Biological function unfolds over timescales far beyond the reach of brute-force molecular dynamics. Rather than only applying existing methods, we develop algorithms that make rare events, kinetics and high-dimensional landscapes tractable — and release them for the community.
WeTICA runs a directed weighted-ensemble search within a machine-learned (TICA) reduced space to estimate rare-event kinetics with high statistical precision (J. Chem. Phys., 2025), and CoWERA introduces a temporal-coherence-guided, binless resampling scheme for robust weighted-ensemble rate estimates (J. Chem. Phys., 2026).
PathGennie uses direction-guided adaptive sampling from ultrashort ‘monitored’ trajectories to rapidly generate rare-event pathways (J. Chem. Theory Comput., 2025). IceCoder applies a variational autoencoder to automatically identify ice/crystal phases during simulation (J. Chem. Theory Comput., 2025), and we have systematically benchmarked dimensionality-reduction and clustering techniques for protein folding on the Trp-Cage mini-protein (Biophys. Chem., 2025).
PPIscout combines mixed amino-acid–water molecular dynamics with structural analysis to map protein–protein interaction hotspots (J. Chem. Sci., 2025) — the engine behind several of our cryptic-pocket discoveries. We also developed machine-learning surrogates such as support-vector-regression Monte Carlo for flexible water clusters (ACS Omega, 2020).