Predicting the non-compact conformation of amino acid sequence by particle swarm optimization

Abstract

Hydrophobic-hydrophilic (HP) model serves as a surrogate for the protein structure prediction problem to fold a chain of amino acids into a 2D square lattice. By the fact that the number of amino acids is equal to the number of lattice points or not, there are two types of folding conformations, i.e., the compact and non-compact conformations. Non-compact conformation tries to fold the amino acids sequence into a relatively larger square lattice, which is more biologically realistic and significant than the compact conformation. Here, we propose a heuristic algorithm to predict the non-compact conformations in 2D HP model. First, the protein structure prediction problem is abstracted to match amino acids to lattice points. The problem is then formulated as an integer programming model and we transform the biological problem into an optimization problem. Classical particle swarm optimization algorithm is extended by the single point adjustment strategy to solve this problem. Compared with existing self-organizing map algorithm, our method is more effective in several benchmark examples.