Adaptive Methods for Free Energy Computation and Coarse-Graining Strategies

Wednesday, July 25, 2007 - 2:00pm - 3:00pm
EE/CS 3-180
Eric Darve (Stanford University)
Free energy computation is one of the main goals of
molecular dynamics simulations. Numerically, these calculations are
made difficult by the presence of free energy barriers separating
meta-stable basins. Several methods have been developed to address
this issue such as the adaptive biasing force (ABF), in which an
external force is progressively adapted. Upon convergence, the
reaction coordinate along which the free energy is computed exhibits a
diffusive behavior in a flat free energy profile, leading to a rapid
convergence of the computed averages and a rapid reduction of
statistical errors. Such methods can serve as the starting point for
computing coarse grained models of large proteins in which the number
of degree of freedom is greatly reduced. We will show the role the
free energy plays in formulating these models. In this case as well,
ABF can be applied to yield faster convergence.