Relatedness is relative: How can I be 85% genetically similar to my mom, but only related to her by half?

First of all, no. I am not the lovechild of star-crossed siblings, or even cousins, or even second cousins. 

This is a gee-whiz kind of post. But the issues are not insignificant.

Hear me out with the background, first, before I get to the part where my eyes bug out of my head and I pull out my kid's Crayola box and start drawing.

If you've learned about sociobiology, or evolutionary psychology, or inclusive fitness, or kin selection, or the evolution of cooperation and even "altruism," or if you've read The Selfish Gene, or if you've been able to follow the debate about levels of selection (which you can peek at here)...

... then you've heard that you're related to your parents by 1/2, to your siblings by 1/2 as well, to your grandparents and grandchildren by 1/4, to your aunts and uncles and nieces and nephews by 1/4 as well, and to your first cousins by 1/8 and so on and so forth.  (Here's some more information.)

So, for example. For evolution (read: adaptationism) to explain how cooperative social behavior could be adaptive in the genetic sense, we use the following logic provided by Bill Hamilton, which became known as "Hamilton's Rule": 

The cost to your cooperation or your prosocial behavior (C) must be less than its benefit to you (B), reproductively speaking, relative to how genetically related (r) you are to the individual with whom you're cooperating. That could have come out smoother. Oh, here you go:

C < rB, or B > C/r

If you're helping out your identical genetic twin (r=1.0), then as long as the benefit to you is greater than the cost, it's adaptive.

C < B, or B > C

If you're helping out your daughter (r = 0.5) then as long as the benefit to you is greater than twice the cost, it's adaptive.

C < (1/2)B, or B > 2C

So already, the adaptive risk to helping out your daughter or your brother is quite higher. And it's even harder to justify the cooperation between individuals and their sibs' kids, and grandkids, especially ESPECIALLY non-kin. But, of course creatures do it! And so do we.

As relatedness gets more distant and distant, we go from 2 times the cost, to 4 times, 8 times, 16, 32, 64 etc... You can see why people like to say "the math falls away" or "drops off" at first or second cousins when they're explaining where the arbitrary line of genetic "kin" is drawn.  If you offer up a curious, "we're all related, we're all kin," someone out of this school of thought that's focused on explaining the evolution of and genes for social behavior may clue you in by circumscribing "kin" as the members of a group that are r = 1/8 or r = 1/16 but usually not less related than that.

This has long bothered me because we're all genetically related and so much cooperation beyond close kin is happening. And it's been hard for me, as someone who sees everything as connected, to read text after text supporting "kin selection" and "kin recognition" (knowing who to be kind to and who to avoid bleeping), to get past the fact that we're arbitrarily deciding what is "kin" and it seems to be for convenience. I'm not doubting that cooperation is important for evolutionary reasons. Quite the contrary! It's just that why is there so much math, based in so many potentially unnecessary assumptions about genes for behavior, gracing so many pages of scientific literature for explaining it or underscoring its importance? 

(It could just be that as an outsider and a non-expert I just don't understand enough of it and if I only did, I wouldn't be gracing this blog with my questions. But let's get back to my reason for posting anyway because it's potentially useful.)

Right. So. Even for folks who aren't part of evolution's academic endeavor, it's obvious to most that we're one half dad and one half mom. The sperm carries one half of a genome, the egg another, and together they make a whole genome which becomes the kid. Voila!

There's even an adorable "Biologist's Mother's Day" song about how we've got half our moms' genome... 

... but there's biology above and beyond the genes we get from mom (and not from dad). And that song is great for teaching us that the rest of the egg and the gestational experience in utero provide so much more to the development of the soon-to-be new human. So "slightly more than half of everything" is thanks to our mothers. Aw!

But, genetically, the mainstream idea is still that we're 50% our mom. 

