Genetics Question

[bunn]

I have been asked to find a geneticist and ask for an opinion on something. Can I just find someone's email address and bother them that way, or is there a better way?

What I want to know is whether it is possible to determine definitely from DNA analysis what breed a dog belongs to.

the people at http://www.deednotbreed.org.uk/ are concerned about 'dog amnesties' where people are encouraged to hand in their dogs and have them killed FOC, and also about some of the dogs that are being seized.

The police say that (among other things) DNA tests are used to prove the dogs are of a banned breed. They would like to know if this technology is actually available and likely to be being used.

I have found a recent paper that seems to suggest it is, plus an abstract of another - but also much material that says it isn't possible. I am not sure I am reading them properly.

If the technology is available, in use and reliable, that will be very reassuring for people who can have their dogs tested and prove they are just random crossbreeds, and who also will then know that if their dog goes missing and ends up in the pound, they stand some chance of getting it back.

It won't help those poor souls who have accidentally adopted an unidentified pitbull, but at least they will know where they stand.

If the technology is of dubious reliability, then it would be useful to know that too. At the moment basically the only option is to take the authorities at their word, and there is a lot of distrust about the judgement process.

People who lose dogs have started reporting them as breeds that they don't belong to (eg a mastiff x was reported as a lost French Bulldog!) because they are afraid that they might not get their dogs back. This greatly complicates the effort to find the owners of lost dogs.

I want to get an authoritative opinion from someone who is independent, knowledgeable, and ideally, prepared to be quoted. I was thinking of contacting the authors of some of the papers I found: does that make sense?

Comments

[pellegrina.livejournal.com]

I think at work we have a directory of specialists to be consulted for evidence in court cases, I'll have a look tomorrow if there's anything useful.

[bunn.livejournal.com]

Thanks, that would be very helpful.

[pellegrina.livejournal.com]

Annoyingly, the White Rabbit has discarded our directory of expert witnesses! But I'm going to see if I can track it down in another library service who might be able to answer by phone/email.

[pellegrina.livejournal.com]

One of our law directories has a small Expert Witnesses section, with a sole representative under Animal Health with more postnominal letters than nominal ones (DVSc PhD MSc MRCVS FBIAC ECVPath FRCPath). I'll email you the details separately if you're still interested.

[tovaglia.livejournal.com]

Seems unlikely, but then my group works on human population genetics ie polymorphisms (gene variants (alleles) whose frequencies vary between populations) not mutations (a bad gene that causes disease). Here's my attempt at reasoning it out from my knowledge of human gentics (no knowledge of dogs at all, I'm afraid).

I assume that there aren't any breeds that are like diseases in that they are entirely caused by the presence of a single gene variant? (I can believe it with regard to some of the funnier-looking breeds, but not the breeds that the police might be interested in).

Assume for the sake of argument that the police have found a gene variant (which we'll call allele A) that is unique to a particular banned breed of dog, which we'll call breed X. (leave aside the question of how they would validate this: ?by getting cheek swabs from Kennel Club-certified dogs? how many would they need to prove that this mutation is ONLY found in this breed and not in some individuals belonging to the thousands of other breeds that exist?)

A dog that was a cross between Breed X and, say, a labrador, would then have a 50% chance of having allele A. And a dog that was one-quarter Breed X would have a 25% chance of having allele A. And a dog that had Breed X somewhere in its ancestry would have a smaller, but still real chance of having allele A. Therefore allele A would not be unique to Breed X.

So the only way this would work is if there were 2 copies of allele A in Breed X (one on each chromosome) but not in any other breed. But THEN, what if two labrador-Breed X crossbreeds, or other individuals with just 1 copy of allele A interbred? Then their offspring would have a 25% chance of having 2 copies of allele A, but they would obviously still have a lot of "labrador" in them.

The argument could be extended to more than one marker gene using exactly the same form of argument. The only difference is the % probabilities.

In summary, I don't think it would be possible to definitely identify an individual as being pure breed X from a DNA test. However, if you used enough marker genes (the criminal DNA database uses 8 or 10, I think) then the DNA might give you a probability (eg 90%, 99%, 99.9%) that the individual is more than a certain % (eg 90%) breed X. This would depend on excellent bioinformatics and an extensive, well-validated DNA database of thousands of dogs of known breeds. This seems unlikely.

Realistically, though, you could make the odds a lot better using bayesian stats and a subjective visual inspection by an experienced assessor to give a "prior probability" of the individual being of breed X.

So yes, it sounds like it might be feasible in certain situations, but a DNA test by itself would be unlikely to be sufficient.

The problem is rather similar to the one of inferring a person's likely ethnic background from a DNA sample alone, which the police are currently working on using the national criminal DNA database. Maybe they're doing the dog study to validate the human study?

[bunn.livejournal.com]

That's a brilliant explanation - thank you so much.

Would you be willing to let this be reused on their website as an explanation of some of the ideas involved?

[tovaglia.livejournal.com]

That's fine (although maybe not the bit about the criminal DNA database!).
And/Or you could try this version, if you like:

Here you have a molecular test (what gene variants are present?) being used to answer a question at the whole-animal level (what breed is this dog?). No one test alone can ever give an absolutely definitive answer on this kind of question, especially since cross-breeding can blur the boundaries. I think what you'd get would be a result and a probability (eg 90%, 99%, 99.9%, 99.99%, etc) that the given result is the correct one. In practice the important thing is the exact figure of that percentage, which depends on a lot of factors to do with the test and the way it has been validated.

[bunn.livejournal.com]

Brill, thanks.