A Bayesian Hierarchical Model for Correlated Microarray Data Sets

• Activity Number: 404
• Type: Contributed
• Date/Time: Tuesday, August 2, 2011 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Bernard Omolo*+ and Ming-Hui Chen and Haitao Chu and Joseph G. Ibrahim
• Companies: The University of North Carolina at Chapel Hill and University of Connecticut and University of Minnesota and The University of North Carolina
• Keywords: Bayesian hierarchical model ; cell-line ; correlation coefficient ; gene expression ; microarray data ; probe

Abstract

Assessment of gene-specific correlation between two independent expression datasets may help in deciding whether to use an original expression data or one updated with additional samples, for differential gene expression analysis. This can be accomplished through modeling the parameters measuring association between variables, for instance, the correlation coefficient. Typically, the correlation coefficients would be computed using the mean expression value for each common gene and cell-line between the two datasets. However, this approach does not utilize the replicated expression values for each gene and instead averages over them, thereby ignoring the effect of multiple probes per gene. We propose a three-level Bayesian hierarchical model for the gene-specific correlation coefficient between two independent datasets that utilizes replicated expression values for each gene. A comparison with the standard approach indicates that the Bayesian approach performs better and hence is more preferable for differential gene expression analysis.