What’s my breed?
To gauge the accuracy and consistency of canine DNA breed identification laboratory analyses marketed to curious dog owners, we ran these six dogs through a series of tests. They are (1) Tag, (2) Finn, (3) Salem, (4) Solara, (5) Annie, (6) and Laika.
Divining the breeds that make up a mutt is amusing sport for most dog lovers, and veterinarians are no exception. So when the first DNA test for dog breeds hit the market in 2007, interest and curiosity were keen. Soon, so was skepticism.
On a message board of the Veterinary Information Network (VIN), an online community for the profession, veterinarians reported a variety of results in friends, clients' and their own dogs. Some were plausible and some were wacky.
On the wacky side was an 80-pound dog whose mother was known to be a Labrador retriever that tested as mostly miniature poodle plus Yorkshire terrier.
There was a pedigreed dachshund who turned up as predominantly Siberian husky, with a dash of dachshund and Ibizan hound.
There was a dog that looked classically Chihuahua that came back as "an extremely complex mixed-breed dog ... (with) distant traces of Afghan hound, Cavalier King Charles spaniel and toy fox terrier."
And so on.
Five years since companies introduced commercial dog breed identification genetic tests, veterinarians continue to wonder if the tests are valid. Hoping to find the answer, the VIN News Service put the tests to the test.
We submitted DNA samples on six dogs to compare the results and check for consistency. We knew the breed backgrounds of four of the dogs, so were able to determine whether the tests gave correct answers.
When we started the project, five tests were available. Four were variations of Wisdom Panel, which is owned by Mars Veterinary. The fifth test was Canine Heritage by Scidera, LLC.
What we found: Wisdom Panel was the superior competitor. Though not flawless, the test tended to give reasonable and usually consistent, if not necessarily enlightening, results. Veterinary geneticists we consulted also pointed to Wisdom Panel as the most scientifically reputable.
"Mars, I know, is constantly reviewing and analyzing and upgrading what they're doing," said Dr. Jerold Bell, a clinical associate professor of genetics at Tufts University Cummings School of Veterinary Medicine. "They're one of the few that have done the research and continue to do the research to make it as scientifically valid as possible."
In its early days, the test was nicknamed "Witless" by some skeptical veterinarians. Wisdom Panel's more dependable results today may be a product of improvements developed over the years, chiefly expansion of the company's breed database.
Breed genetics concept scientifically credible
The concept of determining a dog's breed background by analyzing its DNA is grounded in science. In a paper published in the journal Science in 2004, canine geneticist Elaine Ostrander and colleagues described a technique they developed for identifying dog breeds based on genetic markers. Ostrander, who currently works at the National Human Genome Research Institute, part of the National Institutes of Health, described the approach, along with her broader work on the canine genome, in an article published in 2007 by American Scientist.
The markers the researchers used are not genes themselves, but repeating sequences of DNA known as microsatellites. The commercial tests use a different kind of marker known as single-nucleotide polymorphisms — abbreviated SNPs and pronounced like "snips" — that are small mutations within the genome. Whether using microsatellites or SNPs, the markers, taken together, form signatures particular to each breed.
Because the identification technique is not based on genes, it does not specifically relate to physical or behavioral traits that characterize particular breeds. In other words, the technique doesn't recognize a bulldog by finding the genes that give it a snub nose, beefy head and squat stature.
Consequently, breeds that are vastly different in appearance might, by chance, have similar genetic signatures, which helps explain some head-scratching results. For example, portions of the signatures of Chihuahuas and some mastiff breeds are maddeningly similar, according to Dr. Angela Hughes, a veterinarian and geneticist at Mars Veterinary.
On the other hand, some breed combinations truly result in dogs that resemble other breeds entirely. Hughes recalled a case of a dog that looked like a black Labrador retriever that tested as a golden retriever mix, which made sense to her as a geneticist.
"You can lose that longer coat in one generation," she explained. "Goldens carry a black gene. They don't express it in their coat because the yellow gene blocks all black. They do express it in their nose, eye rims and the pads of their feet. But the golden is recessive. So if you breed a golden retriever to a dog that doesn't have the genes for yellow and long hair, you're likely to have a black dog with a short coat."
