Direct and indirect genomic evaluations in beef cattle

Abstract

We tested two modifications into single-step genomic BLUP (ssGBLUP) that allow it to work with a large amount of genotyped animals. The first method is based on genomic recursions (APY) to construct the inverse of the genomic relationship matrix without directly inverting it; all available genotyped animals are included in ssGBLUP with APY, but they are split into base and non-base, and the method returns direct predictions. The second method is an interim genomic evaluation (IP) for young genotyped animals; only a reference set of animals are used in ssGBLUP with IP, and the method returns indirect predictions for young genotyped animals. A dataset from American Angus with records for growth traits was used. Over 8 million animals were in the pedigree, of which 51,883 were genotyped. The ssGBLUP with APY was as accurate as regular ssGBLUP when the number of genotyped base animals was at least 10,000; this method was also faster and required less memory. The ssGBLUP with IP mimicking the previous official evaluation returned the same accuracy of GEBV for young animals as the regular ssGBLUP. While the first method enables complete genomic evaluations for huge genotyped populations, the second allows for quick genomic predictions on young animals without including their information into a new run of evaluation.