Page Banner

United States Department of Agriculture

Agricultural Research Service

Title: Detection of Quantitative Trait Loci with the Aid of Genetic Markes, the Problem of Multiple Comparisons

item Weller, Joel - ARO, VOLCANI CTR, ISRAEL
item Ron, Micha - ARO, VOLCANI CTR, ISRAEL
item Vanraden, Paul
item Wiggans, George

Submitted to: Proceedings of the International Conference on Multiple Comparisons
Publication Type: Proceedings
Publication Acceptance Date: October 1, 1996
Publication Date: N/A

Technical Abstract: Effects of individual quantitative trait loci can be isolated with the aid of linked genetic markers. Saturated genetic maps including hundreds of DNA-level markers have been developed for humans and mice and are being developed for many agricultural species. Most studies have analyzed each marker or pair of linked markers separately for each trait included in the analysis. Thus, the number of contrasts tested can be quite large, and a comparison-wise type I error of 0.01 or even 0.001 is meaningless. The overall experiment-wise type I error can be readily derived from the nominal type I error if all contrasts are statistically independent, but different markers on the same chromosome are correlated, and different traits are generally correlated. With a saturated genetic map, we propose to reduce the number of comparisons by collapsing information on all markers in a linkage group into the specific events of recombination for each individual. Thus, it is necessary to perform only a single test for each chromosome, and the hypothesis being tested is closer to the question of interest. With multiple traits, we propose analysis of canonical variables derived from the actual traits. Each experimental unit, the chromosome-canonical variable combination, will be independent from all other experimental units, and the experiment-wise type I error can be readily calculated. Furthermore, if the original number of traits is large, it should be possible to reduce the number of variables analyzed by deleting variables with very low eigenvalues. In addition, by analysis of canonical variables, it is possible to determine whether an effect observed on two traits is due to two different quantitative trait loci or to a pleiotropic effect of a single locus.

Last Modified: 4/19/2015
Footer Content Back to Top of Page