Chengwei Dong

I am a Ph.D. student in the Theoretical Chemistry Institute at University of Wisconsin–Madison, where I conduct research at the interface of molecular dynamics and machine learning under the supervision of Prof. Xuhui Huang. Prior to my doctoral studies, I received a B.S. in Chemistry from Peking University, where I worked on molecular docking methods in the group of Prof. Yi Qin Gao.

My research focuses on developing physics-informed AI methods for biochemical discovery, aiming to address the challenges that limit our understanding of complex molecular systems. In particular, I work on deep generative models that integrate statistical physics principles to enable molecular ensemble sampling. In parallel, I have several collaborations with experimental researchers in which I employ protein language models to facilitate protein design.

My work draws on methods and ideas from machine learning, theoretical chemistry, and bioinformatics, and I believe that such interdisciplinary integration can spark scientific breakthroughs that go beyond well-defined problems.

selected publications

JCIM

DSDPFlex: Flexible-Receptor Docking with GPU Acceleration

Chengwei Dong, Yu-Peng Huang, Xiaohan Lin, Hong Zhang^*, and Yi Qin Gao^*

Journal of Chemical Information and Modeling, 2024

Abs DOI HTML

Molecular docking is an essential tool in structure-based drug discovery, widely utilized to model ligand–protein interactions and enrich potential hits. Among the different docking strategies, semiflexible docking (rigid-receptor and flexible-ligand model) is the most popular, benefiting from its balance of docking accuracy and speed. However, this approach ignores the conformational changes of proteins and hence demands suitable protein conformations as input. When the binding interaction adheres to an induced-fit model, flexible methods such as molecular dynamics simulation can be utilized, but they are computationally demanding. To balance between speed and accuracy, the flexible docking approach is an effective choice, as exemplified by AutoDock Vina and AutoDockFR, which treat selected protein side chains as flexible parts. However, the efficiency of flexible docking methods is yet to be improved for virtual screening usage. In this article, we introduce DSDPFlex, an improved flexible-receptor docking method accelerated by GPU parallelization. Beyond acceleration, optimizations with respect to sampling, scoring, and search space are implemented in DSDPFlex to further improve its capability in flexible tasks. In cross-docking evaluation, DSDPFlex demonstrates superior accuracy compared to AutoDock Vina and is 100 times faster than Vina in flexible-receptor tasks. We also show the advantage of flexible-receptor methods on suboptimal pockets and validate the advantage of DSDPFlex in screening on apo and AlphaFold2-predicted structures. With improvements in both efficiency and accuracy, DSDPFlex is expected to hold potential in future docking-based studies.