Nonnegative Matrix Factorization: a New Paradigm for Large-Scale Biological Data Analysis

• Activity Number: 111
• Type: Contributed
• Date/Time: Monday, August 7, 2006 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): Karthik Devarajan*+
• Companies: Fox Chase Cancer Center
• Keywords: nonnegative matrix factorization ; molecular pattern discovery ; high-throughput biology ; class prediction and class comparison ; gene expression ; microarray

Abstract

Nonnegative matrix factorization (NMF) was introduced as an unsupervised, parts-based learning paradigm by Lee and Seung (Nature, 1999). It involves the decomposition of a nonnegative matrix V into two nonnegative matrices W and H, via a multiplicative updates algorithm. In the context of a gene expression matrix V ~ WH, each column of W defines a metagene and each column of H represents the metagene expression pattern of the corresponding sample. NMF has been primarily applied in an unsupervised setting in facial pattern recognition and text mining; and more recently for molecular pattern discovery in high-throughput biological studies. In this paper, we discuss and review the potential applicability of NMF in a supervised learning framework, and as an exploratory tool for dimensionality reduction. We illustrate our methods with several examples using cancer microarray data.