Validation of Single Step Genomic Best Linear Unbiased Prediction in Beef Cattle

Abstract

This study investigated the accuracy of predicting future phenotypes of young Angus and Hereford cattle using Single-step Genomic BLUP (SSGBLUP) compared to the traditional pedigree-based BLUP evaluation (NRMBLUP). Forward cross-validation, using two comparison methods, was used to quantify the predictability of the two evaluations. For each breed, two data sets named ‘full’ and ‘partial’ were generated. The ‘full’ data set included all relationships, all genotypes and phenotypes of animals born up to November 2018. For ‘partial’ data sets, phenotypes of animals born after December 2014 were removed and the data for animals removed after December 2014 were used as the ‘validation data set’. SSGBLUP and NRMBLUP analyses were performed separately for the full and partial data sets and EBVs were predicted for animals in the validation data set. In Method 1, R squared values (R²), regression coefficients (REG) and adjusted correlation (ACOR), between pre-corrected phenotypes and predicted EBVs were compared. In Method 2, correlation ratios between EBVs from full and partial evaluations were estimated to calculate the increase in predictability between the SSGBLUP and NRMBLUP. The estimated R², REG and ACOR using SSGBLUP were higher than those from NRMBLUP. A similar pattern was observed for correlation ratios from Method 2. The increase in ability to predict future phenotypes using Method 1 ranged from 30 to 50% and 10 to 36% for genotyped and 2 to 4% and 1 to 2 % for non-genotyped Angus and Hereford cattle, respectively. Using Method 2, the ability to predict future phenotypes ranged from 22 to 40% and 6 to 28% for genotyped and 1 to 2% and 0.5 to 1 % for non-genotyped Angus and Hereford cattle in the validation set, respectively. This study showed that there was an increase in the accuracy to predict future performance from SSGBLUP compared to NRMBLUP in Angus and Hereford cattle. The increase in predictive ability varied according to the heritability of a trait, the number of phenotypes and genotypes included in the evaluation and whether the animals were genotyped or not in the evaluation.