In a similar vein, test results suggesting pairings of large with small dogs may raise doubts but such combinations aren't impossible, Hughes said. She explained: "The larger female will lie down" for the mating.
Breeds have distinctive genetic signatures owing to the fact their members are genetic isolates — that is, bred from a limited population of dogs. The more unique the breed characteristics, the easier to identify a breed's members via their DNA. Explained Dr. Urs Giger, a veterinarian who heads the clinical program in medical genetics and pediatrics at the University of Pennsylvania (UPenn) School of Veterinary Medicine: "The more you're inbreeding ... that will clearly make the genes among the dogs more similar."
A mixed-breed dog whose parents or grandparents are purebred generally is easier to identify than a dog descended from generations of mixes.
Giger said improvements in DNA-based breed identification may contribute to the development of genetic tests for medical conditions. "The mixed-breed test is the first complex-trait test and thereby is showing the way (for) testing for other complex traits, like hip dysplasia, in the future," he said.
The science of dog-breed detection may be solid, but that doesn't mean any given DNA test is reliable. Its accuracy depends upon the quality of information upon which the analysis is based.
For example, when the Canine Heritage test debuted in early 2007 as the first such test on the consumer market, it detected 38 breeds. Results for any dog with a background outside of those 38 breeds would have been inconclusive.
Wisdom Panel launched in fall 2007 with 134 breeds. Now Canine Heritage is up to 120 breeds, while Wisdom Panel lists 203.
In setting out to test the tests, the VIN News Service surveyed the market to determine how many competing tests were available in the United States. Looking at website domain names, there appeared to be multiple options. On closer examination, we discovered the websites led to only two test brands. They have sites under their own names — www.canineheritage.com and www.wisdompanel.com — plus others. For example, www.dog-dna.com peddles the Canine Heritage test. The site www.whatsmydog.com and www.happydogdna.com promote Wisdom Panel.
A third laboratory, BioPet, whose dog breed DNA tests were sold under the brand names BioPet and PetSafe, left the market last year in the face of a patent infringement suit pressed by Mars and the patent owners.
At the outset, Mars obtained an exclusive license from Fred Hutchinson Cancer Research Center and Argus Genetics, LLC, on their patent-pending process of discerning the breeds in a mixed-breed dog through DNA analysis. (Ostrander and team worked at Fred Hutchinson at the time they invented the process.) After the U.S. patent was granted in 2010, Mars took BioPet, PetSafe and the owner of Canine Heritage to court.
Soon, Mars' dominance will be complete: Canine Heritage is leaving the market. Mars is acquiring the test, according to Canine Heritage spokeswoman Anna Rogatkin. The Canine Heritage website states that the laboratory will coordinate with Mars to continue processing samples submitted within 90 days, starting June 19.
At the time our project began, we still had two different laboratories to compare. We looked at Canine Heritage's single test aimed at mixed-breed dogs and Wisdom Panel's four tests.
The original Wisdom Panel test is a "professional" version that must be purchased through a veterinarian and requires a veterinarian's services in taking and submitting a blood sample. The resulting report is more detailed and includes information for the veterinarian about medical conditions to which the identified breeds are prone.
In 2009, Wisdom Panel introduced a mixed-breed test like Canine Heritage's. Both are do-it-yourself tests in that the dog owner or handler collects the DNA via a "cheek swab" using a small brush run along the inside of the dog's cheek. Test kits come with brushes and return envelopes.
In 2011, Wisdom Panel introduced two new variations. One is for purebred dogs, the other for designer dogs, which are deliberate crosses of two purebred dogs along the lines of Labradoodles (Labrador retriever-poodle mix) and puggles (pug-beagle mix). The purebred and designer dog tests use cheek swab samples.
In terms of the quality of DNA extracted from cheek swabs versus blood, there is no difference, according to Bell, the veterinary geneticist at Tufts.
"We do cheek-swabbing all the time. I think the validity of cheek swabs is pretty high," Bell said. "It's pretty hard to mess up a cheek swab. At most, they don't get DNA. It's not that they get the wrong DNA or DNA from what the dog just ate."
