Presenter Ros Craddock
Authors Rosalind Craddock (1), Mateja Janes (1), Cathryn Mellersh (2), Joanna Ilska (3), Pamela Wiener (1), Steph Smith (4), Gregor Gorjanc (1)
Affiliations 1. The Roslin Institute, University of Edinburgh 2. The Canine Genetics Centre, University of Cambridge 3. The Kennel Club, United Kingdom 4. Global Agriculture and Food Systems, University of Edinburgh
Presentation Type Talk
Abstract
Most monogenic diseases in dogs are recessive. Where possible, dog owners use genetic tests to identify dogs that carry specific disease-causing alleles and the Kennel Club, UK (KC) records the result against the dog’s individual pedigree as either “clear” (homozygous wild), “carrier” (heterozygous), or “affected” (homozygous mutant). These results, along with “hereditary status” (an assigned offspring genotype where both parents are homozygous for either the normal or disease-causing allele), are used to monitor changes in allele frequencies over time. However, this only considers a proportion of the pedigree population since not all dogs have genetic test information.
This work used probabilistic inference for pedigree-based tracking of alleles for two monogenic recessive eye diseases in five KC-registered breeds. With this method we (1) estimated genotype probabilities for all individuals in the pedigree using available genetic test information and family relationships, (2) estimated allele frequencies by year of birth, and (3) examined the quality of the estimated genotype probabilities using leave-one-out cross-validation and simulations.
Results show that disease-causing allele frequencies were underestimated after the introduction of a genetic test when considering only the proportion of the pedigree with either genetic test information or hereditary status. With the estimated genotype probabilities, the allele frequencies were up to 0.092 higher per year after genetic test introduction. This reflects probabilistic inference’s ability to take into account the entire pedigree, thereby diluting the reporting bias toward clear dogs.
The quality of the estimated genotype probabilities varied across breeds, largely depending on population allele frequency, with accuracies ranging from moderate (0.47) to high (0.67) as measured by Pearson correlation. Simulations indicated a tendency of the probabilistic inference to underestimate the disease-causing allele frequency prior to the 2000s due to only recent availability of genetic tests.
Overall, this study shows improved population monitoring of recessive disease-causing alleles by estimating genotype probabilities for all pedigree individuals, aiding better breeding decisions to reduce disease incidence in KC-registered dogs. Nonetheless, estimates were limited by the uptake of genetic testing, highlighting the need for continued genetic testing of dogs.