Nonlinear Dimensionality Reduction Methods for Accelerated Molecular Dynamics

MARINO ARROYO; BEHROOZ HASHEMIAN; DANIEL MILLÁN; CARLOS BRAGA

Sao Paulo

Congreso; 10th World Congress on Computational Mechanics (WCCM 2012); 2012

International association for computational mechanics

Extracting thermodynamic information such as free energy from molecular dynamics simulations for complex biomacromolecules is very challenging, as the sampling is seriously impaired by a huge configuration space with many metastable basins. A myriad of method have been proposed to overcome this difficulty, for instance by enhancing the sampling and the phase space exploration, or by focusing on transition paths between specific conformations. A family of successful methods rely on the identification of reaction coordinates, and applying biases along them to overcome the energy barriers. We propose a method that automatically identifies low dimensional nonlinear collective variables on the basis of MD trajectories or other ensembles characterizing the large deviation flexibility of the molecule. These data-driven (as opposed to insight-driven) collective variables a mechanistic description of the conformational transitions, and are the basis of accelerated MD. We will illustrate the methodology with conformation ensembles obtained from experimental data, and with load-bearing proteins (titian) probed with steered molecular dynamics.