Improving de novo protein binder design with deep learning

Nathaniel R Bennett; Brian Coventry; Inna Goreshnik; Buwei Huang; Aza Allen; Dionne Vafeados; Ying Po Peng; Justas Dauparas; Minkyung Baek; Lance Stewart; Frank DiMaio; Steven De Munck; Savvas N Savvides; David Baker

doi:10.1038/s41467-023-38328-5

Improving de novo protein binder design with deep learning

Nat Commun. 2023 May 6;14(1):2625. doi: 10.1038/s41467-023-38328-5.

Authors

Nathaniel R Bennett^#^{1

2

3}, Brian Coventry^#^{1

2

4}, Inna Goreshnik^{1

2}, Buwei Huang^{1

2

5}, Aza Allen^{1

2}, Dionne Vafeados^{1

2}, Ying Po Peng^{1

2}, Justas Dauparas^{1

2}, Minkyung Baek^{1

2}, Lance Stewart^{1

2}, Frank DiMaio^{1

2}, Steven De Munck^{6

7}, Savvas N Savvides^{6

7}, David Baker^{8

9

10}

Affiliations

¹ Department of Biochemistry, University of Washington, Seattle, WA, USA.
² Institute for Protein Design, University of Washington, Seattle, WA, USA.
³ Molecular Engineering Graduate Program, University of Washington, Seattle, WA, USA.
⁴ Howard Hughes Medical Institute, University of Washington, Seattle, WA, USA.
⁵ Department of Bioengineering, University of Washington, Seattle, WA, USA.
⁶ VIB-UGent Center for Inflammation Research, Ghent, Belgium.
⁷ Unit for Structural Biology, Department of Biochemistry and Microbiology, Ghent University, Ghent, Belgium.
⁸ Department of Biochemistry, University of Washington, Seattle, WA, USA. dabaker@uw.edu.
⁹ Institute for Protein Design, University of Washington, Seattle, WA, USA. dabaker@uw.edu.
¹⁰ Howard Hughes Medical Institute, University of Washington, Seattle, WA, USA. dabaker@uw.edu.

^# Contributed equally.

Abstract

Recently it has become possible to de novo design high affinity protein binding proteins from target structural information alone. There is, however, considerable room for improvement as the overall design success rate is low. Here, we explore the augmentation of energy-based protein binder design using deep learning. We find that using AlphaFold2 or RoseTTAFold to assess the probability that a designed sequence adopts the designed monomer structure, and the probability that this structure binds the target as designed, increases design success rates nearly 10-fold. We find further that sequence design using ProteinMPNN rather than Rosetta considerably increases computational efficiency.

Publication types

Research Support, Non-U.S. Gov't
Research Support, U.S. Gov't, Non-P.H.S.

MeSH terms

Deep Learning*
Protein Binding
Protein Engineering
Proteins / metabolism

Substances

Proteins