Accurate structure prediction of biomolecular interactions with AlphaFold 3

Josh Abramson,Jonas Adler,Jack Dunger,Richard Evans,Tim Green,Alexander Pritzel,Olaf Ronneberger,Lindsay Willmore,Andrew J. Ballard,Joshua Bambrick,Sebastian W. Bodenstein,David A. Evans,Chia-Chun Hung,Michael O’Neill,David Reiman,Kathryn Tunyasuvunakool,Zachary Wu,Akvilė Žemgulytė,Eirini Arvaniti,Charles Beattie,Ottavia Bertolli,Alex Bridgland,Alexey Cherepanov,Miles Congreve,Alexander I. Cowen-Rivers,Andrew Cowie,Michael Figurnov,Fabian B. Fuchs,Hannah Gladman,Rishub Jain,Yousuf A. Khan,Caroline M. R. Low,Kuba Perlin,Anna Potapenko,Pascal Savy,Sukhdeep Singh,Adrian Stecula,Ashok Thillaisundaram,Catherine Tong,Sergei Yakneen,Ellen D. Zhong,Michal Zielinski,Augustin Žídek,Victor Bapst,Pushmeet Kohli,Max Jaderberg,Demis Hassabis,John M. Jumper

DOI: https://doi.org/10.1038/s41586-024-07487-w

IF: 64.8

2024-05-09

Nature

Abstract:The introduction of AlphaFold 2 1 has spurred a revolution in modelling the structure of proteins and their interactions, enabling a huge range of applications in protein modelling and design 2–6 . In this paper, we describe our AlphaFold 3 model with a substantially updated diffusion-based architecture, which is capable of joint structure prediction of complexes including proteins, nucleic acids, small molecules, ions, and modified residues. The new AlphaFold model demonstrates significantly improved accuracy over many previous specialised tools: far greater accuracy on protein-ligand interactions than state of the art docking tools, much higher accuracy on protein-nucleic acid interactions than nucleic-acid-specific predictors, and significantly higher antibody-antigen prediction accuracy than AlphaFold-Multimer v2.3 7,8 . Together these results show that high accuracy modelling across biomolecular space is possible within a single unified deep learning framework.

multidisciplinary sciences

What problem does this paper attempt to address?

The main goal of this paper is to introduce the AlphaFold 3 (AF3) model, which can accurately predict the structures of various biomolecular complexes, including proteins, nucleic acids, small molecules, ions, and modified residues. Specifically, the AF3 model achieves significant improvements over previous tools in the following areas:

1. **Protein-ligand interactions**: AF3's accuracy in protein-ligand interactions is far superior to traditional docking tools.
2. **Protein-nucleic acid interactions**: AF3's accuracy in protein-nucleic acid interactions is also significantly better than predictors specifically designed for nucleic acids.
3. **Antibody-antigen prediction**: AF3's accuracy in predicting antibody-antigen structures surpasses that of AlphaFold Multimer v2.

Through these improvements, the paper demonstrates the possibility of achieving high-precision biomolecular spatial modeling within a unified deep learning framework. Additionally, AF3 can handle various types of covalent modifications and performs better than AlphaFold Multimer v2 in protein complex prediction.