|
Activity Number:
|
225
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 3, 2009 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
| Abstract - #303854 |
|
Title:
|
Hierarchical Bayesian Variable Selection for Genetic Association
|
|
Author(s):
|
Deukwoo Kwon*+
|
|
Companies:
|
National Cancer Institute
|
|
Address:
|
6120 Executive Blvd., Rockville, MD, 20852,
|
|
Keywords:
|
Bayesian variable selection ; SNPs ; genetic association study
|
|
Abstract:
|
Complex diseases are functionally caused by a combination of environmental and genetic factors. Epidemiologists are posed with the problem of determining the specific causes and combinations of risk factors. With inexpensive genotyping technology available, in genetic association studies we analyze several thousands single nucleotide polymorphisms for each individual. Due to the large numbers of predictors available in genetic studies, we need to use variable selection techniques to decide which effects to include in a model that relates risk factors to phenotypic outcomes. Therefore, we present a hierarchical Bayesian variable selection method, which is an extension of the stochastic search variable selection. We introduce two latent binary vectors in a hierarchical manner to model the relationship between genes and SNPs and use generalized linear models to relate them to a phenotype.
|
