Evolutionary Metaheuristic for Biclustering based on Linear Correlations among Genes

Abstract

A new measure to evaluate the quality of a bicluster is proposed in this paper. This measure is based on correlations among genes. Moreover, a new evolutionary metaheuristic based on Scatter Search, which uses this measure as the fitness function, is presented to obtain biclusters that contain groups de highly-correlated genes. Later, an analysis of the correlation matrix of these biclusters is made to select these groups of genes that define new biclusters with shifting and scaling patterns. Experimental results from human B cell lymphoma are presented.