Detection of both positive and negative correlated rows in biclusters using Squared Transposed Virtual Error

Abstract

Biological Laboratories produce huge amounts of data every day. Biologists without proper processing tools and software are not able to analyze and discover hidden knowledge of these huge volumes of data. Biclustering technique is one of the bioinformatics approaches which is used to analysis obtained data from microarrays. Each microarray represents a data matrix of real numbers and biclustering algorithms are used to extract some sub-matrices including some specific patterns. HEvo-Bexpa is an evolutionary biclustering algorithm which can find biclusters including shift, scale and shift-scale patterns using Transposed Virtual Error (VET). VET is equal to zero for biclusters which containing positive correlated rows but it is not responsible for both positive and negative correlated rows at the same time. In this study, VET is extended to Squared Transposed Virtual Error (SVET). Obtained results demonstrate that it is possible to find rows with positive and negative scales using SVET.