A Variable Selection Approach to Gene Set Enrichment Analysis

Dongmei Liu
Research Fellow
Department of Epidemiology and Population Health
London School of Hygiene and Tropical Medicine
*Candidate for the Biostatistics/ Cancer Center Faculty Position

Monday, March 31
3:30pm
Mayo D199
Minneapolis Campus

Abstract:

Several methods and programs are developed to study association of functional classes and disease using microarray data. Most of them are "enrichment tests'' based on the assumption that the association between disease and functional classes is captured by significant differential expression of some genes in the classes. These approaches usually take dichotomas input, i.e. if a gene is differentially expressed or not, and consider the functional classes one at a time. We propose a Bayesian hierarchical model to study the association between disease and functional classes by taking continuous measure of gene-specific disease effect as input and testing all possible disease associated functional classes at the same time. There are three important improvement of Bayesian hierarchical model over conventional enrichment test. Firstly, it doesn't set arbitrary cutoff on differential express. The dichotomous status of whether a gene is differentially expressed is stochastic. It depends on the distribution of gene-specific disease association measure across the whole genome. Secondly, it does not only detect association due to enrichment and/or depletion of significantly differentially expressed genes, but also detect the association due to slightly up or down regulated expression as a group. Thirdly, the model tests multiple functional classes at the same time and distinguishes the classes that are most responsible for enrichment where there are multiple enriched overlapping classes. In the end, we apply the model to analyze two microarray data sets, one studying differential expression associated with age-regulated macular degeneration, the other studying association between breast cancer and various KEGG pathways.

A social tea will be held at 3:00 P.M. in A434 Mayo. All are Welcome.
For more details contact 612-624-4655 or see http://www.biostat.umn.edu/seminar_academic.html