基于Toy模型蛋白质折叠预测的多种群微粒群优化算法研究

Study of Multi-PSO Algorithm for Protein Folding Prediction Problem of Toy Model

基于Toy模型的蛋白质折叠结构预测问题是一个典型的NP问题。提出了多种群微粒群优化算法用于计算蛋白质能量最小值。该算法采用了一种新的算法结构,在该结构中,每一代的种群被分为精英子种群、开采子种群和勘探子种群三部分,通过改善种群的局部开采能力和全局勘探能力来提高算法的性能。分别采用Fibonacci蛋白质测试序列和真实蛋白质序列进行了折叠结构预测的仿真实验。实验结果表明该算法能够更精确地进行蛋白质折叠结构预测,为生物科学研究提供了一条有效途径。

Protein folding prediction problem with Toy model is a classical NP problem. A multi particle swarm optimization （MPSO） is proposed and applied successfully to protein folding prediction. MPSO introduces a new architecture that is characterized by balancing exploitation capability and exploration capability of particle swarm optimization （PSO）. In the architecture,the population in each generation consists of three parts：an elitist part,an exploitative part, and an explorative part. With enhance of the global search and local search ability, MPSO can be effectively used for protein folding prediction. The algorithm has been tested in the two-dimensional Toy model for several Fibonacci protein sequences and real protein sequences. The ground state energies predicted are lower than those reported in the literatures and show that MPSO is correct and effective.