Tag and Finn: pilot dogs
We started with a pilot phase, putting two dogs through all five tests each. The dogs were Tag and Finn. Tag is a purebred standard poodle registered with the American Kennel Club (AKC). Finn is a short-haired black-and-white mutt who was picked up as a stray. He was billed as border collie-spaniel mix but that was a guess on the part of the rescue group that put him up for adoption.
Tag came back as all-poodle on the Canine Heritage test and on the Wisdom Panel professional and purebred tests. But results from the mixed-breed test detected breeds in addition to poodle — namely, Labrador retriever, Pembroke Welsh corgi, shih tzu, Tibetan spaniel and miniature poodle.
Wisdom Panel Professional is designed to pick out the breeds in a mutt just like the mixed-breed do-it-yourself kit. However, we found in Tag's case that the professional test brought more detailed scrutiny. As the letter to his owner explained:
"When we analysed (this dog's) DNA data with the Wisdom Panel Professional computer analysis, the ancestry chart only appeared to contain Standard Poodle. To confirm this, we performed a complimentary Wisdom Panel Purebred analysis, and (the) results are indeed consistent with the purebred Standard Poodle samples in the Wisdom Panel database."
Hughes, the veterinary genetics research manager for Wisdom Panel, reviewed Tag's results afterward with the VIN News Service. She explained that customers who pay the premium for the professional product get more information, along with additional breed tests if the results warrant.
Tag's designer test results differed from his purebred and professional test results. When submitting the sample, the customer must specify what two breeds the dog is supposed to be. Tag's owner specified poodle and Labrador retriever. The results correctly said he wasn't equal parts poodle and Lab, but it did show some Lab in his background, along with an unspecified mix.
Hughes said the next-best breed match after poodle in Tag's genetic signature happened to be Labrador. "The Labrador comes out very, very low, at the great-grandparent level," she said. "Our test protocols would therefore not consider this a true Labradoodle."
She allowed that the Labrador finding was low enough that it could be a false positive.
Hughes noted that Wisdom Panel had an accuracy rate of greater than 90 percent in more than 200 tests of mixed-breed dogs with known ancestry. The validation-test dogs were offspring of purebred parents of differing breeds. Hughes acknowledged that pinpointing more complex mixes than first-generation crosses is difficult, but added, "We believe that we are good at this, as well."
Our second pilot test dog, Finn, a mix of unknown lineage, came up with consistent Wisdom Panel results. Every test of that brand identified him as predominantly boxer, plus Akita, Australian kelpie and a light blend of other breeds.
Canine Heritage agreed that the chief detectable breed in Finn is boxer. That test identified Australian shepherd and Norwich terrier "in the mix."
"Boxers are pretty inbred, so it's not hard to identify them," Hughes said. "They have a pretty strong signature."
Initially, we planned to repeat some tests on the same dogs to check for consistency. However, in testing Tag and Finn, we learned that the people at Wisdom Panel could tell when they'd seen a dog previously, even though we submitted each sample under a different name — assigning pseudonyms to the dogs and tapping friends and family to send in samples for us. With results of successive tests, the lab wrote: "Our analysis of the DNA of this sample indicates that we have tested the DNA from this dog before..."
Asked whether early results from a dog undergoing repeated testing is factored into successive tests, Hughes acknowledged that they may be. "When a sample is received that we have seen before, the data from both samples is tested independently and then combined such that any missing (genetic information) from one sample can potentially be filled by the other – basically we get a consensus set of the data and that is also analyzed," she said. "Therefore, a previous sample may influence a second sample but they are all independently assessed as well."
Salem and Solara: cross breeds with known ancestry
Expanding our pool of test subjects, we recruited two dogs from a research colony at UPenn with the help of Dr. Margret Casal, an associate professor of medical genetics, pediatrics, and reproduction at UPenn's School of Veterinary Medicine.
The beauty of the research dogs is that they are crossbreeds of known lineage. As with Tag, we could see whether the tests delivered the correct answer.
Because of Wisdom Panel's ability to detect repeat samples, we opted to run the dogs through just one of their tests — the mixed-breed cheek-swab version — plus Canine Heritage.
Casal provided a precise rundown on the dogs' backgrounds. The trickiest mix came in the form of Salem, who is "exactly 68.4 percent beagle, 14.4 percent German shepherd, 9.4 percent giant schnauzer and 7.8 percent basset," she told us.
