Supporting the implementation of genomic selection in a guide dogs' population using simulation

For service dog populations, applying genomic selection would enable more efficient breeding for complex traits such as health, welfare, and trainability. However, the transition from phenotypic to genomic selection requires genomic information. Different data collection scenarios can be envisioned based on the number of individuals, the number of markers, and the genotyping technology. The aim of this study was to identify the optimal scenario for data collection to implement genomic selection and investigate complex trait architecture with whole genome sequence (WGS) information. To do so, we simulated the UK Guide Dogs' population based on their pedigree and existing high-coverage WGS data with AlphaSimR, and then phased and imputed with AlphaPeel for various scenarios. The existing pedigree was extended with additional generations to evaluate scenarios' outcomes in the future. The scenarios considered were composed of diverse genotyping densities and sequencing coverages for the puppies. All scenarios were compared using individual imputation accuracy against the true simulated WGS. Low-pass sequencing scenarios (0.5 to 2X depth) achieved accuracy of 0.986 to 0.998. SNP array genotyping (25K to 710K markers) was inferior, with an accuracy of 0.560 to 0.732. For the UK Guide Dogs, the simulation revealed low-pass sequencing as the best strategy for obtaining WGS information for downstream use in genomic selection and analysis of complex traits.