Genetic Analyses Of Discrete Traits Controlled By Polygene Using Generalized Linear Methods

Linear mixed models provide a powerful means of genetic analyses. However, for many traits of economic importance the assumptions of linear responses, constant variance, and normality are questionable. So methods for continuous data are not appropriate for discrete data. Many economically important characters in livestock vary in a discontinuous manner but are not inherited in a simple Mendelian fashion. These traits are presumably controlled by polygenes with expressions modified by environmental effects like quantitative traits. Owing to the recent advances in marker technology, the molecular dissection of heritable quantitative traits into their individual Mendelian components or quantitative trait loci (QTL) is now possible, and novel statistical techniques to detect and map individual genes affecting quantitative traits have been developed and subsequently refined. However, most QTL mapping methods share a common assumption that the phenotype follows a normal distribution and hence are not appropriate for categorical traits. Thus, new methods are needed for QTL mapping for categorical traits.Generalized linear models (GLM) have attracted considerable attention over the years. GLM is extend from ordinary linear mixed models, and can deal with discrete or continuous data conveniently. It differs from linear mixed model in that it focuses more on the inverse link function to model the relationship between the linear predictor and the conditional mean. GLM provide a flexible way to model production traits which do not satisty the assumption of a linear mixed model.In this study, methodology of genetic analyses and QTL mapping for discrete traits in livestocks basing on the framework of a generalized linear model was present. The parameters were estimated using Maximum Likelihood method. The efficiency and power were compared with the linear model. The factors of influencing genetic analyses and QTL detection efficiency were simulated in this study too. Daughter design with multiple families was applied, and the number of segregating population is 500. From the results of simulation study, following conclusions can be derived. (1) GLM has a great advantage in heritability and breeding value estimation, and has nice accuracy for discrete traits. (2) The accuracy of estimation of heritability depended on the true heritability and incidence of the trait directly. (3) GLM can be extended for different situation such as general populations. (4) The GLM method has a certain advantage in location estimation and power of QTL mapping, and has nice efficiency and accuracy for ordinal traits. (5) The accuracy of QTL mapping largerly depended on the effect of putative quantitative trait loci and the value of heritability. (6) The accuracy and power of QTL mapping depended on the distance between the two flanking markers. (7) The threshold value and the number of underlying liability can influence the power and accuracy of QTL mapping, too. (8) The GLM method for QTL mapping can be extended for different situation (e.g. general populations, multiple markers and different distribution of phenotype).