Wisdom Panel found the beagle in Salem, along with a touch of German shepherd. But it also detected Ibizan hound in her background, a finding that made Casal chuckle.
"Hmm, there were never any of those in our colony, so I am not sure where this came from," she commented.
Wisdom Panel provides an ancestry chart going back three generations that it says "best fits the DNA marker pattern observed." Salem's result demonstrates that such charts should not be accepted as a precise family tree but merely a guideline when dealing with complex mixes. The chart shows one of her parents as purebred beagle. Casal said Salem's maternal grandmother was a purebred beagle but not her parents.
Overall, though, Wisdom Panel did much better with Salem than its competitor, Casal concluded, noting, "The Canine Heritage test failed to even recognize the beagle!!"
That test discerned no primary breeds. It named German shepherd as a secondary breed, and Chesapeake Bay retriever as "in the mix."
Rogatkin of Canine Heritage suggested that Salem's DNA signature may have been harder to decipher because she comes from a research colony that is genetically isolated from, and therefore genetically dissimilar to, the pet beagle population.
That may be, Casal said, but she pointed out that Mars' Wisdom Panel had no trouble with Salem. "That shows the Mars database is much bigger and thus more powerful," she said.
The other dog we tested from UPenn was Solara, whose mother is purebred golden retriever and father is purebred English springer spaniel from Australia.
Wisdom Panel identified both breeds plus a minority blend of other breeds. Canine Heritage found golden retriever only, possibly indicating that Australian-bred English springer spaniels were not in its database, Casal said.
Annie and Laika: mixed results
Our fifth test dog was Annie. Like Solara, Annie is a cross of two known purebreds. Her owner, Carl Doby, obtained her from a breeder whose pedigreed Australian shepherd had an unauthorized liaison with a certified Jack Russell terrier.
Wisdom Panel identified Annie as half rat terrier, one-quarter border collie and one-quarter potpourri.
Doby was impressed with Wisdom Panel's presentation of Annie's results ("They emailed me a lot of stuff — charts and bloodlines") and said he would recommend the test to others. He found the results reasonable because of the close relationship between rat terriers and Jack Russells, and the similarities between Australian shepherds and border collies.
Reviewing Annie's results, Hughes commented, "You could not have picked two more difficult breeds due to their diversity" as the Jack Russell and Australian shepherd.
In Australian shepherds, she said, field lines and show lines diverge.
As for the Jack Russell, she said, its genetic signature clusters closely with that of the rat terrier, toy fox terrier and, to some degree, Chihuahua. Jack Russells themselves are "not a very consistent breed," she said. The breed is not recognized by the AKC — the pre-eminent purebred registry in the United States. Annie's mother's purebred certification is from the Continental Kennel Club (CKC).
Canine Heritage identified no primary or secondary breeds in Annie; only Pembroke Welsh corgi "in the mix."
Canine Heritage spokeswoman Rogatkin told the VIN News Service that Annie might not be the cross Doby believes her to be, considering the source of her certification. "A purebred dog has a very exact breeding history if you're with a rigorous and demanding registry, for example, AKC," she said. "Different registries have different demands. The CKC is by far not as rigorous."
Doby separately contacted Canine Heritage to discuss Annie's results and came away dissatisfied. "I called them to tell them how far they were off," he said. "I told them I had papers on the mom and dad and they got offended and said I didn't know what I was talking about."
The final dog we tested was Laika. Her owner, Dr. Maureen Roberts, is a veterinarian in California who tested Laika in 2007 when Canine Heritage first became available. We repeated that test to see if the results would be consistent, and to compare the results with Wisdom Panel's. Here's what we got:
2007: Canine Heritage reported Chinese shar-pei, Akita, Siberian husky and border collie as secondary breeds.
2012: Canine Heritage reported Siberian husky as a secondary breed, with border collie and Bernese mountain dog "in the mix."
2012: Wisdom Panel determined Laika to be half Siberian husky and a melange of other breeds, possibly bull terrier, shiba inu, basenji, dachshund and greyhound.
Roberts found the results no more informative than her own guess about Laika's lineage.
