Integrative Clustering of High-Dimensional Data Using Non-Negative Matrix Factorization

• Activity Number: 61
• Type: Topic Contributed
• Date/Time: Sunday, August 9, 2015 : 4:00 PM to 5:50 PM
• Sponsor: Biometrics Section
• Author(s): Prabhakar Chalise* and Brooke Fridley
• Companies: University of Kansas Medical Center and University of Kansas Medical Center
• Keywords: Integrative Clustering ; NMF ; survival ; molecular

Abstract

Non-negative Matrix Factorization (NMF) method has been widely used in classifying disease subtypes. Most of the available NMF methods are developed for the application of a single data type at a time. Integrative clustering is then carried out by manual integration of the results obtained from individual analysis on several data types which is tedious and highly subjective in the assignment of cluster membership. We propose a novel approach for integrative clustering of multiple high dimensional molecular data in a single comprehensive analysis in order to utilize the strengths across multiple genomic assays on the same tumor samples. The novel method was applied to the cancer genome atlas (TCGA) studies on ovarian cancer that have both gene expression and methylation assays on tumor available on 499 subjects. The integrative method using 500 most variable features from each data resulted in three optimum clusters of samples. Significant survival differences were found among the identified clusters using Kaplan Meier and log rank test. Further studies are needed to determine the genes associated with these clusters and their functional relevance in the ovarian cancer etiology.