Bioinformatics Seminar

Wednesday, April 8, 2009

3:00 - 4:00

Room 457 Blocker

Use score statistic to aid model selection for mapping multiple QTL

Zhao-Bang Zeng

Department of Statistics and Department of Genetics

Bioinformatics Research

North Carolina State University

Many quantitative traits are affected by multiple genes. To map those quantitative trait loci (QTL) in genomic positions, we can fit data with a model that includes multiple QTL and perform a model selection analysis. For such a model selection analysis, the choice of good criteria is an open problem. A common problem in many of the methods for mapping QTL that scan positions across the genome is the determination of a threshold value for the test statistic. Many factors, such as genome size, genetic map density, informativeness of markers, and proportion of missing data, may affect the distribution of the test statistic. Permutation test has been widely used to empirically estimate the null distribution and threshold. It works well for the null of no QTL, not so well for the null of some QTL. Also it is computationally intensive. Zou et al (2004 Genetics 168:2307-2316) proposed a model- and data-based resampling procedure using score statistics to empirically calculate the relevant threshold. It is much more efficient computationally and flexible to be adapted to complex models. We extend this procedure for multiple QTL to aid the model selection in the framework of multiple interval mapping (MIM). In this talk, I will discuss how to use this procedure for model selection with MIM and present some simulation results. Based on this procedure, we also come up general recommendations on practical procedures for QTL analysis based on MIM.