Dramatic performance enhancements for the FASTER optimization algorithm

Abstract

FASTER is a combinatorial optimization algorithm useful for finding low-energy side-chain configurations in side-chain placement and protein design calculations. We present two simple enhancements to FASTER that together improve the computational efficiency of these calculations by as much as two orders of magnitude with no loss of accuracy. Our results highlight the importance of choosing appropriate initial configurations, and show that efficiency can be improved by stringently limiting the number of positions that are allowed to relax in response to a perturbation. The changes we describe improve the quality of solutions found for large-scale designs, and allow them to be found in hours rather than days. The improved FASTER algorithm finds low-energy solutions more efficiently than common optimization schemes based on the dead-end elimination theorem and Monte Carlo. These advances have prompted investigations into new methods for force field parameterization and multiple state design.

Additional details