"Based on looking at Laika, I know she is a husky mix, but I don't know what she is mixed with," Roberts said. Of the possible breeds named in the three tests, the only one she finds believable is border collie.
From her experience, Roberts would not recommend DNA testing to determine breed. "I don't think the test really tells us anything more than we can tell just by looking at the dog and making a guess," she said. "Since I got different results even with the same company, it makes me pretty skeptical."
Photo courtesy of Dr. Luis Tarrido
Is this dog part beagle? By chance, Sidney’s DNA was run through the Mars Wisdom Panel Professional test twice last year. Both showed toy fox terrier and flat-coated retriever,
but one detected 25 percent beagle, while the other showed no beagle at all. The inconsistent results led his owners to doubt the test’s validity.
Unplanned repeat test yields fickle result
In the midst of our project, we became aware of another dog that by chance was tested twice. The dog, Sidney, belongs to Dr. Luis Tarrido and his wife, Dr. Kelly Czech, veterinarians in New York state.
Using the Wisdom Panel Professional kit, they mailed a blood sample from Sidney that got lost in transit. Wisdom Panel provided a second kit at no charge, and the veterinarians sent in a second sample. A few weeks later, the lost sample made it to the laboratory. This time, the people at Wisdom Panel apparently did not notice they had run the same dog before. Sidney ended up getting back-to-back analyses.
The results didn't quite match.
One identified Sidney as a mix of flat-coated retriever, toy fox terrier, beagle and an assortment of other breeds.
The other showed Sidney as toy fox terrier, flat-coated retriever and assorted others.
The list of possible other breeds named were fairly consistent: The first test results identified Chihuahua, Australian shepherd, Finnish spitz, miniature dachshund and weimaraner. The second had the same breeds with the exception of the spitz. That analysis detected rat terrier instead.
But the beagle's disappearing act Tarrido found hard to swallow.
Reporting his experience on a message board of VIN, Tarrido wrote: "I know that the mixed-breed make-up percentage might be a little off and might vary a little but (for) the beagle to not be there ... come on!?"
Tarrido called Wisdom Panel and spoke with geneticist Hughes. She reviewed the results, and discovered that the statistical confidence in the beagle finding was marginal. "When you delve into the data, it was at a low confidence; as low as you can get and still make it onto the chart," Hughes told the VIN News Service.
She dubbed the beagle report a mistake. "It falls into our 10 percent (possibility of) false-positive or false-negative," Hughes said.
Interestingly, Tarrido noted that Sidney "does have a beagle bark, unfortunately ... that I get to hear every night."
To the general consumer, a DNA test for breed may project an aura of scientific precision. But as our six test dogs plus Sidney demonstrate, the analyses are not exact. Bell, the veterinary geneticist at Tufts, said the commercial tests should not be confused with highly accurate medical diagnostic genetic tests.
"A mixed-breed ancestry test is a non-diagnostic novelty test that is consumer-driven," he said. "The science of these tests can be compared to trying to deconstruct the ingredients of a complicated recipe – maybe there's some of this, or maybe some of that. While we would like it to be as accurate as possible, no medical decisions are going to be based on it. For the consumer, it is probably more important that they are happy with the results than their exactness."
As Hughes sees it, DNA testing is much more accurate than someone looking at a dog and surmising its heritage. "Our brains are able to handle one, maybe two breeds and put them together, but when you talk about 9 percent giant schnauzer and 14 percent German shepherd, we can't figure it out," she said. "Genetic testing is a significant improvement over visual identification."
That hasn't been definitively proven, although one team of researchers did compare the results of visual breed identification with genetic identification in a study of 20 dogs with unknown parentage. Looking at what breeds the dogs were said to be by their respective adoption agencies and what the DNA results showed, the researchers found a sizable discrepancy: the two methods agreed less than one-third of the time.
The study, "Comparison of Adoption Agency Breed Identification and DNA Breed Identification of Dogs," appeared in the Journal of Applied Animal Welfare Science in 2009. The lead author, Dr. Victoria Voith, a professor of animal behavior at Western University, presented the findings at a conference of the American Veterinary Medical Association that year, saying the results raise questions about the validity and enforcement of breed-specific policies.
