The English blog hasn’t really gotten much love, so here we go with something very simple that is often on people’s minds: inbreeding calculation, COI, coefficient of inbreeding. A big deal and an important tool especially when managing a small breed like ours, but one that is often used a little bit inefficiently when considering its full potential.
First, to start this off right we’re going to define what coefficient of inbreeding actually means. We get 2 definitions:
The probability that two alleles at a given locus are identical by descent.
A measure of the proportion by which the heterozygosity of an individual is reduced by inbreeding.
So for 10% inbreeding, we’re talking about a 10% probability that any given allele at a locus is identical by descent, meaning its not only the same type, but also came from the same ancestor. Further, for 10% inbreeding the heterozygosity of the individual has been reduced by 10%. Heterozygosity is translatable to genetic diversity, something that is required for a healthy, vital population. Low heterozygosity / low genetic diversity is furthermore linked to health problems like reduced fertility, higher puppy mortality, shorter lifespan, smaller adult size, deteriorating immune systems (vulnerability to immune-mediated illnesses) etc.
COI is not just a number, it’s a tool that can tell us a lot of useful things in breeding:
- It’s half of the kinship of the parents. So if a dog has 25% inbreeding, the parents would have been 50% related (siblings, on average are genetically 50% alike).
- It can be used as a risk % for deleterious alleles (genes that cause harmful traits): in 10% inbreeding, all alleles have a 10% chance of being homozygous, and 10% of all alleles will be homozygous. Even the ones creating diseases and other problems. 10% is generally considered the limit before problems become blatantly obvious.
- Because it gives an estimate on how much heterozygosity has been reduced, it can be used to give an estimate on genetic diversity. One way is to take GD data measured directly from the genome, and compare it with COI. I will show you how later in this post.
- If you know the mean kinship of your dog (its average relatedness with the whole breed), you will know the average inbreeding of it when combined with every dog in the breed. You can make intelligent breeding choices to help the breed when MK & COI are used together. I hope I have time to write about the magic of MK in another post soon.
All of these above mentioned things are great and give us lots of tools to use to improve both our individual litters of puppies, but also to better the breed as a whole. But there is one twist: they only work if you have the full data, or something as close as possible to it. So none of that “my dog’s inbreeding is 1% in 5 generations”, unfortunately, you can’t do anything with it. Currently the breed has about 20 generations of data, which makes 5 generations 1/4 of that. You can’t go ahead and make predictions or breed-benefiting decisions with just 25% accuracy.
Let me tell you a story. Back in 2006 or so when I fired up my first own database to record info on my breed, I bought the desktop version of a software called ZooEasy. Very soon I came across a problem that made me contact customer support: there was no way to choose how many generations I wanted the COI calculation to take into account. All I got was these weird numbers that didn’t correspond to what I knew was “good” or “bad”. The reply from the customer support was equally confused: why on earth would you want to limit it? You’re supposed to use all the data available, what’s the point otherwise? Soon after they did add the option to limit calculation with any number of generations (I guess I wasn’t the only one with the same problem), but what they said sat with me and it didn’t last me long until I took that new feature out of use completely. I learned to look at the numbers as is and I started to understand “good” and “bad” on a scale offering me a wealth of new information.
So what are the reasons we do tend to limit the calculation for inbreeding?
- Wright’s Coefficient of Inbreeding was “invented” in the 1920’s. Back then, and for at least 80-90 years forward from that breeders did not own a computer that could have made the calculation for them. It’s possible to calculate 3-5 generations by hand, but after that it becomes increasingly difficult and eventually impossible. It was only natural to talk about levels of inbreeding on a scale that was more accessible for everyone.
- It’s a common idea in dog breeding that you will get all the knowledge you need about the pedigree if you know the dogs in 3-5 generations. There is really no use to look beyond that, since pretty much all the qualities you can expect in a litter should be present somewhere within those generations. It’s solid advice when considering traits, but somehow, this idea has been expanded to mean inbreeding calculation as well.
I personally believe these are the true reasons. There are a lot of other arguments, but after I started listing them here it felt like my post was turning into a “trump that argument” type of listing that I don’t want to do. Instead I’ll try to explain here why it’s finally time to leave old habits behind and start expanding our horizons. 🙂
And I’m going to use examples, because I love examples. First I’m going to pick some longhaired litters born in the last 2 years, and put their COI values on a graph. This is something I do pretty much always when I talk about inbreeding, so some of you who have followed my stuff for a while might already be sick of this one. I’m not listing my own litters, and I’m not telling which litter is which, all I can say is that none of these figures are extremely rare. Most commonly we limit inbreeding calculation to 5 generations (and many breed clubs have specific allowed & not allowed levels set for this number of generations), so I’m going to always use this as the strictest example of limiting, and full data for full data.
Five litters were selected and their inbreeding plotted into the chart below.
I didn’t manage to pick as clearly contrasting cases as I could have, but a few interesting things can be seen here. For example, it seems impossible to predict the actual inbreeding from 5 or even 10 generations, and this is exactly why no tools relying on COI calculation can be used with such low accuracy data. Also, we can see that inbreeding can never go down with more accuracy, only up, or stay the same when no more data from further generations can be obtained. As a breed-specific trait for the Dutch Shepherd it could be said that around 15 generations should be enough data to at least see realistic rankings of high and low inbreeding litters. The last 5 generations bring some fractions on top of that, but not enough to change anything majorly.
