Impact of adding cows to the current EuroGenomics bull reference population on genomic prediction

Abstract

Genomic evaluation has been based on a reference population comprising only daughter-proven bulls in most countries until recently. With more and more cows routinely genotyped, adding non-selectively genotyped cows to current bull reference population may improve the accuracy of genomic evaluation. Using approximately 20,000 cows genotyped within the framework of a German research project Kuh_L, we wished to quantify the impact of adding cows to the current EuroGenomics bull reference population on genomic prediction. Four types of reference population: only cows from the research project (Cows), the current EuroGenomics reference bulls (Bulls), a mixed of the bulls and cows from the project (Mixed) and a mixed of the bulls and all cows (Mixed_All),  were compared in accuracy and bias of genomic prediction using phenotype and genotype data from April 2017 monthly genomic evaluation for German Holsteins. Analyzed traits included all routine national and MACE traits and novel health traits on claw health, udder health, reproduction diseases and metabolic disorders. Among all the four types of reference population, the scenario Cows had the lowest validation R² value due to the limited number of reference cows, this suggested a reference population composed of only cows being not yet accurate enough for routine genomic evaluation. We observed an increase in validation R² values for all the traits when the cows were added into the current bull reference population. No single trait was identified with a decrease in the accuracy of genomic prediction. For the novel traits accuracy of prediction was significantly lower than the regular ones, due to the much smaller size of reference population. The regression coefficient of validation bulls’ deregressed EBV on their candidate GEBV showed that under the scenarios Bulls, Mixed and Mixed_All genomic prediction remained unbiased, having the cows included in the bull reference population. In summary, adding the non-selectively genotyped cows to the current EuroGenomics bull reference population increased the accuracy of genomic prediction for all the regular traits. No bias of genomic prediction was seen when the reference cows were observed to the current EuroGenomics bull reference population.