Selection of Tagging SNPs and Imputation Efficiency of the 670K Commercial SNP Chip

With the collaboration of the equine genome community, we are leading the development of a high-density (~670,000 SNP) array with markers selected to optimize genotype imputation in the horse. Haplotype sharing among horse populations means that many polymorphisms discovered de novo by whole genome sequencing can effectively mark, or tag, the same haplotype in diverse breeds. Selecting the optimal subset of tag SNPs which capture the largest amount of haplotype information is a difficult, though well described computational problem. To address this issue in the horse, we processed whole genome sequence from 163 horses representing 32 breed groups to choose 2 million candidate SNPs for prototype commercial array design. These 2M SNPs were genotyped in 348 horses representing 10 major breed groups to assess SNP performance and provide independent sampling for tag SNP selection. Over 1.8M SNPs passed quality control and provided data for the commercial array design. We identified tagging SNPs in both inter- and intra-breed scenarios. Inter-breed tag SNPs comprise less than 15% of the original 2M candidate SNPs, yet imputation accuracy is more than 93%. With intra-breed specific tagging SNPs, highly diverse breeds required substantially more SNPs (20% of 2M panel) than less diverse groups (9% of 2M panel) to effectively tag all haplotypes. Inclusion of intra-breed tag SNPs in the array design improves imputation accuracy to over 95%. This commercial 670K array, the haplotype information from the 2M arrays, and the computational tools for imputation will be valuable resource to the equine genome community.