| The inclusion of genomic information into animal breeding predictions can provide more accurate breeding values for selection candidates at a younger age and facilitate selection for expensive or difficult to measure traits. Genomic prediction involves using large training populations with phenotypes and high density genotypes to estimate genetic effects throughout the genome. Implementing genomic selection in the beef industry presents challenges, both in development and application. To investigate the potential for accurate across breed genomic prediction, Bayesian algorithms were used to derive prediction equations for 52,156 SNP loci using approximately 2,000 mixed purebred bulls or 3,400 crossbred cattle with deregressed EPD or phenotypes for growth and carcass traits. MBV accounted for up to 18% of genetic variation in a pooled, multi-breed analysis and up to 42% in single breed subpopulations, depending on trait, and were more accurate in Angus and Hereford breeds as these were highly represented in both multi-breed training populations. These multi-breed derived genomic predictions were compared with Angus-derived genomic predictions in a commercial cattle population consisting of Angus bulls bred on four commercial cow-calf ranches in Northern California and producing nearly 6,000 phenotyped progeny. Compared with Angus-derived genomic predictions, multi-breed genomic predictions had lower accuracy (<0.3). In contrast, Angus-derived genomic predictions using 384 or 50,000 SNP had higher average accuracy (0.38-0.61). Further, Angus-derived 50,000 SNP genomic predictions were more predictive of future progeny performance than breed association EPD and low accuracy ranch progeny derived EPD, suggesting that single breed genomic predictions derived from training populations of several thousand animals are a better selection tool for commercial producers than breed association EPD lacking any genomic input. |
