Genomic prediction using qtl regions identified from regional heritability mapping for parasite resistance in australian sheep

Genomic selection uses genomic information to predict the breeding value of animals and can achieve higher prediction accuracy than pedigree based selection. This study aimed to compare the accuracy of genomic prediction using a medium-density (50k) SNP panel, as well as an imputed high-density (600k) SNP panel, with and without including pre-selected SNPs from QTL regions identified by regional heritability mapping (RHM). The proportion of variance explained by the preselected SNPs combined in a genomic relationship matrix (GRM) was considerably smaller than that explained by all SNPs from the 600k panel (25% of the genomic heritability). To obtain a better estimate of the variance explained by the pre-selected SNPs, both GRMs from the pre-selected SNPs ( GRMs) and their complementary SNPs from the 600k panel ( GRMc) were fitted in a single model. The total heritability explained by both GRMs and GRMc when fitted together was similar to the heritability explained by fitting all SNPs in a single GRM. The GRMs explained a smaller proportion (18%) of the total heritability, whereas the GRMc explained 82%. Fitting either the 50k or the 600k SNP panels resulted in similar prediction accuracy for parasite resistance (~0.37). However, when both GRMs and GRMc were fitted together in the prediction model, genomic accuracy was increased by 10%. These results indicate that accuracy of genomic prediction can be improved by including QTL information explicitly in the prediction models.