A Bayesian neural network and PSO clustering applied to gene expression data

Abstract— over the last decade many researchers are studying in gene expression data clustering. Gene expression data is the process which extracted useful information from genes. Analysis and clustering of this data such as microarray technology is very complex task. Microarray technology measured expression level of thousands genes at the same time, Therefore knowing the gene expression levels of same sample described molecular scenario and it helps to cells and tissues. Microarray technology plays an important role in detection of diseases and lead to significant progress in drug discovery and clinical diagnostics which these technologies by using gene expression profiling and classify samples based on expression patterns is able to respond to the many genetic questions. In this paper a new method based on particle swarm optimization and Bayesian neural network is proposed to find subscription clusters. So the position of each particle will be showed in binary, therefore ‘0’ indicates that the corresponding gene has not been selected particle and ‘1’ indicates selected gene. Thus, in order to evaluate gene selected by a particle we used Pearson correlation coefficient. Next, density concept is used for improve nearest neighbor of clusters and measure the number of genes in a cluster. Experimental results on CSH press database show that the proposed method is accurate than DBscan to classifying similarity gene.

Keywords:

clustering , data expression , Bayesian neural network , PSO