Comparison of methods for predicting genomic breeding values for growth traits in Nellore cattle

Abstract

The objective of this study was to evaluate the accuracy of genomic predictions of growth traits in Nellore cattle. Data from 5064 animals belonging to farms that participate in the Conexão DeltaGen and PAINT breeding programs were used. Genotyping was performed with the Illumina BovineHD BeadChip (777,962 SNPs). After quality control of the genomic data, 412,993 SNPs were used. Deregressed EBVs (DEBVs) were calculated using the estimated breeding values (EBVs) and accuracies of birth weight (BW), weight gain from birth to weaning (GBW), postweaning weight gain (PWG), yearling height (YH), and cow weight (CW) provided by GenSys. Three models were used to estimate marker effects: genomic best linear unbiased prediction (GBLUP), BayesCπ, and improved Bayesian least absolute shrinkage and selection operator (IBLASSO). The prediction ability of genomic estimated breeding value (GEBVs) was estimated by the average Pearson correlation between DEBVs and GEBVs, predicted with the different methodologies in the validation populations. The regression coefficients of DEBVs on GEBVs in the validation population were calculated and used as indicators of prediction bias of GEBV. In general, the Bayesian methods provided slightly more accurate predictions of genomic breeding values than GBLUP. The BayesCπ and IBLASSO were similar for all traits (BW, GBW, PWG, and YH), except for CW. Thus, there does not seem to be a more suitable method for the estimation of SNP effects and genomic breeding values. Bayesian regression models are of interest for future applications of genomic selection in this population, but further improvements are needed to reduce deflation of their predictions.