Increasing the Efficiency of Ensemble Molecular Dynamics Simulations with Termination of Unproductive Trajectories Identified at Runtime
On the occasion of the PASC Conference (Platform for Advanced Scientific Computing), Professor Michel Cuendet just presented an abstract on work that was also published in The Journal of Physical Chemistry A special issue “Charles L. Brooks III Festschrift”.
Written in collaboration with Jack Marquez, Silvina Caino-Lores, Trilce Estrada, Ewa Deelman, Harel Weinstein and Michela Taufer, the article is entitled:
Increasing the Efficiency of Ensemble Molecular Dynamics Simulations with Termination of Unproductive Trajectories Identified at Runtime
The application of molecular dynamics (MD) simulations to study increasingly large and complex biomolecular systems is challenged by the required amounts of trajectory data needed to sample their conformational space appropriately. These massive data sets need to be stored and then fed to the various algorithms to reveal the dynamic behaviors of the systems and the underlying energetics underlying functional mechanisms. To address this challenge, the research team developed a software framework that can increase the efficiency of MD pipelines by performing key analysis tasks on-the-fly, as the simulation proceeds. This article presents one component of this framework that can reduce the computational effort while maintaining the structural properties and relative stability ranking of the different molecular conformational states identified in the simulation. This framework component utilizes early termination of individual trajectories identified as unproductive in the sampling of conformational space. The criteria for termination are derived from time series of secondary quantities calculated from atomic coordinates and are computed in situ during trajectory generation. As a proof of concept, the approach is applied to simulations of the FS peptide. Comparisons between the free energy surfaces calculated from ensembles of complete, unabridged simulations and those obtained from ensembles with increasingly stringent stopping criteria show that ~30% of simulation time can be saved while conserving the correct ranking of peptide metastable states. Our in situ early termination approach is an important step towards more sustainable high-performance computing.
17 Jun 2025