RABBIC: Rank-Based BIClustering Algorithm

Abstract

Biclustering performs simultaneous clustering on the row and column dimensions of the data matrix, it could discover data modules in the data matrix. Gene module is an important concept in systems biology. In this paper, gene modules are specifically defined as a set of genes whose expression levels share the same linear order on each member of a subset of samples. In order to discover such modules, a novel algorithm, the Rank-Based BIClustering algorithm (RABBIC), is designed and developed. RABBIC, when applied to the real ovarian cancer gene expression data, identifies 93 modules, and 25 are biologically significant according to the gene set functional enrichment analysis. This paper deals with the gene expression data from the aspect of rank, which is helpful in reducing the noise of the data. It provides new thoughts for the researches of gene module identification.