8.15.3 The size of the reference population
- Tessa Jacobs
- Kor Oldenbroek
In the figure above you see the increase in accuracy with increasing size of the reference population for 4 different heritabilities, assuming a q of 500, assuming that the gEBV are estimated using only genomic information. The upper line represents the trait with the highest heritability (0.9), and the lowest that with the lowest heritability (0.05). To achieve the same level of accuracy a decrease in the size of the heritability requires a large increase in the size of the reference population. For example, to achieve an accuracy of 0.6, 5630 are required for traits with heritability of 0.05, whereas only 320 are required for traits with heritability of 0.90. This illustrates that, even though genomic selection is a very nice tool, it is not feasible for small populations, and especially not for traits with a low heritability. As a potential solution, populations (studbooks) could combine forces in composing a reference population, so that they can share the costs, but also the benefits. This is current practice in dairy cattle breeding, where a number of international breeding organisations share a reference population. There is no example that reference populations of different breeds are combined. Theoretical this seems only to be effective when the density of SNP's is very high.
Thus: the size of the reference population can be a imiting factor for achieving accurate gEBV. Solution can be to combine forces across breeding associations.
Apart from being able to have an accurately estimated breeding value already at very young age, genomic selection is especially useful for selection on traits that are expensive to measure accurately because only a relatively limited number of phenotypes are sufficient to improve the EBV of many animals. Even though genomic selection does allow for selection without phenotypic information of the animal itself or on close relative, the accuracy of selection still is also determined by the accuracy of the recording of the phenotype. Especially in the reference population the recording of the phenotypes should be done as accurately as possible because those phenotypes are used for selection of the rest of the population through their associations with the SNP. Inaccurate phenotypes result in suboptimal estimations of association between SNP and phenotypes, and thus in suboptimal estimated gEBV. Remember, the effect of inaccurate observations is directly reflected in the size of the heritability, and that effect is illustrated in the figure above.
It is possible to combine genomic and conventional breeding value estimation. Phenotypic information on the animal or its relatives will add to the accuracy of the gEBV. Details on how that would add go beyond the scope of this course.