I teach very basic genetics because I teach evolution and anthropology.And I'm not (usually) a dummy.* I get it. It's a fact! I'm half, genetically, my mom and I'm also half my dad. 

r = 0.5

Okay! But, given these facts about relatedness and how it's imagined in evolutionary biology, facts that I never ever questioned, I hope you can see why this report from 23andMe (personal genomics enterprise) blew my mind:

Percent similarity to Holly Dunsworth over 536070 SNPs (single nucleotide polymorphisms or, effectively/rather, a subset of known variants in the genome; Click on the image to enlarge).
I am 85% like my mom and I am at least 76% like my students and friends who are sharing with me on 23andMe. Names of comparisons have been redacted. As far as I know, this kind of report is no longer offered by 23andMe. I spat back in 2011/12 and the platform has evolved since.

Okay, first of all, it is a huge relief that, of all the people I'm sharing with on 23andMe, the one who squeezed me out of her body is the most genetically similar to me. Science works.

But that number there, with my mother, it is not 50%. It's quite a bit bigger than that. It says I'm over 85% the same as her.

What's more, I am also very similar to every single person I'm sharing with on the site, including example accounts from halfway around the world. Everyone is at least 60-ish% genetically similar to me, according to 23andMe. I know we're all "cousins," but my actual cousins are supposed to be 1/8th according to evolutionary biology. How can my mom be related to me by only one half? How can my actual cousins be only an eighth (which is 12.5%)? 

What is up with evolutionary biology and this whole "r" thing?

Hi. Here is where, if they weren't already, people just got really annoyed with me. Evolutionary biology's "relatedness" or "r" is not the same as genetic similarity like that reported by 23andMe.

Okay!

But why not? 

Let me help unpack the 85% genetic similarity with my mom. Remember, it's not because I'm inbred (which you have to take my word for, but notice that most everyone on there is over 70% genetically similar to me so...).

It's because my mom and dad, just like any two humans, share a lot in common genetically. Some of the alleles that I inherited from my dad are alleles that my mom inherited from her parents. So, not only is everything I got from her (50%) similar to her, but so are many of the parts that I got from my dad. 

Let me get out my kid's arts supplies.

Here is a pretty common view of relatedness, genetically. In our imagination, parents are not related (r = 0) which can lead our imagination to think that their alleles are distinct. Here there are four distinct alleles/variants that could be passed onto offspring, with each offspring only getting one from mom and one from dad. In this case, the sperm carrying the orange variant and the egg with the blue variant made the baby.

1. (Please, if you're horrified by the "r" business in these figures, read the post for explanation.)

But few genes have four known alleles, at least not four that exist at an appreciable frequency. Some could have three. What does that look like? 

The green allele doesn't exist in the next example. As a result of there being only three variants for this gene or locus, mom and dad must share at least one allele, minimum. That means, they look related and that means that, depending on which egg and sperm make the kid, the kid could be more related to mom than to dad. 

2. (Please, if you're horrified by the "r" business in these figures, read the post for explanation.)

Now here's where people who know more than I do about these things say that the kid is not more related to mom than dad because she got only one allele from mom and that keeps her at r = 0.5. 

Well, that's just insane. What does it matter whether she got the allele from mom or dad? I thought genes were selfish? (Sorry, for the outburst.)

Again, I realize I'm annoying people and probably much worse--like stomping all over theory and knowledge and science--by mixing up the different concepts of genetic similarity (e.g. 50%) with "r" (e.g. 0.5) and horribly misunderstanding all the nuance (and debate) about "r," but I'm doing it because I'm desperately trying to know why these two related ideas are, in fact, distinct. 

One last pathetic cartoon. 

In this third example, as is common in the genome, there are only two alleles/variants in existence (at an appreciable frequency, so not accounting for constant accumulation of de novo variation). An example of such a gene with only two known alleles is the "earwax gene" ABCC11 (there's a wet/waxy allele and dry/crumbly one). Here, the two alleles are orange and blue. Most humans in the species will have at least one allele in common with their mate for a gene with two alleles, and it's not because most humans are inbred, unless we want to redefine inbreeding to include very distant relatives (aside: which may be how the term is used by experts). 

3. (Please, if you're horrified by the "r" business in these figures, read the post for explanation.)

