All-atom protein folding with stochastic optimization methods

Abstract

The development of an all-atom free energy force field (PFF01) for protein structure prediction with stochastic optimization methods was made. It showed that PFF01 correctly predicts the native conformation of several proteins as the global optimum of the free energy surface. Recent folding studies, which permitted the reproducible all-atom folding (not CHn groups) of the 20 amino acid trp-cage protein, the 40 amino acid three-helix HIV accessory protein, and the 60 amino acid bacterial ribosomal protein L20 with a variety of stochastic optimization methods. These results showed that all-atom protein folding can be achieved with present day computational resources for the proteins of moderate size. Suitable optimization methods are required to speed up the simulation by avoiding high-energy transition states, adapt large-scale moves, or accept unphysical intermediates. The chapter focuses on the four different optimization methods: the stochastic tunneling method, the basin hopping technique, the parallel tempering method, and a recently employed evolutionary technique. A strong correlation is found between energy and RMSB deviation to the native structure for all simulations. Results indicate that the simple basin hopping method is best in the determination of the global optimum of the free energy surface of realistic all-atom protein models. It also discusses the disadvantages. In conclusion, protein structure prediction with stochastic optimization methods requires two separate key ingredients: an accurate force field and efficient optimization techniques. ？ 2006 Elsevier B.V. All rights reserved.