Enhancing Molecular Dynamics with Smooth and Nonlinear Data-driven Collective Variables

BEHROOZ HASHEMIAN; DANIEL MILLÁN; MARINO ARROYO

Lugano

Workshop; Long time dynamics from short time simulations; 2014

CECAM - Centre Européen de Calcul Atomique et Moléculaire

The time-scale disparity between molecular vibrations and conformational transitions in systems with metastability makes it very difficult to obtain accurate sampling from molecular dynamics simulations, which hinders the connection between simulations and experimental observations. To overcome this issue, a number of enhanced sampling methods have been proposed, such as Metadynamics, and Adaptive Biasing Force. However, the effectiveness of these methods relies on a good set of collective variables, which need to compactly characterize molecular conformations, capture metastability, and be differentiable. We present here a methodology based on nonlinear dimensionality techniques that, starting from pre-existing computational or experimental ensembles representative of the conformational variability, builds Smooth And Nonlinear Data-driven Collective Variables (SandCV) and integrate them in enhanced sampling methods. SandCV relies on an intrinsic manifold model for the molecular system. SandCV is minimally invasive with respect to the underlying molecular dynamics code. We show how this methodology can deal seamlessly with intrinsic manifolds of complex topology described with multiple parameterization patches.