Study Of Particle Swarm Optimization For Protein Structure Prediction

Protein engineering is the frontier of the development of modern biotechnology, solving protein structure prediction problem is not only one of the most important issues for protein engineering on the post-genome era, but also one of the most challenging problems in the field of bioinformatics.Ab initio prediction method of protein structure prediction with thermodynamic hypothesis "the natural conformation of protein is the conformation with the lowest free energy " as the theoretical basis, by calculating the lowest energy value of protein can predict protein structure. Therefore, using ab initio prediction method to predict protein structure can be reduced to a global optimization problem.AB (Toy) model is one of the classical models of protein structure prediction. It is a kind of continuous polymer models, it has more realistic of protein structure than the other models and the practical significance is greater. Based on AB (Toy) model of protein structure prediction problem is a typical NP problem, finding an effective global optimization algorithm is the key to solve the problem.Currently, there are many heuristic algorithms applied to the two-dimensional AB model to predict protein structure, but the prediction accuracy of these methods is not high enough and the calculated speed is not fast enough. The particle swarm optimization (PSO) algorithm for protein structure prediction is a new application for PSO algorithm in recent years.Based on the two-dimensional AB model, this paper studied the protein structure prediction problem using the global version and local version of the PSO algorithm. For the local version of the PSO algorithm, this paper studied the five-neighborhood of the ring structure, nine-neighborhood of the ring structure,five-neighborhood of the von ? Neumann structure and nine-neighborhood of the von ? Neumann structure respectively. Simulation experiments were carried four Fibonacci protein sequences.In order to deal with the shortage of PSO algorithm in tackling optimization problem with multiple variable, this paper present a PSO framework with iterative improvement strategy which can deal with the interfere phenomena among different variable and improve the intensification ability of PSO algorithm. The experiments, which were carried on benchmark function optimization problems, showed that the iterative improvement strategy can improve the performance of PSO algorithm remarkably. We use this PSO algorithm framework for protein structure prediction problem. Simulation experiments were carried four Fibonacci protein sequences and two real protein sequences. Simulation results show that PSO algorithm with iterative improvement strategy has better performance than classical PSO algorithm.