How much physics can we learn from transition path sampling trajectories?
Seminar Computational Chemistry
Event details from: How much physics can we learn from transition path sampling trajectories?
- Date
- 20 May 2025
- Time
- 11:00
Venue
Room G5.29
Speaker
Dr. Fabio Pietrucci, (Sorbonne Université)
Abstract
Molecular dynamics simulations coupled to transition path sampling techniques provide detailed reactive mechanisms of rare events. The simplest among the latter approaches generate hundreds of unbiased trajectories relaxing from a set of transition-state configurations towards free-energy minima. Albeit insightful, such trajectories are not usually regarded, as far as I know, as sufficient to obtain the thermodynamics and kinetics of the process without more sophisticated algorithms like transition interface sampling. I will argue that the contrary is (often) true: a fascinating amount of physical information seems to be “packed” inside simple relaxation trajectories.
I will discuss how, and under which conditions, it is indeed possible to extract barrier heights and kinetic rates, starting from about 100 trajectories relaxing from a barrier top. The tools we developed are based on likelihood optimization of Langevin models describing the dynamics of the system projected on a collective variable. In a nutshell, it appears crucial, for the strategy to be successful, to assess whether the projected dynamics is of the “right” Markovian kind, and whether the entropic contribution to the average force is thoroughly sampled.
Our findings indicate that the reaction coordinate can be automatically optimized without the need to perform committor analysis: we are investigating different strategies where the loss function for the optimization is physically motivated and easily accessible. Intriguingly, committor-like coordinates can be obtained – at least in some cases we investigated – directly from the initial set of relaxation paths, based, e.g., on predicted rates or on calorimetry. I will present results on the estimation of protein-protein dissociation rates from one-directional association trajectories, for the barnase-barstar complex. I will also show how rather different processes, like the nucleation of crystals from liquids or of dislocations from crystals under stress, can be tackled with the same techniques.
L. Mouaffac, K. Palacio-Rodriguez, F. Pietrucci, J. Chem. Theory Comput. 19, 5701 (2023)
D.D. Girardier, H. Vroylandt, S. Bonella, F. Pietrucci, J. Chem. Phys. 159, 164111 (2023)
M. Perrone, D.D. Girardier, G.M. Pavan, F. Pietrucci, arXiv:2504.20211 (2025)