In silico validation of pooled genotyping strategies for genomic evaluation in Angus cattle

In this study, we aimed to assess the value of genotyping DNA pools as a strategy to generate accurate and cost-effective genomic estimated breeding values (GEBV) of sires in multi-sire mating systems. In order to do that, we used phenotypic records of 2,436 Australian Angus cattle from 174 sires, including yearling weight (YWT; N = 1,589 records), coat score (COAT; N = 2,026 records), and Meat Standards Australia marbling score (MARB; N = 1,304 records). Phenotypes were adjusted for fixed effects and age at measurement and pools of 2, 5, 10, 15, 20, and 25 animals were explored. Pools were created either by phenotype or at random. When pools were created at random, 10 replicates were examined to provide a measure of sampling variation. The relative accuracy of each pooling strategy was measured by the Pearson correlation coefficient between the sire's GEBV with pooled progeny and the GEBV using individually genotyped progeny. Random pools allow the computation of sire GEBV that are, on average, moderately correlated (i.e., r > 0.5 at pool sizes [PS] ≤ 10) with those obtained without pooling. However, for pools assigned at random, the difference between the best and the worst relative accuracy obtained out of the 10 replicates was as high as 0.41 for YWT, 0.36 for COAT, and 0.61 for MARB. This uncertainty associated with the relative accuracy of GEBV makes randomly assigning animals to pools an unreliable approach. In contrast, pooling by phenotype allowed the estimation of sires' GEBV with a relative accuracy ≥ 0.9 at PS < 10 for all three phenotypes. Moreover, even with larger PS, the lowest relative accuracy obtained was 0.88 (YWT, PS = 20). In agreement with results using simulated data, we conclude that pooling by phenotype is a robust approach to implementing genomic evaluation using commercial herd data, and PS larger than 10 individuals can be considered.