Speaker: Dr. Jiahui Chen, University of Arkansas | Department of Mathematical Sciences
Date: Wednesday, September 6, 2023, 4:00 – 5:00 PM
Location: Science Engineering Hall (SCEN 408)
Title: Mathematics-AI modeling for protein-protein interactions
Abstract: Artificial intelligence (AI) has emerged as a new paradigm for scientific discovery. However, AI modeling of biological data remains a challenge due to their intricate structural complexity, excessively high dimensionality, severe nonlinearity, and intrinsic multiscale. We devise differential geometry and algebraic topology to address these challenges. Specifically, we utilize persistent homology, a main workhorse in topological data analysis (TDA), to simplify biomolecular structure complexity and reduce their dimensionality. Since persistent homology is insensitive to homotopic shape evolution, we developed persistent Laplacians to capture non-topological shape changes in data by their non-harmonic spectra. For volumetric data, like molecular electron density of proteins, we proposed an evolutionary de Rham-Hodge method to extend the traditional Hodge Laplacian to a multiscale formulation. We introduced boundary-induced graph Laplacians to further reduce computational complexity. These new mathematical tools are paired with advanced machine learning algorithms, such as ensemble learning, manifold learning, graph neural networks, and transformers, to reveal the mechanisms of SARS-CoV-2 transmission and evolutions via infectivity strengthening and antibody resistance. We had successfully predicted the incoming dominance of Omicron BA.2 and BA.4/BA.5 variants.
Bio: Jiahui Chen received his Ph.D. degree in computational and applied mathematics from Southern Methodist University, where his research focused on the implementation of mathematical methods for biophysics, including the Poisson−Boltzmann equation, boundary element method, Treecode method, fast multipole method, and parallel computing. After graduation, he joined the Department of Mathematics at Michigan State University as a visiting assistant professor in Prof. Guo-Wei Wei’s group. He is now an assistant professor at the University of Arkansas. His current research focuses on topological and geometrical data analysis and machine learning algorithms with their modeling of and application to biomolecules. His model studies were applied to the prediction of mutation-induced binding free energy changes of the SARS-CoV-2 spike protein binding to ACE2 and antibodies.
Recent Comments