Genomic selection methods in aquaculture breeding programmes
Theo Meuwissen (NMBU)
The showcase for genomic selection is dairy cattle, but aquaculture breeding programmes also yield substantial opportunities for genomic selection. Many traits of interest to fish breeders are not directly recorded on the candidates for selection for elite breeding, such as disease resistance traits and carcass traits. The recording of the former traits is performed in disease challenge tests on sibs of the candidates in order to avoid a disease outbreak amongst elite breeding animals, and the recording of the latter requires slaughtering of the fish. In addition, the large family sizes that are typical to fishbreeding schemes make it possible to set up reference populations within the families and apply the genomic selection technology within the families, i.e. the so-called within family genomic selection. Within families, the associations between genetic markers and genes are strong, so we do not need such high marker density as in traditional population-wide genomic selection, and the number of within family reference individuals can be relatively small. Theoretical developments show that within family genomic selection can reach an accuracy of 100% in fishbreeding because of the potentially large family sizes. This implies a substantial increase in accuracy of selection compared to the traditional family selection based fish breeding schemes, which reach a theoretical maximum accuracy of 70%. This is because e.g. a traditional challenge test for disease resistance reveal only which family has best disease resistance, whereas genomic selection also predicts which individual within a family has the best genes for disease resistance.
Within family and the more traditional population wide genomic selection were compared by UMB in computer simulation studies. The computer simulation studies showed that at very low marker densities, within family genomic selection yielded higher selection accuracies, but within family genomic selection was more sensitive to small family sizes. In analysis done by Nofima, UEDIN and UMB of common carp, European seabass, gilthead seabream and turbot, population wide genomic selection gave 1-2 percent point higher accuracy than within family genomic selection. The results also seemed to depend on whether there were large genes affecting the trait and markers that are associated to these genes, since in this situation population-wide genomic selection may successfully find these associations by combining data across families. Generally, both population-wide and within family genomic selection were clearly more accurate than traditional pedigree based selection (10-22 percent points).
DNA pooling strategies were investigated by Nofima where DNA of e.g. the survivors of a challenge test is pooled and similarly the DNA of the diseased fish is pooled. Next, the marker testing or sequencing of these DNA pools was performed instead of genotyping individual fish. This may reduce genotyping costs by a factor 100 or more, whilst the accuracy of genomic selection was still high, although not as high as when marker genotypes on individual was used. Moreover, INIA compared methods to estimate inbreeding and genetic diversity based on genomic data, as well as alternative mating strategies based on genetic (dis)similarities. In the absence of genomic data, associative mating was shown to increase genetic progress in fish breeding schemes.