P0553 Utilising information from heterogeneous SNP panels for genome analysis in dairy cattle

Mehar Khatkar , THE UNIVERSITY OF SYDNEY, Camden, Australia
Gerhard Moser , THE UNIVERSITY OF SYDNEY, Australia
Benjamin Hayes , Department of Primary Industries (Victoria), Melbourne, Victoria, Australia
Herman Raadsma , The Univ. of Sydney, Camden NSW, Australia
Genotyping with high density SNP panels is useful for accurate prediction of direct genomic breeding values, fine mapping and discovery of causal mutations affecting health and production phenotypes in dairy cattle. However, the cost of genotyping large number of study samples with very high density SNP chip is still high. We investigated various strategies of imputing genotypes from four different SNP chips i.e. 15k, 25k (Affymetrix), 50k and 800k(Illumina) The effect of using different sizes and types of reference panels on accuracy of imputation was investigated. We compared three commonly used software programs and IMPUTE2 gave higher accuracies of imputation compared to Beagle and fastPHASE. Accuracies of imputation increase with the number of SNPs in the test (imputed) set and with the increase in the number of samples in the reference set. SNP genotypes up to 800k can be imputed with high accuracies (genome-wide 0.79 % allelic error rate) from 50k and with moderate accuracies from panels with lower SNP densities. Similarly cross supplier SNP platforms (15k, 25k and 50k) could be successfully imputed (with 0.8 to 2.9 % mean allelic error rate). Using a panel of 800K imputed from 50K resulted in higher accuracies of genomic selection compared to the 50k panel, however the additional gains in accuracies were marginal. Genome wide association based on 800k imputed genotypes identified additional significant regions for a number of production and reproduction traits in dairy cattle.