Machine Learning in Multiscale Modelling of Molecular Systems

BEHROOZ HASHEMIAN; DANIEL MILLÁN; MARINO ARROYO

Manchester

Conferencia; Multiscale Modelling of Condensed Phase and Biological Systems; 2014

Collaborative Computational Project for Biomolecular Simulation

Multiscale modelling of complex molecular systems, such as proteins, requires to bridge the gap between femtosecond time scale of molecular vibrations and millisecond-and-up time scale of thermodynamic observables. Due to this huge gap, molecular dynamics (MD) simulations experience insufficient sampling, which hamper their connection with experiments. Enhanced sampling methods are thus in great demand to overcome this challenge; however, the effectiveness of such methods relies on having a good set of collective variables (CVs), which govern the essential dynamics of the system. Identifying such a coarse model is far from obvious for complex systems. We present here a general method, smooth and nonlinear data-driven collective variables (SandCV), based on machine learning techniques to identify such CVs from available computational or experimental ensembles, and integrate them in enhanced sampling methods. SandCV is a versatile method and can be non-intrusively combined with the available molecular dynamics implementations.