Identifying Different Types of Biclustering Patterns Using a Correlation-Based Dilated Biclusters Algorithm

Abstract

An essential step in the analysis of gene expression profiles is the identification of sets of co-regulated genes or genes tend to be active under only subsets of experimental conditions or participate in multiple cellular processes or functions. Biclustering is a non-supervised technique exceeds the traditional clustering techniques because it can find groups of both genes and conditions simultaneously. In this paper, we proposed a biclustering algorithm called Correlation-Based Dilated Biclusters CBDB to find sets of biclusters with correlated gene expression patterns. This algorithm has many phases starting with the preprocessing phase, determination of elementary biclusters, then the dilation phase depending on a heuristic searching approach with Pearson correlation coefficient as a measure of coherency, after that, the removal phase to exclude sets of genes and conditions that show low level of coherency, finally, the elimination of duplicated and overlapped biclusters phase. This approach showed reasonable results on both synthetic and real datasets compared with other correlation-based biclustering techniques.