Within-breed stratification for across-breed reference population

In endangered breeds, it is difficult to have large reference populations needed for accurate genomic predictions. To solve this issue, one solution can be to build an across-breed reference population. In that case, candidate animals should however be genetically close enough to the reference population. Therefore, special attention must be paid to the animals from other breeds that are selected to be part of the across-breed reference population. In our study, we explored different measures of genetic diversity, namely principal component analysis (PCA) and ADMIXTURE clustering based on genotypes, to detect stratification within Dutch red-pied cattle breeds with the aim to build an across-breed reference population for two genomically related endangered red-pied cattle breeds from Belgium and Luxembourg. The PCA analysis showed that the Dutch, Belgian and Luxembourgish breeds were part of a genomic continuum. We detected stratification within the Dutch breeds and consequently also genomically related animals to the Belgian and Luxembourgish breeds. Although empirical validation of across-breed genomic prediction is needed to confirm which reference population yields the best predictions, our results do suggest that stratification within breeds should be considered when construction across-breed reference populations.