DynamicBind: predicting ligand-specific protein-ligand complex structure with a deep equivariant generative model

Wei Lu,Jixian Zhang,Weifeng Huang,Ziqiao Zhang,Xiangyu Jia,Zhenyu Wang,Leilei Shi,Chengtao Li,Peter G. Wolynes,Shuangjia Zheng

DOI: https://doi.org/10.1038/s41467-024-45461-2

IF: 16.6

2024-02-05

Nature Communications

Abstract:Abstract While significant advances have been made in predicting static protein structures, the inherent dynamics of proteins, modulated by ligands, are crucial for understanding protein function and facilitating drug discovery. Traditional docking methods, frequently used in studying protein-ligand interactions, typically treat proteins as rigid. While molecular dynamics simulations can propose appropriate protein conformations, they’re computationally demanding due to rare transitions between biologically relevant equilibrium states. In this study, we present DynamicBind, a deep learning method that employs equivariant geometric diffusion networks to construct a smooth energy landscape, promoting efficient transitions between different equilibrium states. DynamicBind accurately recovers ligand-specific conformations from unbound protein structures without the need for holo-structures or extensive sampling. Remarkably, it demonstrates state-of-the-art performance in docking and virtual screening benchmarks. Our experiments reveal that DynamicBind can accommodate a wide range of large protein conformational changes and identify cryptic pockets in unseen protein targets. As a result, DynamicBind shows potential in accelerating the development of small molecules for previously undruggable targets and expanding the horizons of computational drug discovery.

multidisciplinary sciences

What problem does this paper attempt to address?

This paper aims to address an important issue in protein-ligand interactions, which is how to predict the dynamic structure of protein-ligand complexes. Traditional docking methods usually treat proteins as rigid, while molecular dynamics simulations can propose appropriate protein conformations but require high computational demands and are difficult to capture rare transitions between biologically relevant equilibrium states. The paper proposes a deep learning method called DynamicBind, which utilizes equivariant geometric diffusion networks to construct a smooth energy landscape and facilitate efficient transitions between different equilibrium states. DynamicBind is able to accurately recover ligand-specific conformations from unbound protein structures without requiring full structures or extensive sampling. This approach demonstrates state-of-the-art performance in docking and virtual screening benchmarks, can handle a wide range of large-scale protein conformational changes, and identify cryptic pockets of previously unseen protein targets. Therefore, DynamicBind has the potential to accelerate small molecule development against previously considered undruggable targets and expand the scope of computational drug discovery.