Sunday, July 29, 2018

Being Very, Very Distantly Related Is Not inbreeding



A Facebook friend was told by Embark Veterinary that a Mastiff and Greyhound cross had a COI of 2% based a few overlapping segments of mDNA.

No, that dog does not have a COI of 2%. Embark is entirely wrong.

COI stand for "Coeffcieint of Inbreeding" and it is a mathematical construct, developed by Sewell Wright, based on pedigree.

Embark was not using pedigree, but mDNA which is far from the same thing as the video, at top, suggests. Small segments of mDNA overlap are NOT how you calculate COI.

So what's the story?  It's not complex.  Dogs have not been around very long; maybe 40,000 years with very, very small, but continuous, shots of wolf.

Breeds, as we know them today, are only 150 years old, and most types are only about 600 years old. The oldest type (greyhounds) is perhaps 6,000 years old.

Greyhound mDNA is going to show up in the background of almost every mid-sized to large dog as it’s in the root stock of so many breeds.

But is a tiny scrap of vestigial mDNA the same as inbreeding?


Nope.

To be clear, Coefficient of Inbreeding is not a perfect metric.

If you have only 30 dogs, close the gene pool, and breed forward from there for 100 years, resulting in a pool of 250 dogs, a 5-generation pedigree will obscure the amount of true inbreeding.

But if you do the same thing over 150 years, with a breed pool that ends up with over 200,000 dogs, that 5-generation COI number is going to be a bit more useful. Why? Because with greater population size you get both more genetic drift AND you also get more folks slipping in dogs from other lineages (i.e. faking pedigrees). Faked or contrived pedigrees are terrific as they enter a lot of genetic wobble into the mix.  In fact, they are why puppy mill Scottie’s are healthier than those produced by so-called “good breeders”.

The fact that science can show that a few short scraps of MDNA are common to both a sire and a dam does NOT mean that any meaningful level of inbreeding has occurred, and it certainly cannot give you a COI.

4 comments:

Glenn Dorsey said...

Your genetics is a bit shakey here. I’m not an apologist for this breed make-up of random dogs which brings in some cash for Embark but their SNP/Haplotype reports of COI (more appropriately named IBD (inherited by decent) is viewed as a legitimate metric by researchers. As one who is learning to analyze the raw data on over 200,000 SNP’s provided by Embark for download on each dog analyzed by Embark, I suggest you go deeper in this science before you claim Embark is wrong.

Glenn Dorsey

PBurns said...

COI means something. Please read up on Sewall Wright and how this metric was and is actually calculated. COI is NOT properly calculated based on SNP/Haplotype reports. In fact, when Wright created the metric, such things did not exist.

PBurns said...

"Coefficient of inbreeding (COI), also known as coefficient of relationship, is a method of evaluating the degree of relatedness of two individuals based on pedigree records, not DNA testing. As a result, the COI is exactly the same for all puppies from a particular mating. Although genetic recombination in mammalian breeding is unpredictable, COI does not take into consideration the actual DNA inherited by an individual or variation between individuals within litters, nor does it correct for inaccuracies or unknowns in pedigree. DNA testing for diversity is individual-specific and can therefore be more sensitive, identifying opportunities for gains in diversity that could not otherwise be discerned by COI calculations."

From >> https://help.wisdompanel.com/s/article/How-is-Optimal-Selection-s-Genetic-Health-Index-result-different-than-coefficient-of-inbreeding-or-haplotypes

PBurns said...

"…. With the advent of high-scale genotyping technologies such as single nucleotide polymorphism (SNP) arrays, it is possible to estimate realized relatedness directly from molecular data without knowledge of genealogy, and a variety of ways to do this are available [8, 9]. These generally take the form of genome-wise averages of single-SNP statistics, which have the disadvantage of not taking the lengths of genomic regions shared between two individuals into account."

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4642635/