Protein structure prediction with evolutionary algorithm.

Abstract

Three-dimensional protein structure prediction is one of the fundamental problems of Structural Bioinformatics. The use of problem information through fragment insertion, secondary structure, and contact maps can help explore the search space better. An evolutionary algorithm is proposed in this work, which uses this problem information for protein structure prediction. In the proposed method, a dynamic speciation technique and fragment insertion are used to promote the diversity of the population. The fragment library is generated based on the Rosetta Quota protocol to provide fragments with increased diversity. The information from contact maps and secondary structure are used in two selection strategies to better explore the conformational search space. The results of an experiment with nine proteins are presented. Results obtained are competitive with the literature and are compared in terms of RMSD, GDT, and processing time.