Protein sequence design with a learned potential

• Published in: Nature communications Vol. 13; no. 1; pp. 746 - 11
• Main Authors: Anand, Namrata, Eguchi, Raphael, Mathews, Irimpan I., Perez, Carla P., Derry, Alexander, Altman, Russ B., Huang, Po-Ssu
• Format: Journal Article
• Language: English
• Published: London Nature Publishing Group UK 08.02.2022; Nature Publishing Group; Nature Portfolio
• Subjects: 631/114/1305; 631/114/469; Amino acid sequence; Amino Acid Sequence - genetics; Artificial neural networks; Backbone; BASIC BIOLOGICAL SCIENCES; Computer Simulation; Crystal structure; Crystallography, X-Ray; Deep Learning; Design; Humanities and Social Sciences; Learning algorithms; Machine learning; Models, Molecular; multidisciplinary; Neural networks; protein design; Protein Domains - genetics; Protein engineering; Protein Engineering - methods; Protein Folding; Protein structure; Proteins; Science; Science (multidisciplinary); Topology
• ISSN: 2041-1723; 2041-1723
• DOI: 10.1038/s41467-022-28313-9

The task of protein sequence design is central to nearly all rational protein engineering problems, and enormous effort has gone into the development of energy functions to guide design. Here, we investigate the capability of a deep neural network model to automate design of sequences onto protein backbones, having learned directly from crystal structure data and without any human-specified priors. The model generalizes to native topologies not seen during training, producing experimentally stable designs. We evaluate the generalizability of our method to a de novo TIM-barrel scaffold. The model produces novel sequences, and high-resolution crystal structures of two designs show excellent agreement with in silico models. Our findings demonstrate the tractability of an entirely learned method for protein sequence design. Rational protein design to achieve a given protein backbone conformation is needed to engineer specific functions. Here Anand et al. describe a machine learning method using a learned neural network potential for fixed-backbone protein design.