FISHBOOST results: Genomic selection using pooled DNA sequencing
Anna Kristina Sonesson (Nofima)
Great reductions in costs of genotyping
Genomic selection is a new powerful method to select the best individuals based on DNA data. It estimates the effects of genome-wide dense genetic markers (SNPs) and uses these estimates to select the parents for the next generation. This selection method is today used in Atlantic salmon breeding programs and its accuracies are tested in FISHBOOST on all the species of the project.
In aquaculture, the reference population (the group of fish where the effect of each SNP is estimated) is typically made up of sibs of the selection candidates. However, these populations often contain thousands of fish, which results in high genotyping costs even if the genotyping costs per fish have decreased dramatically over the last years. Hence, ways to further reduce the genotyping costs are being sought.
In this study of the FISHBOOST project, we tested the accuracy of selection when DNA of individuals with extreme phenotypes had been pooled and genotyped either with whole-genome sequence data or with an SNP chip. We compared this accuracy to the accuracy of selection when the reference population is made up of individually genotyped fish.
Genomic selection is a new powerful method to select the best individuals based on DNA data. It estimates the effects of genome-wide dense genetic markers (SNPs) and uses these estimates to select the parents for the next generation. This selection method is today used in Atlantic salmon breeding programs and its accuracies are tested in FISHBOOST on all the species of the project.
In aquaculture, the reference population (the group of fish where the effect of each SNP is estimated) is typically made up of sibs of the selection candidates. However, these populations often contain thousands of fish, which results in high genotyping costs even if the genotyping costs per fish have decreased dramatically over the last years. Hence, ways to further reduce the genotyping costs are being sought.
In this study of the FISHBOOST project, we tested the accuracy of selection when DNA of individuals with extreme phenotypes had been pooled and genotyped either with whole-genome sequence data or with an SNP chip. We compared this accuracy to the accuracy of selection when the reference population is made up of individually genotyped fish.
Figure 1. Typical design of aquaculture genomic selection schemes
The typical design for genomic selection schemes in aquaculture species is given in Figure 1. Families are split into two groups: one group of selection candidates that get genotypic records, and one group of test individuals that get genotypic and phenotypic records, e.g. using a disease challenge test. In the latter group, which we call the reference population, the effect of each SNP marker is estimated. These effects are then used to calculate the genomic breeding values in the genotyped group of selection candidates, and to select parents for the next generation.
The reference population is often large resulting in high genotyping costs, because of the large number of families and individuals per family being genotyped in this group. In this study of the FISHBOOST project, we tested the accuracy of selection when DNA of individuals with extreme phenotypes had been pooled and genotyped either with whole-genome sequence data or with an SNP chip. We compared this accuracy to the accuracy of selection when the reference population is made up of individually genotyped fish.
The reference population originated from the SalmoBreed elite population of Atlantic salmon. It had been tested for pancreas disease (PD), an important viral disease in salmon production, in a controlled challenge test. Individual fish that died during the test were chosen for two mortality (M) pools and similarly, individual fish that survived the test were chosen for two survival (S) pools. Each pool consisted of ca. 200 individuals from 30 families. Equal amount of DNA per fish was collected to form the pools. Three genotyping strategies were compared:
1. Individual genotyping with a 50kSNP chip.
2. Genotyping of 4 pools with a 50kSNP chip
3. Whole genome sequencing of the 4 pools (40x coverage).
Genomic EBVs (using the GBLUP methodology) were calculated for the fish with individual genotypes. The GEBV were correlated to the known individual phenotypes, and accuracy of the genomic breeding values was inferred by scaling the correlation with the square root of the heritability of the trait.
The results of this study show that in this dataset, there were small differences in accuracy of selection when the reference population had been genotyped individually or with the whole genome sequencing of pools of individuals with extreme phenotypes.
In conclusion, genotyping pools of DNA of fish with extreme phenotypes seems to be a cost-effective alternative to individual genotyping of the whole test population in aquaculture genomic selection schemes. In further studies, we will compare the different pooling strategies, size of pools, and other traits.
This study was performed by: Anna K. Sonesson (Nofima), M. L. Aslam (Nofima), M. Baranski (Nofima), T.H.E. Meuwissen (NMBU), G. Malacrida (BMR Genomics) and H. Bakke (SalmoBreed).
