Novel Memetic Algorithm for protein structure prediction

Abstract

A novel Memetic Algorithm (MA) is proposed for investigating the complex ab initio protein structure prediction problem. The proposed MA has a new fitness function incorporating domain knowledge in the form of two new measures (H-compliance and P-compliance) to indicate hydrophobic and hydrophilic nature of a residue. It also includes two novel techniques for dynamically preserving best fit schema and for providing a guided search. The algorithm performance is investigated with the aid of commonly studied 2D lattice hydrophobic polar (HP) model for the benchmark as well as non-benchmark sequences. Comparative studies with other search algorithms reveal superior performance of the proposed technique. © Springer-Verlag Berlin Heidelberg 2009.