But as a result of the chance segregation of either the blue or orange allele into each of the gametes, two people with the same genotype can make a kid with the same genotype. 

And of course, making a kid with your same genotype is the only possible outcome if you and your mate are both homozygous (i.e. where both copies are of the same variant so that leaves no chance for variation in offspring unless there is a new mutation). 

So, I wandered a little bit away from my point with these drawings, but I had to because I wanted to get down from where my imagination has me (us?) with "r" versus how things really are with reproduction. We are baby-making with vastly similar genomes to ours, so we are making babies with vastly similar genomes to ours. 

So, I do see why biology says I'm related to my mom by one half. But, on the other hand, what does it matter if I got the thing I have in common with my mom from my mom or whether I got it from my dad? Because I got it. Period. It lives. Period. 

[Inserted graf January 20, '17] Saying it matters where I got the similarity to my mom keeps us at r = 0.5. Saying it matters only that I inherited DNA like hers keeps us always, all of us, at r > 0.5 with our parents and our kids because any two babymakers share much of their genome.

And the fact that this (see 2 and 3) happens so often is why I'm a lot more than 50% genetically like my mom, and the same can be said about my genetic similarity to my dad without him even spitting for 23andMe. 

So, here we are. I don't understand why our relatedness to one another, based on genetic similarity, is not "r."

I hope it's for really beautifully logical reasons and not something political. 

Because...

If "r" was defined by genetic similarity, then would cooperating with my 76% genetically similar students and friends be more adaptive than the credit I currently get from evolutionary biology for cooperating with my own flesh and blood son? 

If "r" was defined by genetic similarity, then could we use the power of math and theoretical biology to encourage broader cooperation among humans beyond their close kin? 

So many questions.

Maybe I should re-learn the math and learn all the other math.

Nah. Not myself. At least, it wouldn't come fast enough for my appetite. Maybe someone who already knows the math could leave a comment and we could go from there... 

And it would be worth it, you know, because despite my relatively weaker math skills, I bet we're more than 50% genetically similar.

*from 23andMe: "You have 321 Neanderthal variants. You have more Neanderthal variants than 96% of 23andMe customers."

My dogs' evolutionary history. Part 3: Inside scoop and why

Recall that we ordered DNA testing kits for our mutts. First we made predictions, then we revealed the results. Today I want to tell a bit more of the story...

yawn

Remember that time a doctor used a vibrator on my hips because I had a cyst on my knee? I was all into it "for science." It was an opportunistic science sting operation. There was totally gum on my shoe. There was all kinds of pseudoscience that day.

Anyway, I was all psyched up for that sort of Dateline-style, hidden camera intrigue when I called up Mars Veterinary, the people who do the analysis for Wisdom Panel doggie DNA kits, last Friday afternoon. I was all psyched up because, as Kevin said, I needed, "to make sure they're not defrauding people."

Me? I do? Little old me?

That sounds about right. I mean, just because I can see all the educational potential here doesn't mean I should be endorsing it without understanding it better. And while I'm understanding it better shouldn't I make sure they're not entirely full of s--t?

But I already know they're not full of s--t because they gave us results for both Elroy and Murphy that made sense. They didn't report crazy results and there are lots of ways they could have... there are 200 breeds on that list, most of which do not look like our dogs or like they've wafted their DNA anywhere near our dogs.

So, they can't be completely fraudulent. Whew. Okay then, less of a burden on me for this interview. But wait. Could it be maybe a little more nuanced, this hypothetical fleecing? After all, they're the ones who saw my blog and reached out to me to have an interview--isn't that a little big brothery? And, after all, their website doesn't tell us what their methods are, not really, not for those of us who have ever operated a PCR machine or run a gel. And they don't tell us what these markers are in our dogs' reports that led them to make the reports. I'm used to having too much information, even, because of 23andMe.

Anyway, there were all kinds of reasons to be professional when I called up the veterinary geneticist. But after she answered my first question so well, I pretty much lost all professionalism and drooled all over her, albeit through the phone.

