ProteinMPNN

ProteinMPNN performs deep learning-based protein sequence design with outstanding performance in both in silico and experimental tests.

Key Features:

• High sequence recovery: ProteinMPNN achieves 52.4% sequence recovery on native protein backbones compared with 32.9% for Rosetta.
• Coupled sequence design: The amino acid sequence at different positions can be coupled between single or multiple chains.
• Broad applicability: The method enables application to a wide range of current protein design challenges including protein monomers, cyclic homo-oligomers, tetrahedral nanoparticles, and target-binding proteins.
• Design rescue capability: ProteinMPNN rescues previously failed designs made using Rosetta or AlphaFold.

Scientific Applications:

• Protein monomer design: ProteinMPNN rescues failed designs of protein monomers as validated by x-ray crystallography, cryo-electron microscopy, and functional studies.
• Cyclic homo-oligomer design: ProteinMPNN enables design of cyclic homo-oligomers with experimental validation.
• Tetrahedral nanoparticle design: ProteinMPNN supports design of tetrahedral nanoparticle structures.
• Target-binding protein design: ProteinMPNN rescues failed designs of proteins that bind specific targets.

Methodology:

ProteinMPNN uses a deep learning-based approach for protein sequence design.