ColabMPNN

ColabMPNN is a Julia wrapper for the MPNN submodule of the ColabDesign Python package, which can be found at https://github.com/sokrypton/ColabDesign/tree/main.

For more details about usage and function arguments, see the original Python documentation

Installation

Add ColabMPNN to your Julia environment in the REPL:

]add https://github.com/MurrellGroup/ColabMPNN.jl

Usage

Create a model using the mk_mpnn_model function.

mpnn_model = mk_mpnn_model()

In order to sample, chains from a PDB file must first be prepared.

prep_inputs(mpnn_model, pdb_filename="example.pdb", chain="A")

Sample sequences using the sample function, or in parallel with sample_parallel, with the model as the first argument. These functions return a Samples instance.

samples = sample_parallel(mpnn_model, batch=10, temperature=0.1)

Sampling returns a Samples instance with the following fields:

• seq::Vector{String}
• seqid::Vector{Float64}
• score::Vector{Float64}
• logits::Array{Float32, 3}
• decoding_order::Array{Int32, 3}
• S::Array{Float32, 3}

Acknowledgements

This is a thin Julia wrapper of a JAX port (by Sergey Ovchinnikov and others) of ProteinMPNN (by Justas Dauparas and others).