Imputation of missing genotypic data in heterogeneous populations

The objective is to assess the effect of population structure on the accuracy of imputing missing SNP genotypes. Data from commercial SNP-arrays used to genotype multiple cattle and sheep breeds are available. The within-breed imputation accuracy will be compared to the multi-breed (multiple breeds in the training and testing datasets) and across-breed (one breed to impute the other) accuracy. Two main scenarios are envisaged: i) gap-filling: after genotyping, a small proportion of locus-sample cells are left uncalled. These missing genotypes usually need to be imputed before moving on to downstream analyses; ii) mixed geneotyping strategies: to optimize costs, typically a fraction of the samples is genotyped with a high-density SNP array, while the remaining samples are genotyped with a low density array. Here the objective is to impute from low to high density (more challenging scenario)