Activity Number:
|
15
|
Type:
|
Contributed
|
Date/Time:
|
Sunday, August 11, 2002 : 2:00 PM to 3:50 PM
|
Sponsor:
|
ENAR
|
Abstract - #301380 |
Title:
|
Model Selection Criteria for Gene Expression Data
|
Author(s):
|
Cynthia Coffman*+ and Marta Wayne and Lauren McIntyre and Sergey Nuzhdin
|
Affiliation(s):
|
Duke University Medical Center and University of Florida and Purdue University and University of California, Davis
|
Address:
|
508 Fulton St., Biostatistics/Bioinformatics, Durham, North Carolina, 27705,
|
Keywords:
|
gene expression ; model selection criteria ; Drosophila
|
Abstract:
|
The high dimensionality of gene expression data makes the description of models relating genes to each other and to a trait of interest difficult. Factor analysis is used to reduce the dimensionality of the data allowing a set of hierarchical models to be defined to describe the relationship between the phenotype and the factors. Now, standard model selection criteria (Akaike's Information Criteria (AIC), Bayesian Information Criterion (BIC), and Non-parametric Cross Validation (NCV) can be applied to the hierarchical models. To determine whether this strategy results in biologically meaningful inference, we simulated data based on several simple genetic regulatory pathways and applied model selection criteria. We then compared these results to the known simulation conditions. We also applied the procedure to a set of expression data gathered on Drosophila melanogaster for the trait body size (thorax length). Body size is related to reproductive success in both sexes in Drosophila melanogaster. Body size is a complex polygenic trait, making this system a good candidate for the analysis proposed.
|
- The address information is for the authors that have a + after their name.
- Authors who are presenting talks have a * after their name.
Back to the full JSM 2002 program |