Here's what I learned. Please bear with me... I didn't take direct quotes because (a) that's hard to do without a hands-free headset (or fancy recording software I don't have because I never needed), and (b) I got a wee bit excited and flailed my arms most of the conversation. 

All questions and answers are my paraphrasing.

Question: Are you using a chip or what? How are you genotyping?

Answer: A chip like 23andMe's isn't cost effective. Yes, there's a chip but it's not specific to the test. They use a sequenom platform. It's PCR markers combined into panels.  They run those PCRs for 321 markers, broken up into several panels, optimized so that they don't interfere with each other clearly and don't have to change primers. They use nested PCR for segmenting around the mutation and then to get the exact SNP they throw the DNA in a mass spec to find the particular A, T, C or G mutation. Each has a different weight, so each can be identified. 

Question: Do you get both alleles from that?

Answer: Yes. They don’t sequence in both directions but they sequence both chromosomes.

Question:  Are those 321 markers for visible or perceptible phenotypes? 

Answer: No. None are visible phenotype. They're ancestry markers/SNPs.

Question:  Have you found any dogs that have none, not one, of those breed markers?

Answer: No. And we don't expect to. But it's not completely out of the question given the spectral nature of variation over space and time. 

Question: How did you validate your mixed breeds family tree methods? Do you have mixed dogs with known parents and grandparents and type them all in the family?

Answer: The pure breed tests validate so well already. And also, yes, they do validate with mixed dogs with known heritage but not many. The methods are built out of Ostrander's research with 85 breeds, 5 dogs from each and 96 microsatellites. She licensed the patent exclusively to WisdomPanel. 

Question: How do your mixed breed reports compare to how people guess based on external cues? Aren't people already good at guessing their mutts ancestries?
Answer: Actually they're not. Ancestry DNA testing is much better. People aren't that good at guessing mutt ancestry based on phenotype, as far as the few studies suggest. (There's one by Victoria Voith that I want to check out and there's one by Levy on pitbulls specifically.)

***

We chatted for much longer than that and she shared all kinds of interesting information about genetics for dogs' visible traits and also lots about their reproduction which I'm keen to learn about. But that right there's all the relevant information regarding our doggie DNA testing experience that I got during the 30 minute phone call. She was completely open with me about everything which is why I'm kicking myself that I forgot to ask why they require the dog's weight when you submit its cheek swabs. I have a hunch, however, that it's got something to do with validating their mixed breed body size estimates given how many people use WisdomPanel on shelter puppies that will be better adopted if their projected size is known. 

So the question remains... If you don't have a puppy or don't care what it will grow up to look like or be like, why do Wisdom Panel? 

The promotional video on the homepage sums up why many people might do it: 

I wasn’t sure what my dog was made up of...Pretty sure she wasn’t just a [breed name]...I wanted to know so I could tell people when they ask me 'what kind is she?'

This is fascinating.

Why do we care where our dogs come from or what kind they are? Same reason we care the same things about humans and ourselves. And it all boils down to the fundamentals of the field of anthropology... and it's complicated.

And finally, why have I titled all three of these posts "my dogs' evolutionary history"? Because evolutionary history is a synonym for family history and for ancestry. And vice versa, all around. And, clearly, I'm trying to make a point about that.

When you type "evolution" into the Wisdom Panel website search, this is what you get:

But that's not going to stop me from using this to teach evolution. I've decided that Wisdom Panel will be a good alternative for students who aren't interested in doing 23andMe in my anthropology courses, so I'm going for it.

Dogs, humans, what's the difference?  A lot and not.

http://evolutionpsa.tumblr.com/

My dogs' evolutionary history. Part 2: Results

Yesterday we made the predictions of the breed signatures we'd find in our dogs' DNA.

If you didn't already, please consider going back to yesterday's post first before reading today's. Predictions are key! It's best if you confirm our guesses and/or add your own. (...what are the reasons for your predictions? Size? Color? Ears? Tails? Fur? Snout? Furnishings?)

