Genomic estimated breeding values on health and behavior traits to inform breeding and management of working dogs – Presented by Heather Huson

Genomic estimated breeding values (gEBVs) are statistically calculated numbers that estimate a dog’s potential for health, conformation, and behavior traits based on their DNA profile. They allow for more accurate selection and refined management of working dogs which increases the rate of change achievable in selection programs to improve overall working dog success rates.  Modern quantitative and statistical genetic methods allow the prediction of breeding values by correlating phenotypic data on health, conformation, and performance traits with genomic markers, specifically single nucleotide polymorphisms (SNPs). These correlations are used to generate predictive algorithms based on a reference population with known genotypic and phenotypic information. These algorithms are then used to predict merit or phenotypic expression of individuals related to the reference population (same species or breed) by using the established relationship between genetic markers and phenotype.  To this end, a reference population continues to be developed for prediction algorithms of over 54 health and performance traits.  The well-established genomic prediction model, GBLUP, was identified as similarly accurate and more computationally efficient than four machine learning models assessed.  Trait prediction was explored in a large cohort of guide dogs and another cohort of personally owned competitive sled dogs allowing us to evaluate our ability to predict traits in a large, single-managed breeding colony using standardized protocols, along with client-owned dogs, scored by their trainers on a prescribed set of traits.  Heritability and prediction accuracy varied among traits and datasets comparing the sled and guide dogs and within the guide dogs, the different breeds.  Indexes, where multiple traits are combined into a single selection value (ie. health index), were also created, to improve ease of selection. In all, our goal is to build a tool for breeding selection and training management that will improve overall working dog success rates.  Here, we explain how genomic prediction works, its utility in working dog management, and the progress our research has made in working dog genomic prediction. We highlight successes, including prediction model optimization and predicted traits, as well as challenges and future directions.   

Coming soon.