ICCFGG 2022
#27 Estimating working dog breeding values for binary health traits J.A. Thorsrud 1 , K.M. Evans 2 , D.M. Holle 2 , H.J. Huson 1 hjh3@cornell.edu, jat325@cornell.edu 1 Department of Animal Sciences, Cornell University College of Agriculture and Life Sciences, Ithaca, NY, USA; 2 The Seeing Eye Inc, Morristown, NJ, USA Breeding values are an effective tool in breeding colony management. By quantifying the genetic merit of an individual for a trait or series of traits, they can help inform mating decisions. Breeding values can be used to either select for positive traits or against negative ones. There are a variety of approaches which can be used in breeding value construction ranging from pedigree based to genomic with different models to fit the data. To aide in understanding model performance given different parameters, 37 health traits were examined with pedigree based BLUP, genomic BLUP, and four machine learning algorithms (support vector machine, random forest, extreme gradient boosting, and neural network). Preliminary data includes 1,097 dogs with 322 German Shepherds, 578 Labrador Retrievers, 127 Golden Retrievers, and 70 Golden Retriever Labrador Retriever crosses. The health traits vary in the estimated heritability, from <0.01 to 0.62. Preliminary results indicate that as the heritability decreases, the pedigree-based approach begins to underperform the other genomic approaches which matches simulated results that predict machine learning techniques may be able to surpass both BLUP methods for complex traits. This information can aide in both future creation of breeding values through optimal strategies and can also directly impact breeding practices by expanding our understanding of the genetics behind diseases. Further work can focus on the creation of indices for different trait types such as dental or dermatologic conditions to simplify selection across many traits. Going forward, breeding values can improve the health and performance of guide dogs. #28 Revealing risk factors for Canine Atopic Dermatitis using Bayesian model and selection signature analyses Tengvall, K. 1 *, Sundström, E. 1 , Wang, C. 1 , Bergvall, K. 2 , Wallerman, O. 1 , Pederson, E. 1 , Karlsson, Å. 1 , Harvey, N.D. 3 , Blott, S.C. 3 , Olby, N. 4 , Olivry, T. 5 , Brander, G. 1,6 , Roosje, P. 7 , Leeb, T. 8 , Hedhammar, Å. 2 , Andersson, G. 9 , Lindblad-Toh, K. 1,6 katarina.tengvall@imbim.uu.se 1 Science for Life Laboratory, Department of Medical Biochemistry and Microbiology, Uppsala University, Uppsala, Sweden, 2 Department of Clinical Sciences, Swedish University of Agricultural Sciences, Up- psala, Sweden, 3 School of Veterinary Medicine and Science, University of Nottingham, Sutton Bonington Campus, Leicestershire, UK, 4 Department of Clinical Sciences, North Carolina State University, Raleigh, NC, US, 5 Department of Clinical Sciences, North Carolina State University College of Veterinary Medi- cine, Raleigh, NC, USA, 6 Broad Institute of MIT and Harvard, Cambridge, MA, USA 7 Division of Clinical Dermatology, Department of Clinical Veterinary Medicine, Vetsuisse Faculty, University of Bern, Bern, Switzerland, 8 Institute of Genetics, Vetsuisse Faculty, University of Bern, Bern, Switzerland, 9 Depart- ment of Animal Breeding and Genetics, Swedish University of Agricultural Sciences, Uppsala, Sweden Canine atopic dermatitis (AD) is an inflammatory skin disease showing clinical similarities to human AD. Several dog breeds are at increased risk for developing this disease. To identify genetic risk factors for canine AD, we applied a Bayesian model adapted for mapping complex traits
46
Made with FlippingBook - Online magazine maker