Sadly (yes, sadly), predictions are not part of the official Wisdom Panel or 23andMe experiences. When I teach with 23andMe predicting the results is a major assignment early in the semester. Not only is making predictions the best practice for later scientific evaluation of the results, and it's the best way to force yourself to come to terms with how inheritance and gene expression work (and don't work), but guessing the outcome first makes reading the results orders of magnitude more fun ... not to mention how it makes things a lot less nerve-wracking when it's about your own DNA with 23andMe.

Briefly, before I reveal our dogs' results, let's consider a couple important things first...

Dog breed markers are mutations. 
Just like anything else alive right now, all dog breeds, no matter how "pure" or revered, are mutants. All of their dog traits just like all of our human traits, good and bad, started out as new mutations. Even the ones we all share that contain little variation now (like our genes involved in the development of five fingers and five toes), but also the ones that vary among our populations (like our genes that affect our pigmentation).

And mutations aren't just a population thing; each of us has a tiny fraction of our genome that's mutated compared to our parents. As far as we know, mutations occur in the making of all babies and puppies, etc. Most are neutral, some are bad, some are good.

We're all mutants because mutation, perpetual change generation upon generation within a lineage, is constant. Stasis is not.

This constant change in every puppy is fundamental to why we can have hundreds of dog breeds today.

Many dog breed traits are genetically simple.
Dogs seem to be particularly simple kinds of mutants.

The mutated genes that determine the traits that distinguish one dog breed from another are remarkably few and remarkably simple, but that simplicity makes a lot of sense.

Since humans coaxed these breeds out of ancestral dog stocks and also out of other breeds (as they still do today), it's easy to imagine that new simple traits, not new complex ones, had the best potential to be easily and quickly propagated into future generations and eventually into new breeds.

If a trait arose that was preferred, and it was caused by the kind of genetic mutation that could be simply and somewhat reliably expressed in offspring that inherited it, that mutation and the trait it produced could be increased in future generations by promoting breeding of those new attractive mutant dogs and their offspring.

(source)

If dog breed traits were genetically complex (based on many genes, for example), they would be terribly difficult to produce through controlled breeding of parents with those traits... at all... let alone during one human breeder's lifetime! That's at least partly because there would be far too many puppies without the preferred trait and it would be far too difficult to preserve a trait at an appreciable frequency in a lineage. Of course, inbreeding with very close relatives that share the mutations helps a great deal with this.

Humans have taken advantage of the simple mutations that have popped up in dogs (as they pop up in all living things!), due to sheer feasibility of the breeding outcomes those simple mutations allowed. Like that new coat curl or those new furnishings? Some puppies will have the exact same look. And we're off and running with a new kind of dog...

(source)

I don't know about the genes for Dalmatian spots (and honestly haven't even looked) but I do know about this paper by Cadieu and colleagues from 2009. Apparently you need only three genes (FGF5, RSPO2 and KRT71), each with two alleles (i.e. gene variants; denoted +/- in the figure below) to explain all this coat variation among dogs:

Cadieu et al., 2009

Dog genes are made of the same goop that ours are, and their genomes are very similar to ours because of our shared mammalian ancestry, but they're described as "simple." Evan Ratliff explains some more about why dog genes are simple in "How to Build a Dog:"

The vast mosaic of dog shapes, colors, and sizes is decided largely by changes in a mere handful of gene regions. The difference between the dachshund's diminutive body and the Rottweiler's massive one hangs on the sequence of a single gene. The disparity between the dachshund's stumpy legs—known officially as disproportionate dwarfism, or chondrodysplasia—and a greyhound's sleek ones is determined by another one. The same holds true across every breed and almost every physical trait. In a project called CanMap, a collaboration among Cornell University, UCLA, and the National Institutes of Health, researchers gathered DNA from more than 900 dogs representing 80 breeds, as well as from wild canids such as gray wolves and coyotes. They found that body size, hair length, fur type, nose shape, ear positioning, coat color, and the other traits that together define a breed's appearance are controlled by somewhere in the neighborhood of 50 genetic switches. The difference between floppy and erect ears is determined by a single gene region in canine chromosome 10, or CFA10. The wrinkled skin of a Chinese shar-pei traces to another region, called HAS2. The patch of ridged fur on Rhodesian ridgebacks? That's from a change in CFA18. Flip a few switches, and your dachshund becomes a Doberman, at least in appearance. Flip again, and your Doberman is a Dalmatian. "The story that is emerging," says Robert Wayne, a biologist at UCLA, "is that the diversity in domestic dogs derives from a small genetic tool kit."

So it's this simplicity that allows companies like Wisdom Panel to genetically distinguish breeds and then look for the signatures/markers of those breeds in our dogs' DNA. It's also the recency of most of these breeds (no more than a few to several hundred years at most) that allows us to assume (maybe not rightly but still...) that so much of what genetically identifies a breed today is similar to what the breed was working with all along.

And so... without further ado...

Here are the results of mailing off our dogs' cheek swabs and having 321 markers for 200 breeds analyzed.

For Elroy...
This is his "Breed Ancestry Certificate"

The report notes how Rottweiler and Chow Chow are the only ones with confidence. 

How do they make the pedigree chart? 

What I think they do is estimate what percent of the breed contributes to the dog's ancestry and if it's something like 50%, it's one parent. If it's like 25%, it's one grand-parent and if it's like 12.5%, it's one great-grandparent. Any % less than that is lumped into "mixed breed" ... the details of those are guessed at below.  So those placements on either side of his chart are just best fit in terms of percentage. 

What I'm not sure about is why we must assume that 100% of Elroy's ancestry comes from any breed. But, remember that 100% of the markers in the Wisdom Panel test distinguish breeds. That means, this Breed Ancestry Certificate is missing all the information about Elroy that's not analyzed by Wisdom Panel. 

Non-surprises.
We guessed the breeds that they were confident about! There must be something to this test. Especially since we had seen his litter mates and were told by the adoption agency that he was probably rott/sharpei (and later on we figured he was just as likely chow chow as sharpei). SCIENCE works.

Surprises.
German Shepherd and Collie aren't terribly surprising either. He came from Ohio. Those breeds are abundant enough in the region, are allowed to roam free and probably mated like that too for decades upon decades in those regions. And he's a big boy!

But Westie? Westie is a bit of a head-cocker.

Or is it? Westies had to have had big ancestors, first of all, since they descended from wolf ancestors like all other dogs. But something that distinguishes Westies from other breeds is in Elroy's DNA. That means a relatively recent Westie left a signature in his genome. So was it kamasutra lovemaking between tiny Elroy ancestor and big Elroy ancestor?  Or are we talking about love between mixed breeds and pure breeds and mixed breeds over time? Probably.

Here are their best guesses about what contributed to those "mixed breed" mysteries on his certificate.

Mostly big dogs, some ancient and awesome. The Dogo! (Ugh, don't google it... so many fighting and abuse videos.)

Nothing to really take home from this list because they're far off guesses without even a report of percentage there with the bar graph.

For Murphy...
This is her "breed ancestry certificate"

Akita, German Shepherd and Lab are the only confident ones (as noted in the report). 

And here are the most likely breeds contributing to those "mixed breed" mysteries on her certificate.

Non-surprises.

Shepherd! And Lab is another good one considering friends and relatives see lab in her.
Many of her mix guesses are herders!

Surprises.
Akita? That's phenotypically (and maybe genetically(?) will have to check) close to Husky which is another guess we often hear for her. So this is only a little bit surprising.

French Bulldog? Now that's a real head-cocker.

But again, like with the Westie, it's not noted as "confident" in the report, but on the other hand, why not French Bulldog in her ancestry? It just means this is a marker, like with the Westie, that's not tied to outward appearance/phenotype.*

Not enough information...
I wonder what these markers are then; if Westie and French Bulldog are showing up, why? As the website explains, “Physical features characteristic of certain breeds, such as the flattened face of the English Bulldog or the extremely curled tail of the Pug, seldom survive even the first crossbreeding.”

