Contribution of Genome-wide IBD to the Accuracy of Genomic Selection

The prediction of genome-wide breeding values (GW-EBV) in genomic selection (GS) is generally thought to be achieved by using linkage disequilibrium (LD) between the markers and QTL. Habier et al. showed that GS implicitly also uses genetic relationships to achieve accurate predictions. Here we tried to quantify to what extent the reliability of GW-EBV prediction is due to genomic relationships between animals generated in the known pedigree and how much is due to LD that already existed in the founders of the pedigree. The study was performed on milk yield, fat yield and protein yield. GW-EBV was predicted using genotype data (35,706 SNPs) and phenotypic data of 1,086 Italian Brown Swiss bulls. GW-EBV was predicted using a genomic identity-by-state (IBS) relationship matrix and a genomic identity-by-descent (IBD) relationship matrix (averaged over all marker loci), respectively. The IBD matrix was calculated by linkage analysis using known family relationship and the Fernando and Grossman approach to set up IBD matrices and using 2-6 generations of pedigree data. The study showed that GW-EBV prediction using only IBD genomic relationships within the known pedigree achieved similar reliability to that of the prediction using IBS, which implicitly also accounts for genomic relationships that occurred before the known pedigree.