Assessing accuracy of imputation using different SNP panel densities in a multi-breed sheep population

Background: Imputation is a robust tool for minimizing costs of genotyping but there are many factors that can influence imputation accuracy, providing opportunities for further improvements and optimal implementation of this technology. The use of imputation is also an important strategy within the New Zealand sheep industries implementation of genomic selection. The objective of this study was to provide practical directions on the implementation of imputation strategies in a multi-breed sheep population genotyped with 3 different marker panels: 5K, 50K and HD (600K markers). Results: Imputation from 5K to HD was slightly better (0.6%) than the imputation to 50K using the same low density panel. The imputation in two steps from 5K to HD (5K to 50K and then from 50K imputed to HD) outperformed the imputation straight from 5K to HD. A slight loss in accuracy with the large fixed reference population, was only seen for a few animals across all imputation scenarios from 5K to 50K, in addition of a high gain in accuracy for a large proportion of animals (purebred and crossbred animals) in the imputed set, justifies the adoption of a fixed and large reference set for all situations. This study showed that only Chr26 had overall imputation accuracy from 5K to 50K of 100 SNPs in each tail higher than 60% (r2). Most of the chromosomes had problems in at least one of the ends. The prediction of imputation accuracy before imputing was also implemented efficiently in this study. FIMPUTE V2.2 outperformed BEAGLE 3.3.2 in all imputation scenarios. Conclusions: Imputation accuracy in sheep breeds can be improved efficiently by following a set of recommendations such as marker panels and software to be implemented, strategies of imputation (one or two step imputation), and choice of animals to be genotyped using both high and low density panels. Incorporation of additional markers in the sparsest panel (5K) is required for imputation accuracy improvement.