Clustering Posterior Distributions: Application to Somatic Embryogenesis in Maize

• Activity Number: 226
• Type: Contributed
• Date/Time: Tuesday, August 9, 2005 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): Tanzy Love*+ and Alicia Carriquiry
• Companies: Iowa State University and Iowa State University
• Keywords: maize ; empirical probability distributions ; clustering ; MCMC

Abstract

In large time series or multiple treatment microarray experiments, we are interested in locating groups of genes that react together. Subject matter theory designates these groups as coregulated by the same biologic pathways. For example, genes responsible for photosynthetic processes may express together in an experiment covering time periods in light and darkness. The statistical problem is then clustering genes based on their expression values over multiple treatments. However, we don't have values for gene expression, rather replicated estimates for each treatment condition. We have used the mean of the gene expression estimates in clustering genes. However, this ignores the uncertainty we have in the actual values of expression. To incorporate this uncertainty, we have modeled expression estimates using hierarchical models. This provides posterior probability distributions for quantities such as expression value and expression ratio for two treatments. We also can construct the joint posterior probability distribution of all expression values or all pairwise expression ratios for each gene.