There is an increasing demand for computing the relevant structures, equilibria, and long-timescale kinetics of biomolecular processes, such as protein-drug binding, from high-throughput molecular dynamics simulations. Current methods employ transformation of simulated coordinates into structural features, dimension reduction, clustering the dimension-reduced data, and estimation of a Markov state model or related model of the interconversion rates between molecular structures. This handcrafted approach demands a substantial amount of modeling expertise, as poor decisions at any step will lead to large modeling errors. In this talk, a variational approach for Markov processes (VAMP) will be introduced, which allows us to find optimal feature mappings and optimal Markovian models of the dynamics from given time series data. This leads to a family of deep learning methods for optimal end-to-end encoding of simulation trajectories for analysis of molecular kinetics.
Hao Wu's research focuses on operator theory of dynamical systems, machine learning approaches to time series analysis, and their applications in biophysics. He received Ph.D. (2007) and B.Eng. (2002) in computer science at Tsinghua University, and worked as a Postdoc from 2007 to 2018 at Institute of Mathematics, Freie Universität Berlin. Wu was also ex-Head of Research Group at Zuse Institute Berlin. He is now a Professor in Computational Mathematics at Tongji University.
Seminar Series by the NYU-ECNU Center for Computational Chemistry at NYU Shanghai