Nonlinear dimensionality reduction for accelerated molecular dynamics calculations

BEHROOZ HASHEMIAN; DANIEL MILLÁN; CARLOS BRAGA; MARINO ARROYO

Barcelona

Simposio; IRB Barcelona PhD Student Symposium; 2011

Institute for Research in Biomedicine

Free energy is the central thermodynamic function in biological systems as it provides the relevant thermodynamic forces that drive structural processes for biomacromolecules. Molecular Dynamics (MD) simulations allow the calculation of these properties in a potentially precise manner. Notwithstanding, extracting such thermodynamic information from a MD simulation is often a very challenging process described by the presence of a large configuration space with many metastable basins. A spate of methods has been proposed to overcome this difficulty either by enhancing the sampling of the phase space, or by focusing instead on the transition paths between specific conformations. A family of successful methods relies on a judicious identification of the reaction coordinates along which the chemical transformation takes place, and applies an appropriate bias along them to overcome the energy barriers. The unambiguous identification of the relevant reaction coordinates for a given chemical process is often not immediately apparent. We propose a method that automatically identifies low dimensional nonlinear reaction co-ordinates 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) reaction coordinates provide a mechanistic description of the conformational transitions, and are the basis of accelerated MD.