Optimization of parameters in macromolecular potential energy functions by conformational space annealing

A general protocol for refining the parameters of macromolecular potential energy functions by optimizing criteria that compare nativelike and normative conformations of one or more benchmark protein(s) is described. The protocol exploits the high efficiency of conformational space annealing (CSA) in finding the lowest-energy conformation of an isolated macromolecule. A novel form of the CSA method, local CSA, is introduced to provide better sampling of nativelike conformations. The computational expense of the protocol is reduced significantly by a linear approximation that estimates the energy of the (reminimized) native and normative conformations after every change of the force field parameters. The protocol is illustrated by optimizing the parameters of two force fields used in the CASP3 and CASP4 experiments, respectively. Another version of this general protocol (with different optimization criteria and optimization methods) was used to determine the parameters for the alpha, beta and alpha/beta force fields used in the CASP4 experiment, as reported in a companion publication.