_030, abmpnn, soluble

temperature · number · default: "1.0"

Softmax temperature for relaxing optimizer logits before ProteinMPNN scoring.

use_ste · boolean · default: "True"

Hard one-hot forward pass with soft-probability gradients.

device · string · default: "cuda"

Device for ProteinMPNN execution, e.g. ‘cuda’ or ‘cuda:0’.

seed · integer

Seed for ProteinMPNN decoding-order sampling. None lets proto-tools choose a fresh seed.

score_mode · enum · default: "ppl"

Return ProteinMPNN perplexity by default, or raw mean NLL when set to ‘nll’.

Options: nll, ppl

logit_scale · number · default: "1.0"

Pre-scale raw logits before ProteinMPNN; gradients are scaled back by the same factor.

sequence_bias · SequenceLogitBiasConfig

Declarative sequence-symbol bias (canonical 20-AA protein) added before ProteinMPNN.

Usage

python

from proto_language.core import Constraint
from proto_language.constraint import mpnn_perplexity_constraint, MpnnPerplexityConfig

constraint = Constraint(
    inputs=[segment],
    function=mpnn_perplexity_constraint,
    function_config=MpnnPerplexityConfig(
        # Configure parameters here
    ),
)

scores = constraint.evaluate()

Metadata

Property	Value
Key	mpnn-perplexity
Function	mpnn_perplexity_constraint
Category	sequence_scoring
Mode	dual
Uses GPU	True
Supported Types	protein