Jesse Hoff, Agrigenomics Product Manager - Apr 28, 2022

Using a Modern Approach to Genotyping to Challenge Breed Stereotypes in Dogs

Do you rate your golden retrievers high on the scale for friendliness to strangers because they’ve been bred to demonstrate that trait or because you house an internal bias that tells you golden retrievers are friendly? These are precisely the types of questions that researchers are answering by utilizing modern, genome-wide sequencing information to better understand the association between genes and traits in dogs. In a new paper published in Science, Morrill et al. applied low-pass sequencing in a diverse population of dogs to power GWAS studies, drawing clear conclusions about the heritability of both the physical traits we know separate dog breeds and the more complicated behavioral traits like human sociability and agonistic behavior.

A better study design for a diverse population

The researchers aimed to explore the complicated relationship between breed and behavior in dogs by asking owners of the dogs a wide array of questions about physical traits and behavioral tendencies. They then looked for trends and connections in the genetic data. Unfortunately, survey data of this sort comes with a set of biases. Humans are prone to attribute many stereotypes to dogs based on their breed. We’ve all heard the stereotypes that golden retrievers are family dogs and pit bulls are dog-aggressive dogs. Therefore, to mitigate these biases, the researchers used many mixed breed “mutt” dogs. By utilizing dogs that were unable to be categorized into specific breeds by humans, answers to questions about a dog’s tendencies could better reflect an unbiased response. This posed another issue for analyzing the genetic data. While several genetic studies have been conducted in dogs, most of them are focused on purebred individuals and using array data. Because mutts have shorter runs of homozygosity and linkage disequilibrium that decays more rapidly than purebred dogs, genotyping arrays designed with sufficient marker density for purebred dog studies miss much of the genetic variation in mutts. To solve this problem, the researchers used Gencove’s low-pass sequencing and analysis platform to genotype 1,715 dogs for more than 32 million SNPs. They then performed several GWAS studies to look at a variety of physical and behavioral traits and the influence of genetics and breeding on those traits.

Figure from Morill, et al. 2022 demonstrating the behavioral traits being explored and the statistical tests applied to help understand the influence of various genetic factors.

The complicated and unexpected results

It turns out that behavioral factors are quite variable within breeds, which suggests that while a dog’s breed may affect the likelihood of a particular behavior to some extent, it is not a very strong predictor. The analysis from this research reveals that the breed of a dog explains just 9% of the behavior variation between individuals. From a genetics perspective, this number is quite high - many traits of interest that geneticists study are much more complicated than a single locus driving heritability. However, from a human perspective, where we are told all our lives that dogs have breed-specific personality traits, it feels like there should be more predictability in behavior based on breed.

And for some behavioral traits, the heritability is quite high. However, the authors note that these tend to be for motor pattern related behaviors rather than what we think of as personality related behaviors. For example, the tendency to howl was much more likely in beagles, bloodhounds, and Siberian huskies. Alternatively, they found no evidence of a higher likelihood of aggressive behavior (termed agonistic threshold) in what humans have typically thought of as the bully breeds.

Figure from Morrill, et al. 2022 showing that physical traits have a much higher correlation to breed compared with behavioral traits

So, what does this mean for those of us who were planning to use breed as a way to choose the perfect furry companion for our family? It means we have to get comfortable with the behavioral variations within breeds. While we can be confident that choosing a great dane will get us a large, short-haired fuzzy friend, we cannot be sure that this same dog will love kids or meeting strangers. Just like humans, each dog is an individual with particular tastes and tendencies. Thankfully, the authors developed an interactive dashboard to illustrate the value of using breed to predict the behavior of an individual dog. Leveraging the data they collected as part of the Darwin’s Ark project to fuel their studies, users can select one or a combination of characteristics to dynamically show the frequency of the traits across 23 breeds and mutts to help you predict as best as possible, what your new furry friend will be like.


This study highlights the power of low-pass sequencing and genotyping for non-inbred populations without the need to build a new microarray. The authors were able to include complex mixes in a behavioral study context to mitigate the influence of breed stereotypes, and enabled them to draw clear conclusions about which traits can be predicted by breed.

Furthermore, by utilizing a diverse cohort of dogs with high density genotyping data, the associated regions are much narrower than those discovered using intrabreed GWAS with less dense marker sets, increasing the resolution for identifying causal variants. This, in turn, can help define more predictive markers in the future. Thankfully, the Gencove platform is easy to update with new markers and reference genomes, providing a platform that will support their research for years to come.


If you’d like to learn more about low-pass sequencing and the Gencove analysis platform, please contact us.