Composite Gene Module Discovery using Non-negative Independent Component Analysis

Abstract

Gene module discovery can provide comprehensive molecular portrait of biological regulation and functional genomics. We present a new analytic strategy - nonnegative independent component analysis to reveal some gene module composite. The results show that by grouping genes in the latent space, we can find statistically more significant enrichment of gene annotations within clusters. Further, this approach has been applied to a muscular dystrophy data set for gene module discovery