Andrej Rotter, 
Department of Pharmacology, The Ohio State University

A SAGE (serial analysis of gene expression) library is a collection of thousands of DNA "tags", each of which represents a distinct mRNA transcript. We model the count of each unique tag in a library as coming from a Poisson distribution, with its intensity parameter representing the abundance of the mRNA transcript. Focusing on the problem of identifying genes that are differentially expressed under different conditions, a Bayesian formulation is established. For genes that are differentially expressed, a different Poisson intensity parameter in each group is needed to explain the observed data. On the other hand, for genes that are similarly expressed across all groups, only a single parameter is sufficient to describe the counts. Under this formulation, the problem is to separate the differentially expressed genes from rest, and reversible jump Markov chain Monte Carlo method is adapted for this purpose. We will discuss our application of the method to analyzing 6 mouse cerebellum libraries, trying to uncover genes that are associated with the process of aging in the cerebellum. Comparison of the results with those obtained from other methods will be presented as well.