Solving Lattice Protein Folding Problems by Discrete Particle Swarm Optimization

Abstract

Using computer programs to predict protein structures from a mass of protein sequences is promising for discovering the relationship between the protein construction and their functions. In the area of computational protein structure analysis, the hydrophobicpolar (HP) model is one of the most commonly applied models. The protein folding problem based on HP model has been shown as NP-hard, to handle such an NP-hard problem, this paper proposes a discrete particle swarm optimization algorithm (DPSO HP) to solve various 2D and 3D HP lattice models-based protein folding problems. The discrete particle swarm optimization method used in DPSO HP is based on the set concept and the possibility theory from a set-based PSO (S-PSO). A selection strategy incorporating heuristic information and possibilities is adopted in DPSO HP. A particle's positions in the algorithm are defined as a set of elements and the velocities of a particle are defined as a set of elements associated with possibilities. The experimental results on a series of 2D and 3D protein sequences show that DPSO HP is promising and performs better than various competitive state-of-the-art evolutionary algorithms.