Genomic prediction ability for carcass composition indicator traits in Nellore cattle

Abstract

Abstract The aim of this study was to compare the genomic prediction ability for carcass composition indicator traits in Nellore cattle using the Best Linear Unbiased Prediction (BLUP), Genomic BLUP (GBLUP), single-step GBLUP (ssGBLUP), Bayesian methods (BayesA, BayesB, BayesC and BayesianLASSO) and an approach combining the pedigree matrix of genotyped animals with both the genomic matrix and Bayesian methods. Phenotypic and genotypic information on about 66,000 and 21,000 animals, respectively, evaluated by National Association of Breeders and Researchers (ANCP) were available for body structure (BS), finishing precocity (FP), musculature (MS), Longissimus muscle area (LMA), back fat thickness (BF) and rump fat thickness (RF). The genotypes were obtained based on the low-density panel Zoetis CLARIFIDE® Nellore version 3.1 containing 30.754 markers. To obtain the prediction ability, the dataset was split into training (genotyped sires and dams with progenies) and validation (genotyped young animals without progeny records and without phenotypes) subsets. For genomic models, the predictive ability was assessed through the correlation between the deregressed expected progeny differences and DGVs. For BLUP model, the prediction ability was evaluated through the correlation between estimated breeding value (EBV) and deregressed expected progeny differences (dEPD). To evaluate the extent of prediction bias the linear regression coefficients between the response variable (dEPD) and DGVs (or EBVs for BLUP model) considering only the animals in the validation set, were calculated. In terms of prediction ability and bias, Bayesian approaches were superior for visual scores traits and the ssGBLUP for carcass traits obtained by ultrasonography, however, more biased results were obtained for BF and RF using the ssGBLUP. The ssGBLUP model showed less biased prediction for low heritability traits, such as LMA, and also it has lower computational demand and it is a straightforward method for implementing genomic selection in beef cattle. Therefore, earlier reliable genetic evaluation of unproven sires trough genomic selection is appealing in order to increase the genetic response for carcass traits in the Nellore (Bos taurus indicus) beef cattle.