Do a mutt's breeds matter?
What breeds comprise a mutt is clearly interesting to most, if not all, dog owners, but beyond simple curiosity, is there any value to knowing?
Giger, head of the clinical program in medical genetics and pediatrics at UPenn's veterinary school, thinks so — especially if the information is available when the dog is a puppy or before it is adopted. "It's important when considering a puppy, and it's tiny and cute and you would like to know how big it's going to get," he said.
Similarly, it would be helpful to know its likely behavior, (such as) if a dog has a herding background, in order to place it in the proper home, Giger added. "One wouldn't want to place it into an apartment where it might go berserk because of boredom."
The test makers maintain that knowing breed background enables owners and their veterinarians to better target a dog's health care, alerting them to watch for diseases to which the dog's breeds are prone.
Hughes often tells a story of her own dog, Rimsky, whom she adopted as an older puppy from a shelter in Sacramento, Calif. The shelter called him a border collie because he was white with black spots, but when his weight topped out around 20 pounds and he developed a feathered coat, Hughes described him as "an overgrown papillon." At age 5, Rimsky had an episode of frantic scratching at his face and seemed to be choking. Hughes rushed him to an emergency hospital but even before they arrived, he appeared to fully recover, and no cause for his distress could be found.
Six months later, Rimsky had a grand mal seizure. Eventually, he was diagnosed with epilepsy. Hughes believes now that during the face-scratching episode, Rimsky was having a smaller seizure. Years later, when the breed test became available, she tested him. The result said he was part cocker spaniel, part Maltese and other blends. "It made so much more sense," Hughes said, because cocker spaniels are prone to epilepsy. Had she known his breeds earlier, she may have made the connection to his condition sooner, she said, resulting in less anxiety for her and earlier treatment for him.
Hughes' personal experience notwithstanding, the notion of tailoring medical care to the results of a DNA mixed-breed lineage test is not widely held. Casal, the veterinary geneticist from UPenn, suggested it is driven more by marketing than science. She said that a dog made up in large measure of one breed might be predisposed to health conditions particular to that breed, but such dogs are likely to be identifiable without needing a DNA test.
"If your dog comes back, for example, 80 percent poodle, then you might actually worry about Addison's," she said, speaking of an adrenal-gland disease common to poodles. But in that case, the dog would be apt to look like a poodle, she said. Its veterinarian, therefore, probably already would be on the lookout for health conditions associated with poodles.
Test makers also maintain that knowing a dog's breed background can make training it easier. Veterinary behaviorists consulted by the VIN News Service said that may be true up to a point.
Dr. Laurie Bergman, a specialist in veterinary behavior, said: "I guess it can help somewhat in terms of helping owners have a better understanding of their dogs but even with purebreds you do get variations in behaviors, even the behaviors that those breeds were selected for ....
"More important than a breed test in successfully training any dog is understanding positive reinforcement training," Bergman said. "... Find what motivates your dog (which may or may not be shaped by breed) and use that to reward the dog."
Another behavior specialist and consultant, Dr. Ellen Lindell, had a somewhat different take. "It could indeed be helpful to know the breed background of your dog," she said.
"We have bred certain types of dogs, to be called purebred, and to exhibit useful behaviors in a somewhat predictable manner," she elaborated. "Some terriers may herd sheep, but if you have plans to participate in herding trials or need a dog to manage livestock on your farm, you would probably seek a dog from the herding group, rather than rely on the hope that your digger/rodent-killer might want a side job herding sheep."
She continued: "Purebred dogs were separated into groups based on their functions, not on appearance." For instance, she said, "Since terriers and hounds work independently, they might not be as sensitive to owner cues. Herding dogs are ultra vigilant — bred to follow subtle movements and notice all sounds. Inexperienced owners might not give these dogs enough information. Most dogs in the working group were bred to guard, so if a person with a working dog wants an open-door policy, then that person will need to work extra hard to teach the dog that everyone is welcome. So yes, it can help people to know what to expect so (as to) make better matches and help establish reasonable goals."
But she cautioned that breed isn't a sole determinant. "Of course there can be variation within a breed," Lindell said. "Each individual dog should always be evaluated based on his own behavior."