GENOME-WIDE SELECTION AND FEATURE SUBSET SELECTION OF MARKERS BY BAYESIAN NETWORKS IN BEEF CATTLE USING MULTIPLE-TRAIT MODEL

Abstract

Several methods have been used for genome-enabled prediction, where multiple regression models describe a target variable with a linear function of a set or subset of covariates. Bayesian Networks has offered interesting tools for a more parsimonious representation of the join distribution of a set of variables, which are useful for prediction purposes, e.g. using Markov Blanket (MB) of the target variable. Genome prediction studies have been focused mostly on single-trait analyses. However, most of economically important traits are genetically correlated, and it is expected to increase the accuracy of prediction of genomic breeding values of genetically correlated traits using multiple-trait model. Thus, this study will be carried out to assess the accuracy of genomic prediction for carcass and meat quality traits in Nellore cattle using multiple-trait model on two different scenarios: i) genome-wide prediction using the whole SNP data; and ii) genomic prediction using subset of SNP markers selected using MB. It will be used data from 8,000 phenotyped and genotyped Nellore cattle. The animals have been genotyped using low-density SNP panel, and subsequently imputed for arrays with 54k and 777k SNPs. The MB learning method will be used to identify a minimum set of molecular markers associated with carcass and meat quality traits. Four Bayesian specifications of genomic regression models, namely Bayes A, Bayes B, Bayes CÀ and Bayes R, will be compared in terms of prediction accuracy using a five folds cross-validation. The results of this study will allow accessing the impact of genomic information upon genetic evaluations in beef cattle using multiple-trait model, which has shown advantageous regarding univariate models, because it takes into account the selection process through the use of several traits at the same time. Therefore, this project will be the technical and economic viability evaluation of multiple-trait genomic analyses, where there is information from both pedigree and genomic data.