Using Informative Set of Genes for Biomarker Discovery Based on Gene Expression Microarray Data

• Activity Number: 383
• Type: Topic Contributed
• Date/Time: Tuesday, July 31, 2012 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Abdus Sattar*+ and Darius Dziuda
• Companies: Case Western Reserve University and Central Connecticut State University
• Keywords: Biomarker Discovery ; Multivariate Feature Selection ; Informative Set of Genes

Abstract

Understanding the association between disease and molecular signatures or genes is critical in the therapeutic drug development and hence the treatment of a disease. But finding a small set of genes (that would constitute a robust multivariate biomarker) from a microarray experiment with thousands of genes is a challenge. Due to high-dimensionality and sparseness of gene expression data, a small set of genes identified by a single run of any feature selection method has a high probability to be a chance results that is neither generalizable nor reflects underlying biological processes. Identification of the Informative Set of Genes (containing all of the information significant for the class differentiation), and then performing feature selection on its most frequent gene expression patterns allows for maximizing the chances that the resulting multivariate biomarker represents the most significant biological processes associated with the phenotypic classes we differentiate. We will apply these methods to breast cancer gene expression microarray data obtained from TCGA.