General Departmental Seminar Series
Identifying differentially expressed genes under multiple treatment conditions: An empirical Bayes model regarding gene clusters
Yuan Ji, Ph.D. candidate, Department of Statistics, UW-Madison
Friday, February 21, 2003, 12-1 p.m.
6225 Medical Sciences Center (MSC), 1300 University Avenue
Statistical models have shown to be effective in identifying differentially expressed genes. We address several important issues in statistical modeling for DNA gene expression data. First, to improve model efficiency and account for gene dependence we construct an empirical Bayes model regarding gene clusters resulted from a preliminary cluster analysis; Second, to improve computational efficiency we propose a distribution approximation method instead of using intensive Monte Carlo simulation; Last, to assess the reliability of the set of selected genes we calculate their selection probabilities via Bootstrapping . We select genes according to a summary measure reflecting magnitudes of gene differential expression levels. Efficient modeling and empirical Bayes estimation enable us to afford the computational intensive bootstrapping, with which not only can we identify the set of target genes, but also provide a measure of confidence level for each selected gene. We demonstrate our methodology with a detailed case study from Toxicogenomics and evaluate model performance with a simulation study.
Back to General Departmental Seminar Series