Genotype Imputation Accuracy and Proof of Principle in the Welsh Pony

Genotype imputation from low-density (LD) to high-density (HD) single nucleotide polymorphism (SNP) chips is a cost-effective approach to increase the power of genome-wide association studies. Genotyping a proportion of a sample cohort on low and higher density arrays and imputing to higher density is a potentially cost-effective strategy. To evaluate the accuracy of genotype imputation using a within population reference sample, and to determine the minimal number of reference individuals necessary, 234 Welsh ponies were divided into reference and validation sets with varying numbers of individuals included in the reference set. Individuals from the validation set had their HD genotypes masked, except for the genotypes of markers present on the EquineSNP50 array, simulating a situation where these ponies were genotyped on the LD array. Criteria used to assess imputation accuracy included genotype concordance and Pearson’s correlation coefficient between imputed and observed genotypes. As proof of principle, a GWAS for height was performed first as if the ponies were genotyped on the LD array, secondly, in the imputed datasets with varying reference population sizes and finally in the real dataset. HMGA2 is a confirmed QTL for height located on ECA6.  The LD array poorly covers this region and thus provides a prime example to evaluate the capability of genotype imputation to improve power to detect causal loci. Imputation accuracy improved with increasing numbers of individuals in the reference population along with improved power to detect a causal variant poorly tagged by the LD array.