Comparison of High-Density Genomic Relationship Matrices in Beef Cattle

Genomic similarity has been used to manage genetic diversity, parentage testing, and selection. The adoption of relationship matrices based on markers allows the use of exact fractions of shared genes, providing more accurate predictions than use of the expected fractions as implemented by the traditional pedigree based relationship matrix (A). New methods have been proposed to adjust the genomic relationship matrices for a multi-breed population (Harris and Johson,2010). In this study, the genomic relationship matrix GRM defined by VanRaden (2008) was implemented and compared with the method based on Euclidean distance matrix (EDM), originally proposed by Gianola and van Kaam (2008) and modified by Harris et al. (2011) in a multi-breed, crossbred taurine beef cattle population genotyped (n = 96) with both the Illumina High-Density Bovine BeadChip Array (777,962 SNP) and Affymetrix Axiom Genome-Wide BOS 1 Array (648,874 SNP). Correlations among the 4 derived matrices (IlluminaEDM, IlluminaGRM, AffymetrixEDM and AffymetrixGRM) were evaluated and the correlations between the two panels were 0.9971 and 0.9994 for the methods EDM and GRM, respectively. Correlations between the two different methods for the same panel decreased the relation by 4.45 and 4.30% for the Illumina and Affymetrix panels, respectively. These results suggest that the genomic relationship matrix using HD panels is influenced more by the method used to calculate the relationships than the underlying panel as matrices with either panel were almost perfectly correlated.  Further research is required to determine which method provides the more accurate relationship matrix.