Across- and within-breed imputation across several genotyping densities in dairy and beef cattle.

D. Berry,M. Mullen,A. Cromie,N. Villalobos
Abstract:Illumina high density genotypes (777,962 SNPs) were available on 3,122 dairy and beef bulls. Animals were partitioned into either a calibration or validation dataset to test the accuracy of imputation. All animals, irrespective of breed, born after 2005 (n=698) were assumed to represent the validation bulls. The high density genotypes were masked in the validation animals to represent a low density (n=6,501) or medium density (n=47,770) genotyping platform. The accuracy of within breed imputation (i.e., correlation between actual and imputed genotype) from medium density to high density (0.987) was superior to that from low-density to high density (0.938) with the difference diminishing as the proportion of back-pedigree genotyped on the highdensity platform increased. Using multiple breeds in the calibration dataset for imputation did not improve the accuracy of imputation.
What problem does this paper attempt to address?