Health Research and Policy


DATE: April 11, 2013
TIME: 1:15 - 3:00 pm
LOCATION: Medical School Office Building, Rm x303
TITLE: Model selection procedures for genome wide association studies
SPEAKER: Prof. Dr. Florian Frommlet
Department of Medical Statistics, Medical University Vienna, Austria

After giving a brief introduction into genome wide association studies (GWAS) we will revise the most common statistical approaches. Till today the analysis of (GWAS) is almost exclusively performed using single marker tests, which means that the association of each single nucleotide polymorphism (SNP) is tested individually. The results from a comprehensive simulation study based on real SNP data will illustrate the shortcomings of single marker tests for complex traits, i.e. traits which are influenced by a larger number of genetic factors.

The second part of the talk will then introduce several different model selection approaches to GWAS. In particular we will discuss the statistical concepts underlying the software packages HLASSO, GWASelect and MOSGWA, and compare their performance in the context of case control studies. Discussion of MOSGWA will include the introduction of FDR-controlling modifications of the Bayesian information criterion, and heuristic search procedure which try to minimize this criterion over the extremely large model space in GWAS. Finally we will discuss potential directions in which to extend MOSGWA.

Suggested readings:
Balding, D.J., 2006. A tutorial on statistical methods for population association studies. Nat. Rev. Genet. 7, 781791.

F. Frommlet, F. Ruhaltinger, P. Twarog, M. Bogdan, 2012. Modified versions of Bayesian Information Criterion for genome-wide association studies.CSDA 56, 1038 - 1051.

Footer Links: