P1026
What's Your Breed Puppy? Genomic Breed Definition for Dogs

Tamer A. Mansour , Department of Population Health and Reproduction, School of Veterinary Medicine, University of California Davis, Davis, CA
Joshua Stern , Department of Medicine and Epidemiology, School of Veterinary Medicine, University of California Davis, Davis, CA
Sara Thomasy , Department of Surgical and Radiological Sciences, School of Veterinary Medicine, University of California Davis, Davis, CA
Emily Brown , Department of Population Health and Reproduction, School of Veterinary Medicine, University of California Davis, Davis, CA
Noa Safra , Department of Population Health and Reproduction, School of Veterinary Medicine, University of California Davis, Davis, CA
Karen Vernau , Department of Surgical and Radiological Sciences, School of Veterinary Medicine, University of California Davis, Davis, CA
Jodi Westropp , Department of Medicine and Epidemiology, School of Veterinary Medicine, University of California Davis, Davis, CA
Claire Wade , School of Life and Environmental Sciences, Faculty of Veterinary Science, University of Sydney, Sydney NSW, Australia
Stan Marks , Department of Medicine and Epidemiology, School of Veterinary Medicine, University of California Davis, Davis, CA
Chris Murphy , Department of Surgical and Radiological Sciences, School of Veterinary Medicine, University of California Davis, Davis, CA
C. Titus Brown , Department of Population Health and Reproduction, School of Veterinary Medicine, University of California Davis, Davis, CA
Danika Bannasch , Department of Population Health and Reproduction, School of Veterinary Medicine, University of California Davis, Davis, CA
Artificial selection of dog breeds has introduced breed specific characteristics and specific disease predispositions. Many of those phenotypes can be explained by fixed monogenic alleles or enrichment for few large-effect variants in each breed because of small founder populations. Classical array based GWAS studies to identify these causative mutations is challenging. Whole genome sequencing (WGS) and subsequent investigation of candidate genes could represent a more powerful approach to identify breed-specific mutations that are responsible for breed predisposition to diseases.

In this study WGS of 100 dogs from 23 breeds enabled us to identify ~22 million high quality variants. Ten breeds that have at least 4 unrelated dogs were selected to identify their genomic profiles. These breeds are Labrador retriever, Toller, Golden retriever, Bulldog, Weimaraner, Whippet, Boxer, Newfoundland, Pug, and Rottweiler. Within each breed, pruned SNPs were used to estimate inbreeding coefficients (mean: 0.169, range: -0.005 to 0.289) and pairwise Identity by Descent (maximum PI-HAT < 0.07). SNPs and indels were used independently to create breed-specific genomic profiles by calculating the frequencies of alternative alleles of high quality bi-allelic variants. Hierarchal clustering of genome-wide allele frequencies was able to delineate the underlying genetic relatedness of tested breeds. PLINK software was used to perform genome wide association analysis. For indels to fit PLINK input, they were converted into fake SNPs by homemade scripts. A false discovery rate of 0.05 was used to select significantly associated variants, which were then further filtered to select those with more than 50% difference in the allelic frequency between the tested breed and all other dogs. On average, there are ~500k and 100k breed-specific SNPs and indels respectively. About half of these variants occur within 5kbs of the boundaries of the Ensembl annotated transcripts. According to VEP tool, each breed has 148 and 1360 variants with expected high and moderate effects on their gene functions respectively.