So these markers for unseen traits are fascinating but not explained in the report.  I'm spoiled rotten by 23andMe that tells you everything, down to the A,T,C, or G you have for your allele for whatever gene. I'm guessing that Wisdom Panel keeps all this under wraps because (a) most consumers don't care to know those details and (b) it's in their best competitive and economic interests to keep their methods to themselves.

What's interesting is how some of the hype about these dog tests is similar to 23andMe: They're "for your health!" But unlike with 23andMe, Wisdom Panel says that you're good just going to your vet and saying "she's got Akita in her blood" when you're dealing with health issues.  While 23andMe advocates that you know each and every SNP.  My health is more important than my dog's I suppose, but what's absolutely not clear is how these two very different approaches (ancestry and family history vs. SNP data) are resulting in different health outcomes.

As you might imagine, there's a lot more to say about the evolutionary and anthropological issues these dog tests raise, so please stay tuned. (Here's part 3.)

Kevin and me and our Westie, Elroy, and our French Bulldog, Murphy.

*Or if it is, it's present in only one recessive copy and would require two copies for the visible trait in the Westie or French Bulldog  to be expressed (added June 14)

My dogs' evolutionary history. Part 1: Predictions

[Click here to skip to Part 2 and Part 3]

You might remember Murphy and Elroy from the time we used science to solve the book-eating mystery. Or the time they figured out evolution. (Still looking for puplisher! [sick]) Or maybe you already know Elroy because you follow him (@ElroyBeefstu) on Twitter.

Elroy (I fit in your phone!)

Murphy (awww)

These are the mutts that inhabit our lives, and we theirs.

It's because of our tremendous love for these dogs but mostly because of our tremendous fascination with evolution that we ordered Wisdom Panel kits for each, for my birthday.

What we do

With Wisdom Panel, the process on the consumer end isn't a whole lot different from 23andMe. You purchase the kits online, they arrive at your house, you activate the kits online, you swab your dogs' mouths, pop the kits back in the mail with the postage-prepaid packaging, and wait for an email with the results.

Wisdom panel needs far less of the contents of a dog's mouth than 23andMe requires. And that's not only because the analysis will be far less extensive, but because dogs' lips and face muscles aren't hooked up for spitting into a test tube.

The turnaround was speedy. We sent in the samples on May 28 and got the results June 6.

At this stage in the process, the only red flag is that they require you to submit your dog's weight during online registration. For the love of science, there should be no phenotypic hints required. I hope they don't use weight to discard or confirm dog breeds for their results report.

What they do
Here's what the FAQ says:

Testing your dog with Wisdom Panel® 2.0 begins when you use the cheek swabs to simply collect a small DNA sample from inside your dog’s cheek and send the swabs into the laboratory. Once your sample is received at our lab it is scanned into our database and assigned to a batch for testing. It then undergoes processing to extract the DNA from your dog’s cells which is examined for the 321 markers that are used in the test. The results for these markers are sent to a computer that evaluated them using a program designed to consider all of the pedigree trees that are possible in the last three generations. The trees considered include a simple pedigree with a single breed (a likely pure-bred dog), two different breeds at the parental level (a first-generation cross), all the way up to a complex tree with eight different great-grandparent breeds allowed. Our computer used information from our extensive breed database to fill these potential pedigrees. For each of the millions of combinations of ancestry trees built and considered, the computer gave each a score representing how well that selected combination of breeds matched to your dog’s data. The pedigree with the overall best score is the one which is selected and provided to you in your dog’s individualized report.

This doesn't really cut it for me. I want to know what the methods are and this is all they provide in answers to "Science Based Questions!" This isn't helping much:

Not only does the computer analyze a dog’s DNA for the breeds and their likely proportions in the dog’s ancestry, but it also models which side of a dog’s ancestry each breed is likely coming from.

