Statistical Consideration of Correlation Among Genes in Gene Set Analysis of Paired Case-Control Studies

• Activity Number: 39
• Type: Contributed
• Date/Time: Sunday, July 29, 2012 : 2:00 PM to 3:50 PM
• Sponsor: IMS
• Author(s): Martina Fu*+ and Wenjiang Fu
• Companies: and Michigan State University
• Keywords: Association test ; Correlation ; gene expression ; Gene set analysis

Abstract

Gene set analysis plays an important role in testing association between a biological pathway and a disease or phenotype of interest. Since the genes in a gene set (often a biological pathway) are co-regulated, they present high correlation and may yield false positive findings if the correlation is not properly modeled. In this paper, we will consider several methods for gene set analysis for paired case-control studies, including PAGE (Kim and Volsky 2005) and GAGE (Luo et al 2009), in which independence among genes in a gene set is assumed. We show that the independence assumption increases false positive rate by simulation. We then propose an improvement to the current methods.