Combining genomic and classical information in national BLUP evaluation to reduce bias due to genomic pre-selection

The existence of a genomic pre-selection step which is ignored in classical national evaluations leads to biased results: pre-selected bulls are underestimated, their reliabilities are overestimated. This is because the mendelian sampling term of the pre-selected animals is incorrectly assumed to have an expected value of 0. A simple method is proposed to include genomic information into national evaluations: genomic breeding values are transformed into genomic equivalent daughter performances, which are added to real observations in national evaluations. A simulation study shows that the bias of estimated breeding values disappears, that their reliabilities are improved but are still overestimated, and the mean squared error of prediction decreases. A crucial assumption to get these encouraging results is that all information on culled bulls are available.