How much extra information is gained using imputed genotype data?

Genotype imputation has been discussed widely as a tool to increase the statistical power associated with genome wide association studies (GWAS) and genomic prediction. Previous studies have examined the performance of imputation by evaluating how well a validation set has been predicted. The aim of this study was to examine the amount of extra information added by utilising genotype imputation. The number of new haplotype combinations, between adjacent loci, were estimated for multiple genotype densities from an Australian Sheep dataset. In our example, using genotypes from OAR6, imputation increased the number of haplotypes for 81% of the regions when imputing from 12k to 50k. Large distances between adjacent low density markers resulted in higher numbers of new haplotypes. This also corresponded to a greater proportion of low frequency haplotypes. When imputing from HD to WGS no information was added for 50% of the regions and there was a greater proportion of haplotypes with only 1 observation. Estimating the number of new haplotypes from imputation provides an understanding about the value of imputation and can be utilised to help design of reference genotype datasets.