A Co-Clustering Technique for Gene Expression Data Using Bi-Partite Graph Approach

Mining microarray data sets is vital in bioinformatics research and medical applications. There has been extensive research on co-clustering of gene expression data generated using cDNA microarrays. Co-clustering approach is an important analysis tool in gene expression measurement, when some genes have multiple functions and experimental conditions are diverse. In this paper, we introduce a new framework for microarray gene expression data co-clustering. The basis of this framework is a bipartite graph representation of 2-dimensional gene expression data. We have constructed this bipartite graph by partitioning the sample set into two disjoint sets. The key property of this representation is that, for a gene×sample matrix, it constructs the range bipartite graph, a compact representation of all similar value ranges between sample columns. In order to produce the set of co-clusters, it searches for constrained maximal cliques in this bipartite graph. Our method is scalable to practical gene expression data and can find some interesting co-clusters in real microarray datasets that meet specific input conditions.