Mapping QTL For Quantitative Traits And Threshold Traits In Animal Outbred Populations Using Bayesian-MCMC Method

Designs based on crosses between inbred lines or outbred lines are developed and worked well for QTL mapping in experimental organisms, which usually cannot suit well husbandry animal populations. Furthermore, it is a great challenge to map QTL underlying threshold traits and quantitative traits as well in animal outbred population because of unknown information of marker linkage phrase and discrete nature of threshold traits which resulting in reduced variation in the phenotypic distribution. In this paper, studies on mapping QTL controlling quantitative trait and threshold traits were performed respectively using Bayesian-Markov Chain Monte Carlo (MCMC) approach combined with Identity-by-Descent (IBD)-Based variance component model by Monte Carlo simulation. The method of multipoint inference of IBD matrices was modified in our study so that all marker information could be extracted and the distribution of IBD between half-sibs was easy to be calculated. Three samplers including Gibbs sampling, Metropolis-Hastings algorithm and reversible jump MCMC, were implemented to generate the joint posterior distribution of all unknowns so that the QTL parameters were obtained by Bayesian statistical inferring. Utilities of the method are demonstrated and examined using simulated animal populations under different considerations.The results of Bayesian analysis on quantitative traits showed that lengths of the 90% credibility intervals of QTL positions varied from 5cM to 22cM in all simulated data. Moreover, with the increasing of the QTL effects and the family size and the decreasing of the family number, the accuracy of the parameter estimates augments. So Bayesian-MCMC method is powerful for mapping QTL for quantitative traits in animal outbred populations.Conclusions can be derived from QTL detection for threshold traits as follows: (1) Basically, estimates of QTL parameters are according to their "true" values which suggested that the method used in the paper was available and robust. (2) QTL effect and family structure play important roles for threshold traits as same as quantitative traits, and more full-sib size could improve accuracy and precision of estimates when family size is small under full-half sib family design. (3) The posterior modes of the QTL numbers overlapped with the true number of QTL almost in all simulated conditions, which proved that RJ MCMC was perfect in QTL number estimating. (4) Higher marker density and marker polymorphism are helpful to gain more accuracy parameter estimates in QTL mapping for threshold traits. (5) When there are more than two QTL linked, the more marker interval between them, the less amount of interference and the higher detection accuracy. (6) Under biallelic QTL model, although detection power is lower compared with normal-effects QTL model, the length of QTL 90% credibility intervals still can be within 14cM. (7) As incident rate of threshold trait decreased, the QTL interval lengthened and support probability became small as well, and the MSE of all other model parameters augmented at the same time. (8) Basing on the identical liability, estimates obtained from traits with continuous, binary and three-ordered categorical distribution showed that detection power for continuous trait is highest among them.Finally, when we utilize some optional mapping strategies, Bayesian-MCMC method showed many advantages over other traditional procedure in QTL mapping, especially for threshold traits or traits with multiple QTL under outbred populations design. Our method and program ran in this paper can also easily deal with actural data in QTL mapping.