Effective Energy Methods for Global Optimization for Biopolymer StructurePrediction

摘要

We developed the fundamental theory and algorithms of the new Packet AnnealingMethod for analyzing biopolymers 3-dimensional structures and tested it on small systems. We showed that the method provides a natural and powerful computational approximation to the stochastic description of biopolymer motions and encompasses other competing potential smoothing methods as special cases. Its main strength is that it uncovers and exploits the intrinsic hidden structures of biopolymer energy landscapes to efficiently perform global minimization using a hierarchical search procedure which concentrates parallel computing effort on a sequence of selected regions of decreasing size. Each search region corresponds to a metastable macrostate of the system, a region of conformation space that is isolated from the remainder of the space by effective energy barriers. The effective energy includes both energetic contributions from the energy potential function and entropic contributions resulting from thermal fluctuations of the biopolymer. It determines the thermodynamic macrostate free-energies which (rather than the energies) determine biopolymer structures. In addition, new methods for computing macromolecular conformational transitions and for molecular dynamics simulation were developed.