I'll just have to assume for now that they use something like a chip (because sequencing is still not thrifty) to identify markers that they've already linked to breeds and then they're applying their probability-based analyses to those markers in our dogs in order to provide an estimate of our dogs' ancestry. And what are those markers?

Wisdom Panel only uses what are called autosomal DNA markers, chromosomes that contain most of the genetic instructions for every canine’s body make up (height, weight, size etc.). There are no markers from either the so-called sex chromosomes (the canine X or Y chromosomes). Mitochondrial DNA, or Y-chromosome DNA testing, is rather different as these parts of the genome are passed on intact from mother to child and father to son respectively, but are therefore only representative of either the female or the male lineage. Autosomal DNA is inherited both from the maternal and paternal lineages equally and constantly shuffled by a process called recombination at each successive generation, and therefore is able to give useful information on the breeds found on both sides of a dog’s lineage.
To find the genetic markers that performed best at distinguishing between breeds, Mars Veterinary™ tested over 4,600 SNPs (single nucleotide polymorphisms or genetic markers, where genetic variation has been found between different dogs), from positions across the whole canine autosomal genome from over 3,200 dogs. To further refine the search, Mars Veterinary determined the best 1,536 genetic variations and ran them against an additional 4,400 dogs from a wide range of breeds. This stage of testing resulted in the selection of the final panel of DNA markers that performed best at distinguishing between breeds, ultimately creating the Wisdom Panel genetic database which presently covers over 200 different breeds.

Predicting our results
Both Elroy and Murphy are mixed breed dogs. Here's the list of breeds they say they can detect.  And here's more from the website:

Wisdom Panel® 2.0 breaks down a dog’s lineage in the form of an ancestry tree.  This allows you to see which breeds are present at a parent, grandparent, or great-grandparent level.  Keep in mind that a parent contributes 50% of their DNA to the puppy while a grandparent contributes about 25% of their DNA on average to the puppy.  It follows that a great-grandparent would contribute approximately 12.5% of their DNA to the puppy on average.
Since each of these different levels can contribute different amount of DNA to the puppy, you can see a variety of influence in the puppy’s physical and behavioral traits.  With a parental breed, you are likely to see some physical and behavioral traits from this breed represented unless some of the genes are recessive (requires two copies of the gene variant to show it).  Examples of recessive traits include longhair in most breeds, a clear yellow or red hair coat, a brown or chocolate hair coat, and prick or upright ear set (e.g. like a German Shepherd Dog).  You may see traits from breeds at the grandparent level and it becomes less likely to see physical and behavioral traits from breeds at the great-grandparent level unless those traits are dominant (requires only one copy of the gene variant to show it).  Examples of dominant traits include shorthair in most breeds, black hair coat, black nose, a drop or down ear set (e.g. like a Beagle), and merle/dapple (e.g. like a Australian Shepherd or Great Dane).

For Elroy (85 lbs)
Our guesses = 50% sharpei or chow chow; 50% rottweiler

Kevin was told he was half sharpei and half rottweiler when he adopted him and all his litter mates were black and looked like rotts. He's got tiny ears and a huge square head and heavy neck relative to his body.

We started to wonder whether he was chow chow instead of sharpei when I checked my best dog reference for another breed with a black tongue.

If he's part chow chow then my envy of his fur is no longer so crazy, since the breed was long made for that... and for dinner too. It makes me chuckle every time I call Elroy's dinnertime, "chow time."

For Murphy (40 lbs)
My guess = 25 % German shepherd; 25 % Border collie; 25 % Hound of some kind; 25% unknown village dog. Kevin's guess = 25% German shepherd; 75% Border Collie

Before her greybeard took over.

Unlike for Elroy, Murphy's behavior came into play for these predictions. Her main occupation is to herd each car that comes down the lane along the edge of our lot. She also stalks squirrels and chases shorebirds. And compared to Elroy, she doesn't have as many breed-specific morphological traits.

Who knows? We could be way off. After all, I just took this Dog Bark Interactive Quiz and failed miserably despite knowing exactly what Elroy and Murphy's vocalizations mean.

Results and analysis, right here, tomorrow...