Comparison of Meta-Analytical Approaches for Gene Expression Profiling

• Activity Number: 102
• Type: Contributed
• Date/Time: Monday, July 30, 2007 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): Heather Adams*+ and Sandra Rodriguez-Zas and Bruce Southey
• Companies: University of Illinois at Urbana-Champaign and University of Illinois at Urbana-Champaign and University of Illinois at Urbana-Champaign
• Keywords: microarray ; mixed effects models ; meta-analysis

Abstract

Meta-analysis allows the integration of information across studies, enhancing the estimation of effects evaluated across experiments. We evaluated two approaches to combine microarray data that have complementary advantages. Results from meta-analysis of summary results and from joint analysis of raw gene expression data pertaining to four studies profiling gene expression of honey bees at two adult maturation stages were compared. A total of 192, 89, 2207, and 5 genes had significant differential expression (P< 0.01) within study, 1030 and 31 genes were significant in the joint and summary meta-analyses, respectively. These comparable results suggest that although joint analysis accounting for within and between studies sources of variation is desirable, the limited availability of raw data and descriptions necessary for adequate modeling make meta-analysis a suitable alternative.