Cross-validation of single step BLUP applied to terminal sire sheep in Australia

Abstract

To inform decision making regarding the implementation of single step for routine genetic evaluations in sheep in Australia, forward cross-validation and 5-fold cross-validation schemes were used to compare single-step genomic BLUP to pedigree BLUP (hereby referred to as SS-GBLUP and ABLUP, respectively). Accuracies and slopes of EBVs regressed against phenotypes were used to validate the two methods used. Accuracies for genotyped animals from forward cross-validation were, on average across traits, 0.073 higher for SSGBLUP compared to ABLUP and were 0.092 higher calculated using 5-fold. Slopes for genotyped animals from forward validation were, on average across all traits for reference flock animals 0.92 for SS-GBLUP EBVs and 0.80 for ABLUP EBVs, whereas for 5-fold cross-validation the respective slopes where 0.93 and 0.89. Greater increases in accuracies were observed in genotyped sheep with these increases also observed in hard to measure traits, including worm egg counts and carcase traits.