Across breed multi-trait random regression genomic predictions in the Nordic Red dairy cattle

Abstract

Current genomic prediction equations, when carried out in multiple populations with admixed structures ignore structure and assume these populations are uniform. The observed reliabilities of direct genomic breeding values (DGV) for unproven bulls in these populations so far have been low. The current study evaluated reliabilities of DGV in selection candidates using multi-trait random regression model which account for interactions between marker effects and breed of origin in the admixed Nordic Red dairy cattle. Our breed-specific model used breed proportions (BP) as random predictors and deregressed proofs of estimated breeding values (DRP) as response variables weighted by approximated reliability of DRP. Reliabilities were explored as squared correlation between DRP and DGV, weighted by the mean reliability of DRP. Estimated reliabilities were low for milk (0.32) and protein (0.32) and slightly higher (0.42) for fat. Observed reliabilities were similar to those estimated assuming homogenous structure. The Nordic Red cattle is admixed but closely related, thus, the model under investigation may have been unable to differentiate additive genetic effects by breed of origin with a medium dense marker data.