Predicting Protein Structure By Modified Particle Swarm Optimization Algorithm

In1980’s, bioinformatics was regard as a new interdisciplinary field and widely studied byscientists. As the important undertaker of the life activities, protein plays a key role in almost allbiological process, and the biological functions of protein are determined by their dimensional foldingstructures, so protein structure prediction remains one of the most challenging problems in molecularbiology research. In order to solve the problem two aspects should be resolved: establishing theeffective simplified model and employing the reasonable optimistic algorithm. In recent years,scientists have focused on introducing simplified models based on thermodynamics hypothesis thatthe native structure of protein is the state in which the energy of the whole system is lowest. HP latticemodel is one of the most typical simplified models. However, protein structure prediction problembased on HP lattice model is a typical NP problem.This paper presents the mathematical energy function, establishes the integer optimization model,so protein structure prediction problem is transformed into a combinatorial optimization problem. Atthe same time, the properties of the combinatorial optimization model are given in this paper.Currently, a majority of various optimization approaches have been applied to predict proteinstructure. Particle swarm optimization algorithm is a new group intelligent algorithm. The advantagesof this algorithm are simple and easy to achieve, profoundly intelligent background. In theoptimization field, it is suitable for continuous optimization problem, but protein structure predictionbased on2D HP lattice model is a discrete problem. Therefore, the concepts of adjustment operatorand exchange operator are introduced to reconstruct the two particle swarm optimization algorithm.They are used to solve the compact grid lattice, diamond lattice and triangular lattice as well as thenoncompact grid lattice, diamond lattice and triangular lattice based on2D HP lattice modelrespectively. At the same time, this paper also gives comparisons between compact grid lattice andcompact diamond lattice, compact diamond lattice and compact triangular lattice, noncompact gridlattice and compact grid lattice. The numerical simulation results indicate that the improved particleswarm algorithm for solving protein structure prediction problem based on HP lattice model isfeasible and effective, which is a valid attemption for protein structure prediction problem.