Molecular pattern discovery using non-negative matrix factorization based on Renyi's information measure
Associate Member, Department of Biostatistics, Fox Chase Cancer Center, Philadelphia
Coauthors: Nader Ebrahimi, Department of Statistics, Northern Illinois University
Current efforts in cancer research focus on predicting drug responses of patients by analyzing gene expression patterns and identifying genes that cause specific responses, with a view to personalizing cancer treatment. DNA microarray technology has made it possible to simultaneously measure the expression levels of tens of thousands of genes. A related problem is that of class discovery, which involves unsupervised clustering of tissue samples or genes based on corresponding gene expression data and to identify the genes that distinguish the clusters. Non-negative matrix factorization (NMF) is a powerful method for decomposing the gene expression matrix V into two matrices with positive entries, V ~ WH, where each column of W defines a metagene and each column of H represents the metagene expression pattern of the corresponding sample (Brunet et al, Bioinformatics, 2004). We propose an unsupervised method for molecular pattern discovery using NMF, based on Renyi's measure of distance between two non-negative matrices, related to the Poisson likelihood. This is a generalization of previous work (Lee & Seung, Nature, 1999), and includes various well-known distance measures as special cases. We demonstrate the applicability of this method using cancer microarray data.
Date received: September 28, 2005
Copyright © 2005 by the author(s). The author(s) of this work and the organizers of the conference have granted their consent to include this abstract in Topology Atlas. Document # caqt-98.