A novel SFLA based method for gene expression biclustering

Abstract

Form the time of its invention, microarray technology is continuously growing and has been taking major role in biological research. This technology generates huge amount of gene expression data for biological analysis. Parallel computation methods are required to find functional associations from this large amount of biological data. An unsupervised machine learning technique, clustering algorithm groups similar genes based on entire conditions. But normal clustering methods cannot find different cellular processes from gene expression data because a biological activity can start functioning in the presence of some specific conditions. So, biclustering techniques are used instead of normal clustering. Biclustering basically identifies a set of genes that are co-expressed for some specific experimental conditions. Here we introduce an improved shuffled frog leaping algorithm(SFLA) based approach to find biclusters. SFLA is a hybrid of evolutionary memetic algorithm and collective intelligence based particle swarm optimization algorithm. Also It has faster convergence speed. By applying the proposed algorithm on yeast (Saccharomyces cerevisiae) cell cycle dataset, large number of biologically significant biclusters are obtained, which are verified by gene ontology database, compared to other existing algorithms. Also the biclusters have small MSR value and large size.