Designs optimized protein sequences from structural backbones using deep learning-based inverse folding.
The ProteinMPNN skill empowers researchers and bioengineers to solve the inverse folding problem by generating high-quality protein sequences for specific 3D backbones. It is highly versatile, supporting sequences for RFdiffusion backbones, the redesign of existing proteins, and the optimization of sequences for expression or stability. With features like fixed-residue design, multi-chain tying, and adjustable sampling temperatures, it provides a robust framework for exploring protein sequence space while maintaining structural integrity. Whether performing de novo design or interface redesign, this skill streamlines the transition from structural coordinates to viable biological sequences.
주요 기능
01High-accuracy inverse folding for structural backbones
02Integration with LigandMPNN and SolubleMPNN variants
03Support for fixed-residue and chain-specific design
04Multi-state and negative design capabilities
05Adjustable sampling temperature for sequence diversity control
0675 GitHub stars
사용 사례
01Redesigning protein-protein interfaces to improve binding affinity
02Generating sequences for de novo protein backbones created via RFdiffusion
03Optimizing protein sequences for better solubility and expression in the lab
