Strings Attached: A Data-Driven Framework for Committor-Consistent Pathways

Please join Yale Chemistry for a Silliman Seminar in Theoretical Chemistry with Chris Chipot, Research Director at the Centre National de la Recherche Scientifique (CNRS) at University of Lorraine, and Adjunct Professor of Physics at University of Illinois Urbana-Champaign.

Summary: Identifying physically meaningful, committor-consistent transition pathways remains a central challenge in the study of rare events in complex systems. Here, we present a neural-network-based framework that, for the first time, simultaneously learns the committor function and its associated committor-consistent string, providing a unified and insightful depiction of transition processes. Rooted in the committor time-correlation function, this approach transcends traditional methods based on infinitesimal time-lag approximations, enabling accurate application across a wide range of dynamical regimes. It effectively captures and discriminates between multiple competing pathways—an essential feature for unraveling the complexity of biomolecular transformations. Validated on both benchmark potentials and biological systems such as peptide isomerization and protein folding models, the method reliably recovers established mechanisms, rate constants, and dynamic behavior. Its adaptability to various collective variables and robustness across neural-network architectures render it a powerful and versatile tool for enhanced-sampling simulations of rare events, offering new perspectives on the intricate energy landscapes of biomolecular systems.

For more information on Professor Chipot's research: https://www.lia-uiuc.cnrs.fr/index.php/research.

Faculty Hosts: Professor Tianyu Zhu and Professor Bill Jorgensen.

This seminar is generously sponsored by the Mrs. Hepsa Ely Silliman Memorial Fund.

Location: Sterling Chemistry Lab (SCL), Room 111