Analyzing the Analysis: Wisdom Panel DNA Tests & Mixed-Breed Dogs
We started a “Guess the Breed Background” contest this week after sending in a DNA sample of Katie’s dog, Leo to the Mars Wisdom Panel lab. Here’s a photo of Leo if you missed it:
Over 292 people sent in their guesses, and congratulations are due to Robin S! Of all the 290+ guesses on the blog, you came closest of anyone to guessing Leo’s genetic background, as analyzed by the Wisdom Panel (Professional), by guessing (in order of genetic representation), Boxer/Shih Tzu/Sheltie . But then, no one actually came very close to guessing what came back on the Wisdom Panel, which was:
One parent was half Boxer and half American Staffordshire Terrier (so Leo is 1/4 of each). (No one guessed both breeds as part of the mix, which is understandable if you think about it, but a good reminder to us all when guessing breed backgrounds.) This result appears to be definitive, in that there is a very high confidence level that one of Leo’s parents was indeed a Boxer/Am Staff cross.
The other parent came back as “Mixed Breed,” as did Leo’s grandparents and great grandparents. The top 5 breeds comprising that ‘mixed breed’ were German Spitz 21.86% , Shuh Tzu 8.5%, Belgian Tervuren 6.35%, Parson Russell Terrier 4.88% and Mastiff 4.49%.
Reports like this have generated a veritable stew of questions, confusions and in some cases, if I may be so blunt, snorts of derision. A cursory look at articles and comments about the results that others have received are full of skepticism about reports that include extremely rare breeds like Dandi Diamont Terriers, Jindos, Cane Corso & Scottish Deerhounds as part of the mixed-breed analysis. Look, for example, at some of the analyses that came up in a study being done by Dr. Julie Levy. I can understand the questions: Exactly how many Salukis or Brussels Griffons are out running the streets breeding willy nilly with someone’s mutt down the road?
In Leo’s case, Katie and I were 100% on board with the Boxer/Am Staff results. We always thought he had some pit-type breeding in him, and Leo behaves as much like a Boxer as, well, a Boxer. He is high energy and clownish and most importantly, uses his paws like a Boxer all the time. In Katie’s words “He thinks any problem can be solved by throwing a paw at it.”
Then we looked at the breeds listed under “Mixed-Breed.” German Spitz? Hummm. Seems possible, but certainly not a breed that comes to mind looking at Leo. (At least not to us.) However, look up German Spitz and you’ll find that it is both a breed and a type, which includes American Eskimo and Pomeranian. And 8.50% Shih Tzu? Seems possible. 4.49% Mastiff? Hmm, what might that mean?
This is when it is wise to actually pay attention to what the good people at Mars are telling you. Here’s the text at the top of the page titled “What does the Mixed Breed Analysis mean for my dog?” It began by saying “We have identified for you the next 5 best breed matches which appeared in the analysis of your dog’s DNA. One or more of these breeds could [my emphasis] have contributed to the genetic makeup of the ancestors indicated by the mixed breed icon.” It goes on to say that there may be a breed or breeds that they can’t detect with their current data base of purebred dogs. In other words, they are confident that one of Leo’s parents is a half Boxer/half Am Staff, and are making likely predictions of the other parent. In other words, the ‘mixed breed’ analyses are guesses. Good ones, but still, just guesses based on probability statements. Neither are they saying that Leo is, for example, exactly 21.86% German Spitz, but rather, “…the relative strength of each result in our analysis with the most likely at the top of the list.” Exactly, then, what the percentages mean were unclear to me, until I spoke to several people at Mars itself.
After networking through just a few channels I was able to speak with two of the top people at Mars Veterinary, Alan Martin, PhD and Senior Bioinformatics Scientist, and Angela Hughes, DVM & PhD and Veterinary Genetics Research Manager. They know more about genetics in their sleep than I do wide awake–I don’t even know what “bioinformatics” is for heavens sake. Happily, after almost an hour and a half speaking to the two of them, I knew a heck of a lot more than I did before we spoke. (Don’t be jealous; anyone who would like to talk about the result they obtained is more than welcome to call and speak to an expert at the company.)
Here are some of the things I learned: First off, the test done by Mars is best described as a pattern recognition test. The computers (each dog’s analysis requires 7 million runs) look at 321 “markers” on the chromosomes in the sample, looking for patterns of base-pairs in the strands of DNA. The computer program is looking for patterns consistent with both different ancestral breed types (“Guard breeds” like mastiffs, Am Staffs, etc, “Ancient breeds” like Chows and Siberian Huskys or “European breeds” like hounds and retrievers), and individual breeds, each of which has its own distinctive pattern of arrangement of base-pairs. (Here’s a good Genetics 101 refresher for anyone who would like the review.) Some breeds have more definitive patterns than others, especially those that have been closely bred as purebreds for a long time. Dr. Hughes reported that breeds like Catahoulas, Jack Rusell Terriers and Rat Terriers, who were bred more for function and less for form for so long, are more varied genetically and thus a bit harder to pin down. (It also explains why some of these breeds show up as false positives, when the computer searches can’t find a match and in some electronic version of desperation, settles on a Catahoula.
