Ancestry-inclusive dog genomics challenges popular breed stereotypes

Morrill, K., Hekman, J., Li, X., McClure, J., Logan, B., Goodman, L., Gao, M., Dong, Y., Alonso, M., Carmichael, E. and Snyder-Mackler, N., 2022. Ancestry-inclusive dog genomics challenges popular breed stereotypes. Science376(6592), p.eabk0639.

One sentence summary from the paper: 

“Pairing owner surveys and genetics in pet dogs challenges breed stereotypes, and finds new behavior-associated loci.”​

This study is included because it breaks new ground in both methodology and findings in the study of the relationship between genetics and behavior in domestic dogs.

 Among the authors’ most important findings are that:

Only 9% of the cumulative variation among dogs on the 8 behavioral factors identified can be attributed to breed.

So little (almost none) of the variability on the agonistic threshold factor (behaviors commonly labeled “aggression”) can be attributed to genetics at all that it is of no utility in predicting behavior in individual dogs.

The majority of mixed breed dogs have more than 4 different breeds in their ancestry.

Morphology varies too extensively among mutts and people typically place too much weight on associations between a few traits and specific breeds to make visual identification reliable.

The dog breed most commonly appearing in the ancestry of mutts is the American Pit Bull Terrier, which appears in the ancestry of 10% of mixed breed dogs.

This research is particularly likely to have a strong impact on public policies regarding dogs in US society because the authors found that with regard to behavior that is commonly described as “aggression,” breed and even genetics in general, had virtually no influence, and that breed is not particularly predictive of the behavior of any individual dog.

This is the first research that compares genetic analysis with behavioral data collected on the same sample of dogs. Moreover, the sample was composed of both pedigreed and mixed breed individuals, the latter labeled “mutts” by the authors. Earlier research has either worked on the assumption that behavioral data and genetic data could be compared on different cohorts of dogs of the same breed to make generalizations regarding breed characteristics (MacLean et al 2019) or that breed club descriptions of a breed could serve as surrogates for behavioral data and then correlated with genetic data (Shan, et al, 2021). Both techniques beg the question they address since their underlying assumption that behavioral traits were largely consistent among members of particular breeds was also part of their findings. The results of such studies could only yield findings that were at best speculative, which is why we have not included them in this research library. Thus Morrill’s findings must, by virtue of their superior sample design, carry more credibility than any of the prior work on the role of genetics in canine behavior.

Moreover, the sample size for the behavior data and genetic sequencing data in Morrill et al is much larger (18,385 and 2,155 respectively) than anything attempted before. The results also suggest a high level of accuracy in the breed identification of the dogs in the sample. The risk of owner breed bias affecting their responses as a possible confound was mitigated (although it must still be acknowledged as a possibility of unknown extent) through the sample size, the inclusion of breed or other bias neutral questions, and the analysis of the behavior of a large cohort of mixed breed dogs.  All of this further establishes this study as the current gold standard in canine behavioral genetics.

Methods: Survey questions (117 in all) for the open-sourced survey sample were primarily drawn from the Dog Personality Questionnaire, an instrument for which both reliability and predictive validity have been assessed (Jones, 2008). Factor analysis was then applied to the responses which clustered around eight personality trait style continua: human sociability, arousal level, toy-directed motor patterns, biddability, agonistic threshold, dog sociability, environmental engagement, and proximity seeking.  The breed identification algorithm used more markers than the commercially used ones and had a high level of agreement with owner reports that their dogs were purebreds, either by pedigree (98.7% agreement) or by reporting only one breed on the survey (85.8% agreement), suggesting a high level of accuracy on the breed identification. The strong correlation between dogs thus “confirmed” as purebreds in the sample and the AKC registration percentages by breed suggest a sample likely to be reasonably representative of the general canine population. The proportion of purebred (49%) to mixed breed (51%) dogs also suggests a representative sample in its agreement with the most widely accepted estimates of the US pet dog population.

Findings: The effect of breed on where an individual dog scored on these 8 factors overall was found to be very low at 9%, while the effect of genetics as a whole (unrelated to breed) was 25%. The effect of genetics in general and breed in particular varied from factor to factor.  The effect was so low for agonistic threshold, “how easily dog is provoked by frightening, uncomfortable, or annoying stimulus” (what is often labeled “aggression”) that this can be described for practical purposes as no effect at all. The effect of breed on behavior was highest for biddability, “how readily dog responds to human direction, especially in the context of training,” but can still only be described as modest. Figure 1 from the paper reproduced below summarizes the  findings on the 8 behavioral factors identified by the authors.

This low correlation between breed and behavior was in stark contrast to the very high correlation between breed and appearance where it became very predictive, but only among purebred dogs. This confirms earlier research regarding the unreliability of identifying mixed breed dogs by their appearance. The study of mutts in this research confirmed that people do poorly on such labeling attempts and that they tend to rely on a few physical traits to guide their guesses with generally inaccurate results. Figure 3 H reproduced below shows the range of difference in appearance of dogs, all of whom have at least 25% American Pit Bull Terrier ancestry.

The study also confirms earlier research which used a sample of dogs living in shelters (Gunter et al, 2018) that most mutts are mixtures of several breeds at least: 66% according to Morrill et. al., and that mixes of only 2 breeds are unusual (17%), and that the most common breed signature among mutts is the American Pit Bull Terrier, appearing in 10% of the mutts’ ancestry.