Particle Swarm Optimization for Triclustering High Dimensional Microarray Gene Expression Data

The study of high dimensional microarray gene expression data represents the large computational challenge due to its huge volume of the data. Many clustering techniques are applied to extract the coexpressed genes over the samples. Biclustering improved the traditional clustering by grouping the genes that similarly expressed over only a subset of samples. However, to cluster the high dimensional data with three dimensions such as genes, samples and time points, Triclustering technique is employed for grouping the coexpressed genes over a subset of samples under a subset of time points which imposes huge computational burden. In this paper, Particle Swarm Optimization technique is applied to extract the triclusters from the high dimensional data with objective function as Mean Square Residue. The algorithm is applied to three real life microarray gene expression data and the performance of the work is analyzed using the objective function. The biological significances of the extracted triclusters from all the three datasets are also analyzed. The biological significance analysis are also compared with other triclustering algorithms and the proposed work outperforms the other algorithms.