Yekutieli D., Benjamini Y. (Tel Aviv University, Israel)
Genetic dissection of quantitative traits using the False Discovery Rate criterion
Genetic dissection of quantitative traits is achieved through a series of individual statistical tests, each testing the effect of the genetic structure in a given locus on one of many quantitative traits. The problem of multiple comparisons is the main statistical problem in quantitative trait mapping, some researchers even stressed that the resolution of this problem has important consequences on the future of the field. Correcting for multiplicity in a QTL study is very difficult due to the large number of hypotheses tested, often exceeding 100,000 tests and complex dependency structure. Dependency between trait measurements. For each trait, test statistics corresponding to closely located genetic loci are highly correlated. The presentation will focus on addressing the problem of multiple comparisons in quantitative trait mapping using the False Discovery Rate. If there are many QTLs the FDR thresholding is lower than the conventional Family Wise Error thresholding. The increase in power is particularly evident in large multiple comparison problems, where the conventional approach lacks power. I will show the validity of the existing FDR controlling procedures in QTL mapping. I will present the results of simulation studies and apply the FDR controlling procedures to real data. Finally a two stage FWE controlling modeling scheme will be presented. In the first stage the FDR criterion is used for screening promising QTLs. The second stage is a confirmatory study on the screened QTLs.