The accuracy of genomic selection due to linkage analysis information and due to linkage disequilibrium

Genomic selection (GS) is generally thought to use linkage disequilibrium (LD) between the markers and QTL to predict breeding values. Recently, Habier et al. (2007) showed that GS implicitly also uses linkage analysis information. Here we studied how much of the accuracy of genomic selection EBV (GEBV) is due to linkage analysis information and how much due to LD that already existed in the founders of the pedigree. The study was performed on milk yield, fat yield and protein yield. GEBV was predicted using genotype data (45,888 SNPs) and phenotypic data of 255 British Holstein bulls. Breeding value prediction using only linkage analysis information was obtained by applying the Fernando and Grossman approach to marker assisted selection across all 45,888 loci, using 5 generations of pedigree data. The study showed that 87~95% of the accuracy of GEBV was due to linkage analysis information.