Impact of genotype imputation and density of markers on the accuracy of genomic prediction

Using SNP markers information and genomic evaluation approach, predicting the genetic merit of individuals without phenotypic records is now possible. However, using high-density panels for genomic evaluation of all individuals is not economically feasible. To achieve high genomic prediction accuracy with reasonable price, it is possible to genotype a proportion of animals with high-density panels and the rest of animals with low-density panels then impute them to high-density genotypes. In this study, the effect of three low-density panels (1k, 2k and 4k), genotype imputation to 10k panel and the relationship between reference and validation populations on the accuracy of genomic predictions and also the correlation between the estimated breeding values using panels with different densities in simulated data were assessed. The low density panels genotypes were actually consisting of 10, 20 and 40 percent of 10k markers selected randomly and FImpute was used for genotype imputation. As a general trend, by increasing the density of markers, the correlation between the estimated breeding values was increased using different panels. So, the accuracy of genomic predictions was similar using 4k and 10k genotypes. Moreover, imputing 4k to 10k genotypes, did not improve the accuracy of genomic prediction. However, the accuracy of estimated breeding values was increased after imputation from 1k or 2k to 10k. The accuracy of imputation was decreased when the reference and validation populations were more distant.