Discovery of QTL using a QTL-Effects Model in a Multi-Generational Pedigree

Abstract Text: Bayesian methods developed for genomic prediction using SNPs have become increasingly popular for QTL discovery in livestock. However, the hypothesis test for association based on SNP markers may not be valid for detecting QTL. This is because the posterior probability that SNPs in a genomic segment have nonzero effects is not equivalent to the posterior probability that a QTL exists in that segment. This discrepancy between association studies and QTL discoveries is expected to be bridged by modeling effects of unobserved QTL genotypes, which permits direct inferences on the QTL effects, their contributions to genetic variance and their locations. In addition, it has been shown that structured sources of information in a reference population, such as multiple generations, can be better exploited in a QTL-effects model. Thus, we hypothesized that using a QTL-effects model would benefit QTL discovery. The performance of the QTL-effects model for QTL discovery was compared with a marker-effects model, BayesB, using a 10-generation population simulated from real chicken genotypes and pedigree. We also demonstrated the manner in which credible intervals can be obtained for the positions of multiple unobserved QTL with an example. The QTL-effects model generally had less bias for the estimation of genetic variance contributed by the simulated QTL. The proportion of false positives at markers (PFP_M) from the QTL-effects model was closer to the proportion of false positives at QTL (PFP_Q) than from BayesB. When PFP_Q was limited to a small value, less than 0.2 for example, the QTL-effects model had greater power than BayesB in detecting QTL, especially for those contributing larger fractions of genetic variance. For a particular region of interest, the number of QTL segregating in that region can be inferred from the QTL-effects model. Conditional on the number of QTL, the posterior distribution for the ordered putative QTL position gives a credible interval for its location. These findings are important to researchers who are interested in characterization of QTL

Keywords: