Breeds with small active populations may lack the scale to undertake accurate genomic selection within the breed itself. This study, performed within Breed4Food in collaboration with partners in Melbourne, Australia, showed that simple pooling of information across breeds provides the highest accuracy of multi-breed genomic selection.
Multi-breed genomic prediction
So far, limited benefits have been observed from combining information on multiple breeds in genomic evaluations. In this study we aimed to increase the accuracy of genomic selection in a small breed (in this case Jersey) by using information of a large breed (in this case Holstein Friesian).
The highest accuracy was obtained by a model that simply analysed information of Holsteins and Jerseys without considering the differences between breeds. A sophisticated Bayesian model that was able to model more subtle differences between breeds, actually resulted in slightly lower accuracies.
Considering that genes with large effect are the main drivers of genomic prediction across breeds, we hypothesized that the pooling strategy’s assumption that gene effects are the same across breeds is especially appropriate for genes of large effect, while it is not detrimental for the estimation of genes of small effect.