What’s pretty much as important, or some would say even more important than the level of inbreeding itself, is the rate of increase in the inbreeding value:
|Litter||gen 5||gen 9||gen 13||gen Full||Increase/gen|
It is estimated that in short term nature can counteract the harmful effects of inbreeding by only 0.5-1% per generation. Higher rate than this means that genetic variation is lost too rapidly. Knowing this, it’s fair to say that any Dutch Shepherd with more than 20-25% inbreeding has surely lost genetic variation too fast, in addition to the vitality loss coming with a flat COI of 20-25%. Notice that even the least inbred litters on this list are pretty much equal to full sibling pairings! The most inbred litter on the list with 41% COI is truly in a terrible situation, but the situation would be much less dire if it was 41% COI in 40 generations instead of 20. Unfortunately, we can only calculate COI as far as our pedigrees go.
Following the idea that >1% COI increase per generation is just too much, we could go ahead and say that in 5 generations 5% COI should be the limit for any dog. But this would be a very situational guideline, since no matter how you put it, Litter 2 with 0.88% COI in 5 generations (but nearly 40% homozygosity) here is much worse off than Litter 3 with 8.59% COI in 5 generations. Short-term inbreeding can actually even be used to lower long-term inbreeding, but I won’t go into it here too much. The basic idea is that breeders can inbreed rarer families together to create dogs that have a low kinship to the whole population (= mean kinship) and this way will create low-inbreeding combinations with almost any other dog in the breed. For example Litter 3 in our example has high 5 generation inbreeding, but a low true inbreeding. It doesn’t say it in the table, but the litter also has a low kinship to the breed, because the short-term inbreeding happened with dogs whose pedigrees are not over-represented in the breed yet. By pairing any dog from this litter with for example any dog from Litter 1, it would be possible to create a combination that would have 0 inbreeding in five generations, and significantly lower than 20% in all generations; an improvement from both parents.
Another limit that could be imposed on breeding would be the 10% that was mentioned at the start of this post. After 10% of inbreeding we’re starting to see signs of increased health problems and lowered vitality. The problem with this is that we would have to stop breeding this breed entirely, since only combinations with a lot of off-variety blood actually reach this “low” level of inbreeding. Even inter-variety crosses to traditional shorthair pedigrees are close to 10% due to several crosses done already. I could have picked a few litters with <10% into my example above, but I chose not to do it to not give the idea that these kinds of combinations are somehow commonplace. A combination like this is always the result of a cross to another variety or breed, and if you make one, you’ll know it.
At this point I could say that most of this post here was written with the longhaired population in mind. The most inbred shorthairs are around 25% COI and working line dogs at much, much less than that, very often <10% COI. The application of everything I say here goes for the shorthairs as well, but the actual numbers aren’t exactly the same. Next I’m going to show the link between COI and GD, and here shorthairs have played a major role as well, since otherwise I wouldn’t have had so many low COI dogs for the graph.
Inbreeding coefficient in itself is just a statistical estimate, and even dogs from the same litter can have different levels of heterozygosity. Today we have many genetic test panels like MyDogDNA & Embark available that will give evaluations on the genome-assessed inbreeding or genetic diversity as well. Since there is a lot of public data on the GD measured by MyDogDNA, I’m going to use it to show you how it correlates with the COI calculation.
Here is GD% data (y) plotted on a graph together with COI% data (x) in 5 generations:
There is no correlation. If you stick a trend line here its accuracy is close to zero. There are low GD dogs in both low and high inbreeding individuals. What this tells us is that there is supposedly no way to tell anything about the genetic diversity of an individual with COI calculation, even if one of its key features is supposed to be giving an estimate of how much heterozygosity was lost!
But let’s not fret, since here is the same GD% data (y) plotted on a graph together will full COI% data (x):
The accuracy of this trend line is very close to a perfect match, and you would be able to see the trend clearly even without the line set in the background (it’s there because I like how beautifully it fits 😀 )! So on this graph the calculated inbreeding value and the genome-tested genetic diversity value correlate with each other strongly. If you know the full COI of your dog or upcoming litter, you can get a really good estimate on the expected genetic diversity as well just by seeing the data collected from other dogs. Your upcoming litter has 15% COI? You can expect a GD of around 33-38%. Conversely, you can get a really good estimate on the full inbreeding value of your dog from a test like this. Your dog has 25% GD in the MyDogDNA panel? Most likely their real COI is around 40%. The more dogs are tested with these tools, the more accurate these predictions will be able to be. So if you ever do test your dog, please make the profile public.
Using pedigree data and the inbreeding coefficient as tools it’s not hard to make decisions that can help the whole breed. But we must have all the data. Imagine this scenario: You want to help the breed and reduce the rise in inbreeding with your upcoming litter, so you make sure the COI of your litter is lower than the average of the parents. Let’s say the dam has inbreeding of 3% COI (5 gen) and the sire has 2% (5 gen), creating a combination with 0% COI (5 gen). You truly believe you just did your part in reducing inbreeding in the breed and feel really good about it. What if I told you just paired a 38% dam with a 39% sire and created a litter with 40% COI? What kind of compromises did you make in selecting breeding animals to try to achieve that 0%? Compromises in character, or in health? How would it feel if you actually had rejected a better individual for your female just because you made your selection based on inaccurate data? Not good, I can believe.
If you recognize yourself from that paragraph, I can tell you one thing: you are definitely not alone. The average COI in the breed is easily above 35%. And even 10% is too much! Instead of putting empty effort into making selection based on partial calculations, we should focus on getting the real COI down now. Fortunately, with the right selection it is possible. Go ahead and grab some software that can do the calculation for you, and start assembling your database. When you have all the data, you will start seeing the breeding values of dogs in a very different light. Welcome to the bright side. 😉
// Disclaimer: the numbers in this post are from my personal database, and they cannot be compared to numbers from other databases. Here as well, it all depends on how accurate the data is.