|
Activity Number:
|
235
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, August 5, 2008 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
| Abstract - #301820 |
|
Title:
|
A Spike and Slab Centering Distribution in Dirichlet Process Mixture Models for Gene Expression
|
|
Author(s):
|
Sinae Kim*+ and David B. Dahl and Marina Vannucci
|
|
Companies:
|
The University of Michigan and Texas A&M University and Rice University
|
|
Address:
|
Department of Biostatistics, 1420 Wasington Heights, Ann Arbor, MI, 48109,
|
|
Keywords:
|
Dirichlet process mixture models ; Gene expression data ; Multiple testing ; Spike and slab prior
|
|
Abstract:
|
Model-based clustering methods using Dirichlet process (DP) mixture models have been proposed to exploit clustering for increased sensitivity in multiple hypothesis testing. Rather than yielding a probability of a hypothesis for each object, existing methods can only provide a ranking of the objects by their evidence for a particular hypothesis. In this work, we adapt the framework on Dahl et al. (2008) to accommodate point null hypotheses. For that, we use a spike and slab distribution which is a mixture of both a point-mass distribution and a continuous distribution as the centering distribution for the Dirichlet process prior. The method yields probabilities that genes follow the hypotheses of interest, whether those hypotheses be sharp or not. We apply our method in gene expression context and show how to simultaneously infer gene clustering and differential gene expression.
|
