Genomic selection is revolutionizing breeding programs worldwide. It requires establishing a reference population to estimate the so-called SNP-effects used to predict genomic breeding values. For breeding companies with multiple lines or breeds, an important question is whether it is beneficial to use a reference population across those lines or breeds, in order to reduce implementation costs.
Whether the accuracy of genomic selection is increased by using information of multiple lines, depends on the “genetic distance” between those lines; closely related lines may partly have the same genetic architecture and therefore harbour relevant information for the other line. This potential benefit was investigated using data from Hendrix Genetics using data from three lines of layer hens, including two brown lines that have been separated for at least 25 years, and an unrelated line of white layers.
The results showed that information of an unrelated line at best does not add to the accuracy of genomic selection, and may in fact reduce the accuracy. The two closely-related lines benefitted slightly from combining both lines in one reference population. It was shown that for this particular data, the reference population should be ten times larger when it comprises a closely-related line rather than the line itself, to achieve the same accuracy of genomic selection. In addition, from the closely related line, only the 20% closest related animals helped to improve the accuracy in the other line, while the remaining 80% in fact led to a reduction in accuracy.
Based on these results, it is concluded that genomic selection across lines or breeds is only beneficial for closely related lines or breeds, and when the additional breed or line contributes considerably larger number of animals to the reference population.
Future research involves the prediction of breeding values for purebred selection candidates, derived from crossbred performance.
For more information, please contact Mario.Calus@wur.nl