Exhaustive Search Method of Gene Expression Modules and Its Application to Human Tissue Data
Yoshifumi Okada, Kosaku Okubo, Paul B. Horton, Wataru Fujibuchi · 2007
Abstract — Recently, several biclustering methods have been suggested to discover modules in gene expression data matrices. A module, namely a bicluster, is defined as a subset of genes that exhibit a highly correlated expression pattern over a subset of conditions. Most existing methods produce sub-optimal solutions by approximation approaches since biclustering requires combinational searches for pairs of genes and conditions in a large search space. In this paper, we propose a fast biclustering method, BiModule, that exhaustively searches modules in real time based on a closed itemset mining algorithm. We show that BiModule can discover functionally-enriched biclusters better than the approximation approaches, while maintaining a comparably fast running time. In addition, we apply BiModule to a gene expression data matrix obtained from various human tissues/cells and demonstrate that genes found in each bicluster well reflect the functions and morphology of specific tissues/cells. Index Terms—biclustering, closed itemset, gene expression module, LCM. I.