Heritability and Genome-Wide Association Study of Dog Behavioral Phenotypes in a Commercial Breeding Cohort

Abstract

Background: Canine behavior plays an important role in the success of the human-dog relationship and the dog's overall welfare, making selection for behavior a vital part of any breeding program. While behaviors are complex traits determined by gene × environment interactions, genetic selection for desirable behavioral phenotypes remains possible. Methods: No genomic association studies of dog behavior to date have been reported on a commercial breeding (CB) cohort; therefore, we utilized dogs from these facilities (n = 615 dogs). Behavioral testing followed previously validated protocols, resulting in three phenotypes/variables [social fear (SF), non-social fear (NSF), and startle response (SR)]. Dogs were genotyped on the 710 K Affymetrix Axiom CanineHD SNP array. Results: Inbreeding coefficients indicated that dogs from CB facilities are statistically less inbred than dogs originating from other breeding sources. Heritability estimates for behavioral phenotypes ranged from 0.042 ± 0.045 to 0.354 ± 0.111. A genome-wide association analysis identified genetic loci associated with SF, NSF, and SR; genes near many of these loci have been previously associated with behavioral phenotypes in other populations of dogs. Finally, genetic risk scores demonstrated differences between dogs that were more or less fearful in response to test stimuli, suggesting that these behaviors could be subjected to genetic improvement. Conclusions: This study confirms several canine genetic behavioral loci identified in previous studies. It also demonstrates that inbreeding coefficients of dogs in CB facilities are typically lower than those in dogs originating from other breeding sources. SF and NSF were more heritable than SR. Risk allele and weighted risk scores suggest that fearful behaviors could be subjected to genetic improvement.

Keywords: genetic risk score; inbreeding coefficient; non-social fear; social fear; startle response.

Figure 1

Manhattan plot and QQ plot for social fear in a CB cohort. Red and blue lines in the Manhattan plot indicate the Wald test p-values of 3.54 × 10⁻⁶ (suggestive threshold) and 4.00 × 10⁻⁵, respectively. The genomic inflation factor (lambda) for the QQ plot is 1.016.

Figure 2

Manhattan plot and QQ plot for non-social fear in a CB cohort. Red and blue lines in the Manhattan plot indicate the Wald test p-values of 3.54 × 10⁻⁶ (suggestive threshold) and 4.00 × 10⁻⁵, respectively. The genomic inflation factor (lambda) for the QQ plot is 1.015.

Figure 3

Manhattan plot and QQ plot for startle response in a cohort of dogs from CB facilities. Red and blue lines in the Manhattan plot indicate the Wald test p-values of 3.54 × 10⁻⁶ (suggestive threshold) and 4.00 × 10⁻⁵, respectively. The genomic inflation factor (lambda) for the QQ plot is 1.037.

Figure 4

Density plots comparing risk allele counts (cGRS) and weighted genetic risk scores (wGRS) between more fearful and less fearful dogs in a CB cohort: (a–c) density plots of cGRS (simple risk allele counts) for SF, NSF, and SR, respectively; and (d–f) density plots of wGRS (weighted risk scores) for SF, NSF, and SR, respectively. Note that the x- and y-axes are not identical for each panel.