Speaking of Rare Breeds: That was my primary question to both scientists at Mars. What do those low percentage breeds really mean? That somewhere way back in time Leo really did have a purebred Tervuren ancestor, or not? The answer is maybe. It could be that Leo simply has some patterns of base-pairs on his chromosomes that just happen to be very much like a similar sequence in Tervs or more likely, a related breed in a similar group. It could be that indeed, somewhere back in time a purebred Terv bred with someone’s purebred Whatchamacallit. Mars has done extensive testing of their computer models and found that their “Mixed-Breed” analysis is about 90% accurate. Given the complexity of genetics, that is impressive. It also means that 10% of the time, it is flat wrong. Thus, an analysis could come up with several very likely breeds and one outlier. Rather than dismissing the results as a whole, the surprising outlier simply deserves more attention, or to be taken with a grain of salt.
If you’ve scrolled ahead you’ve just seen a photo of Leo’s mother. She came into the shelter pregnant, and so half of Leo’s parentage is definitive, if not to breed at least to an individual. Katie did not send the photo in with the DNA sample (she sent in a blood sample; do be careful of cheek samples, they can be easily compromised although it costs less to do). However, I sent the photo in this morning to Dr. Hughes at Mars and she immediately said “Oh, that explains the breed that came up just under our reporting criteria, Miniature Wire-Haired Dachshund.” (I thought PBGV when I first saw her. Shows what I know… which is enough, I would add, to keep my mouth shut whenever anyone asks me to guess a breed.) Based on looking at all the details of the data (only the top 5 breeds making up a “mixed-breed” are reported unless you call and ask), there is a suggestion that Leo’s mom has a lot of Min WH Dachshund and Shih Tzu in her. I also learned that wire hair (or “furnishings” as they as so delightfully called) and short legs are dominant traits, so it is not surprising that they show up in Mom.
Overall, after all this, I have a much more nuanced understanding of what the results of the Wisdom Panel tests actually mean, and find the results to be extremely interesting. Much of the time I would argue that the motivation for the tests is to satisfy our curiosity, a trait I find to be one of humankind’s most useful and endearing. However, there are other more practical applications: Once one knows if there are purebred lines well represented in one’s “mutt,” one can also ask for Genetic Mutation Tests to be done. For example, Terriers suffer from several diseases more often than other breeds (lens luxation for example), and Leo’s test shows that he is negative for the ADAMTS17 mutation that could cause that affliction. Good to know, especially if early treatment would make a difference in the case of some diseases.
But what if you get back a report that says “No Result”? I read several comments by angry or disappointed people who got back a report that said there was no particular breed that could be assigned to their dog. They felt that they had wasted their money, and clearly wanted to get results that listed a breed, any breed, in the report. I asked Dr. Hughes about this, and she said first, reports are now sent even with low probability results, in the understanding that everyone wants to know something about what the analysis found. In addition, all “No Result” reports are now sent with a letter that describes the process that led to the finding, and offers a refund if the owner wants one. One could argue it is possible that a re-analysis would come up with a different result; even computers make mistakes, and sometimes, in Dr. Hughes’ words, “go down a tangent” and need to be redirected. I’m glad to hear it’s not just me that can be easily distracted.
However, in some cases one’s dog really contains such an amalgam of genes that it is truly impossible to associate any breed to it. If I got back such results I would be fascinated, because to me that’s not “No Result,” it’s an extremely interesting one. Biologically, such dogs are called “mongrels,” but many hesitate to use the word because it has negative connotations in other contexts. If my dog came back closer to a village dog than a Labrador, I’d be fascinated, not disappointed. I’m not sure that is, genetically, the correct analogy, but the point here is that “no breed detected” is a definite result to me, and an interesting one at that. My very-much-unsolicited advice would be to re-title “No Result” into “Unique and Special” or “All American” (the All American is actually a favorite phrase of Dr. Martin, who tells me that US dogs have much more genetic variance than dogs in the UK… many more strays and random matings for example). Dr. Hughes tells me that the most common questions they get are questions about breed ID: That black is dominant and so is black and white spotting, so not every all black dog has any labrador in it and not all black and white dog have a snippet of Border Collie.
In summary, I better understand now why some of these rare breeds pop up in the analysis, and have more confidence in the results. Next time I am curious about a dog I suspect I won’t be able to resist sending in a sample. Most importantly, blog reader Robin S came closest to the analysis, and wins the DVD of her choice.(Robin, email us at firstname.lastname@example.org and tell us what DVD you’d like!)
MEANWHILE, back on the farm: Beyond blogs, grading term papers, and trying to keep up with email (ha!), it’s pretty much all about lambing. Four ewes have lambed, with eight lambs total. Two sets of twins are doing well, but things have not gone so smoothly with the others. Rosebud (triplets) and her daughter Oreo (who delivered twins but one was found dead when I returned home from UW on Tuesday evening) both have dsyfunctional udders. It’s not typical mastitis, no fever, the little milk that comes out is fine. But their bags are hard as rocks, poor things, and they are giving almost no milk at all. The condition is called “hard bag” (not very creatively, but certainly descriptive) and isn’t well understood, there are several possible causes but no reliable cures. We’re working on it (vet visits, hot packs to the udders, reaching out to UW specialists) but right now the 4 lambs are being bottle fed as often as possible with goat milk from our good friend Donna H. (Bless you!) I’d write more, but the lambs are way overdue for milk. And we have 3 ewes to lamb yet… if one has a single I might be able to graft a one of Rosebud’s triplets on to her. Rosebud has completely rejected one little lamb, so it would be good to graft her onto another ewe if she has enough milk. Solo looks like she might be thinking about having her lamb(s), I’m hoping for a single.
Here are Buttercup’s two contented, napping lambs, more lamb photos when I get a minute:
And for those of you in the Facebook “When will Spot lamb pool”…. she’s not saying.