Multi-View Gene Clustering using Gene Ontology and Expression-based Similarities

Gene is a fundamental, physical and functional unit of hereditary. A proper grouping of genes is needed for a better understanding of the natural patterns amongst them. Most of the existing approaches utilize gene expression values for the clustering of genes. They mostly ignore the semantic similarity between genes obtained from the global database like Gene Ontology(GO). We present a multi-view multi-objective clustering approach where Euclidean distances between the gene expression values and GO-based multi-factored gene-gene semantic similarity are considered as two complementary views and a consensus partitioning is obtained that satisfies both the aspects or views. Two real-life gene expression datasets are used to demonstrate the effectiveness of the proposed multi-view clustering technique. We have compared our approach with various standard single-view, multi-view and multi-objective optimization based clustering approaches. The results show that better co-expressed and biologically significant gene partitions are obtained by our proposed approach. Also, to validate the significance of the obtained clusters statistically and biologically, various biological, statistical and visual significance tests were conducted, and the corresponding results are reported.