Applied Research On Gene Clustering Through Particle-Pair Optimization And Harmony Search Algorithm

With the in-depth research of life science, more and more scholars focus on exploring the area of the origin of life and species. Nowadays, in the wake of the work on sequencing of the multiple biological genomes, the DNA chip technology has been recognized commonly by the field of scientific research. To confronting with the development of intricate data analysis and computation, bioinformatics was emerged as the interdisciplinary field of science, integrating biology, mathematics, chemistry and computer science, etc. It indicates that HGP (Human Genomes Project) has entered the post-genomes era, and the key point of the genomes research translates from the genomic sequencing into determining the function of organisms possessing the gene in live process and the relationship of intercoordination and mutual guiding in different genes. With such great transition, the traditional experimental technique and method can no longer meet the requirement of dealing with mass data of genomic sequence. On account of rapid development of AI (Artificial Intelligent) and life science, people have urgent need to seek more efficient and convenient processing method of genetic expression. Therefore, if we wish to reveal the ultimate source of life, we should conduct the sophisticated sequence and statistic analysis, which came into being the microarray technology.Microarray technology, commonly known as chip technology, is widely used in sequence analysis, detection of gene mutation, polymorphic analysis, genetic disease diagnosis, and the like. Clustering analysis is one of the most common methods in microarray data analysis, owing the significant academic research value. Though previous clustering algorithm settles the difficult issue of lower dimension and less data quantity smoothly, the existing algorithm may not require the ideal clustering results with the expansion of DBscale and level of interaction constantly, and the extension of investigative analysis. In addition, the existing one is not easy to meet the requirement of efficiency and veracity in data analysis for human beings. Due to the emerging demand of efficiency of algorithm, the rise of swarm intelligence algorithm, such as GA (Genetic Algorithm), PSO (Particle Swarm Optimization), PPO (Particle-pair Optimization), immune algorithms, artificial bee colony algorithm and so on, offer a new direction for data analysis. Swarm intelligence algorithm has characteristic of simulating the evolution learning in biotic community, demonstrating the unique advantage of solving the intricate optimization problem. Hence, it is utilized in many fields successfully, for instance, social science, natural science, managerial economic, medical science, biology, and computer science, etc.The research on clustering algorithm of gene expression data is in exploratory stage. Currently, as the innovative genetic clustering algorithm, PPO is one of the widely used genetic algorithms with better effect, which has great advantage of small population size, being convenient for coordinating the position relationship between particles and owing capacity for achieving the good cluster result. Meanwhile, PPO also has deficiency of sinking into local optimum untimely and global searching ability. Aiming to solve such insufficient, the following innovative work presents a new algorithm mixed the PPO optimization algorithm with harmony search algorithm, which named as DPPO-HS. In the iterative process of generating the elite particles for the first time, One of the elite particles are initialized once by introducing the PSO rapidly. Simultaneously, another elite particle are do the same through the standard PPO clustering algorithm. By combining two different initialized particles into an elite particle pair, it will enhanced the ability of exchange and learning between them. To a certain extent, the quality of the solution and the search ability of elite particles are improved. After obtaining the elite particle pair via hybrid algorithm, a harmony optimization algorithm based on information entropy to adjust the fine tuning probability of HS algorithm is introduced in the second stage of the iterative process to solve the problem where PPO is easy to fall into the local optimal, which improves the global search ability and the ability to jump out of the local optimum region. On the basis of this idea, through four sets of different international standard gene expression data sets and taking the MSE mean square function, the compactness within the D1 class, and the separation degree between the D2 classes as the detection index, this paper make a detailed comparative analysis of the experiment among DPPO-HS and the current popular K-Means algorithm, standard PPO particles on the clustering algorithm, DPPO algorithm, as well as better clustering results of the PPO-DE hybrid algorithm. Five kinds of algorithms are compared in detail by the three test indexes of four sets of different data sets.The experimental results indicates that the fusion algorithm in this paper, compared with the other four algorithms, has achieved a better clustering accuracy, within class DPPO-HS and class separation. It is proved that the fusion algorithm proposed in this paper is effective.