Hierarchical Bayesian Analysis for Temporal Microarray Gene Expression Data

• Activity Number: 277
• Type: Contributed
• Date/Time: Tuesday, August 9, 2005 : 10:30 AM to 12:20 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Xueya Cai*+ and Yulan Liang and Yue Li
• Companies: University at Buffalo and University at Buffalo and University of Rochester
• Keywords: Bayesian hierarchical model ; Monte Carlo Markov Chain ; Multiple testing ; Gene filtering ; Multiple sclerosis ; IFN-ß treatment

Abstract

Microarray measurements provide an enormous amount of data for analysis. A lot of statistical models have been applied in assessing differentially expressed genes. The goal of this study is to provide the statistical evidence for identifying genes with differential temporal expressions and for tracking their dynamic patterns during the IFN-ß-1, a treatment for relapsing multiple sclerosis patients. Several Bayesian hierarchical models are developed and compared for taking account of the multilevel heterogeneous variations of the expression values, the high noise levels, and---most importantly---the correlated temporal information. Markov chain Monte Carlo (MCMC) algorithm is implemented to take this task for simultaneous sampling of the conditional posterior distributions of all parameters. In addition, multiple testing problems are carefully handled through regularized t-posterior statistics. The results from the developed models provide the most probable set of significant genes on the basis of the estimated posterior probabilities and regularized t-posterior statistics and allow us to describe the "stochastic profiles" of these genes under different conditions.