I have no doubt that inbreeding leads to problems and that loss of genetic variation is a problem in dog breeds.

However, Brenda points out that the genetic COI (or genomic COI) are not pedigree-based but then continues to say that a 25% COI (F) is what you get from a full-sib mating. That is correct when you talk about pedigree-based F. For the genetic COI from Embark we do not have a benchmark. We know that they use runs of homozygosity (ROH) and it seems that they allow also very short ROHs. This increases the general level of COI. Going to the extreme limit, you come down to single-snp based COI, i.e., overall homozygosity. This is no longer a measure of identity by descent (IBD) but rather identity by state (IBS). Now, remember that with only 2 alleles the proportion of homozygotes (if we have Hardy-Weinberg equilibrium) is p² + q², where p and q are the allele frequencies. If they are equal, we have 50% homozygosity. If allele frequencies are unequal, the homozygosity increases. If minor allele frequency is 0.25 then the expected homozygosity is 0.625 (without inbreeding). 

So, presumably, the genetic COI are somewhere in-between pedigree-based IBD inbreeding coefficients and the IBS homozygosity F, but where on the spectrum is hard to know. 

As an example, in a snp-dataset I have available, with settings that only allow rather long ROH (hardly any ROHs shorter than 4 mill. base pairs, average of 6.7 Mbp) I get an F_ROH average of 0.05. If I allow shorter ROHs (average 1.7 Mbp) I get an average F_ROH of 0.16 for the same group of animals. (Total homozygosity is 0.65)

I think it would be very useful to benchmark the genomic COI (from Embark or elsewhere) against known pedigree-based average relationships and total homozygosity before we draw the conclusion that all German shepherds, say, are full sibs in the IBD sense.

On 5/5/2022 at 7:48 PM, Erling Strandberg said:

I have no doubt that inbreeding leads to problems and that loss of genetic variation is a problem in dog breeds.

....

...I wonder if they calibrated the minimum ROH length used in their F_ROH estimate by looking at known relationship breedings (They use 1 MB if I understood the material on their site correctly).  A n=1 but, I know of 1 offspring from an accidental father-daughter pairing (both mixed breed dogs) that came back at 26% , so very close to expectation.