Skip to main content

ProteinMPNN for Structure-based Design

ProteinMPNN is a GNN-based AI method for designing a protein sequence given a protein structure, a process known as structure-based design or inversefolding. ProteinMPNN uses various parameters to assist with protein design, including the temperature parameter, which controls diversity in the designs: values from 0.1 to 0.5 result in less diverse designs, while values of 0.5 and above lead to more diverse outcomes. LigandMPNN is coming soon.

How to Use ProteinMPNN​

Use ProteinMPNN to design P00698

Inputs​

PDB file

A protein structure. Note this version only handles single chains. Updates coming soon.

Outputs​

Protein sequence
A designed sequence of the same length as the input structure.

Examples​

Use ProteinMPNN to design P00698 with a temperature of 1.0 and show 3 results

Use ProteinMPNN to design chain A of 4KRL with a temperature of 1.0, and fix all residues of chain B of 4KRL that are part of the interface with chain A