When computing the coefficient of inbreeding from pedigree data, how many generations of data should you use? 

Ask ten breeders and I expect you will get at least five different answers. Many just do as they see others doing, some only have limited pedigree data and they use whatever is available, and often breeders rely on websites that provide COIs based on the data in that site's database, but for a modest number of generation (usually 8 or 10). I suspect lots of breeders don't know that the number of generations used actually matters.

Well, does it?

The coefficient of inbreeding is probably the single most useful statistic for animal breeding. It tells you about the degree of homozygosity due to inheritance of two copies of an allele from an ancestor on both sides of the pedigree. The key thing to know about this is that homozygosity matters in animal breeding. For many genes, the heterozygous state of an allele (e.g., Aa, versus AA or aa) is the most advantageous, a situation called "overdominance". Inbreeding results in loss of heterozygosity, producing homozygous genotypes, either AA or aa. Consequently, the advantages of heterozygosity are lost, and the resulting deleterious consequences for function are known as inbreeding depression.

Animal breeders documented the negative effects of inbreeding depression over a century ago. They also realized that outcrossing improved inbreeding depression by reducing homozygosity, the benefits of which are called "heterosis" or hybrid vigor. Awareness that the level of inbreeding in their animals affected their profit, animal breeders supported the development of the coefficient of inbreeding by Sewell Wright in the 1920s, which could estimate the level of inbreeding of any animal from pedigree data. To this day, COI remains one of the most powerful tools in the design of breeding plans that balance the negative effects of inbreeding with the benefits for consistency and quality.

When COI is computed from pedigree data, the quality of the database is critical. COI cannot predict inbreeding due to an ancestor that is not in the pedigree database you're using. In dogs, much of the most significant inbreeding occurred very early in breed development. Many breeds also suffered population crashes during wars, especially WWII. In small populations, it's difficult to avoid breeding, so these events define the genetic resources of the breed forever into the future as long as the stud book is closed. The dogs we have today carry the genes passed down from the animals before them. Animals lost to breeding because of a bottleneck are genetically irrelevant. If we are interested in the genetics of today's dogs, only the offspring produced that carried on after a bottleneck matter. The gene pool then is as large and diverse as it will ever be, and if the stud book remains closed, some of that diversity will be lost every generation.

Both history and genetics indicate that the number of generations of pedigree data used to compute COI should matter, and indeed it does. But the number of generations used for COI in dogs is both variable and seemingly arbitrary. If we are interested minimizing the deleterious effects of inbreeding, then we should be working with COI data that tell us about actual levels of homozygosity, not relative to other dogs or by a subjective opinion of what is "good" or "bad". If we want useful information about inbreeding, we need to be producing COI values that inform us about actual levels of homozygosity.

[This is a bit of an aside, but it is crucially relevant here. Ideally, breeders should have pedigree data that are complete back to founders. By some oversight of breed or kennel clubs, most breeds do not have a complete database back to founders. This prevents breeders com computing what would be the most accurate estimate of inbreeding using data for genetic history. This is a problem we could (and should) solve, and it should be a priority for all stakeholders. Certainly, the kennel clubs could do a lot to help out this with, since they maintain the actual stud book. There is no benefit in blocking access to this information, so we have to ask why they do this. I'm sure many would urge the kennel clubs to get with the program and unlock a hugely valuable resource so breeders - and others including researchers and scientists - can get access.]

So, if we consider the nature of pedigree data (a record of ancestry and population history) together with the need for information about consequences of breeding strategies for population and individual genetics, the depth of pedigree data used in COI calculations should be based on things that matter for genetics, not some arbitrary number of generations. 

When a complete pedigree database back to founders is not available, I propose that the most useful, non-arbitrary pedigree depth that should be used in estimates of inbreeding should reflect the timing of historical bottlenecks or the lowest historical population size for more recently recognized (or developed) breeds. This pins down a population of dogs (and their genes) as known ancestors from which all subsequent descendants descend. Breeders need to realize that the COI computed from this population assumes that the oldest generation (our bottleneck population) is comprised of dogs that are not inbreed and not related. This means tht the first generation of descendants from those dogs will have computed COIs of zero. For most breeds, we know this isn't true, but this is the same assumption used in any calculation of COI from pedigree data. What this means, however, is that the COI we calculate from these data represents the inbreeding that occurred from that generation to the present. This defines a specific pedigree depth for calculations of COI. Because many breeds went through bottlenecks at the same time, it allows for comparison of historical rates of inbreeding across breeds. 

The other problem this specific "bottleneck" generation for calculation of COI solves is that of the "sliding" COI. If you use some arbitrary number of generations in calculation of COI, as would be the case for online databases and for many breeders that maintain their own databases (e.g., 5 or 8 or 10, for example), adding the most recent (new) generation to the database bumps off the oldest generation to keep the number of generations the same. The calculation of COI must include the historical inbreeding, so cutting off the bottlenecks and other generations where there was lots of inbreeding will result in lower COIs in today's dogs. Remember, the COI calculation assumes the oldest generation of dogs are not related and not inbred, which isn't true. Generation after generation, as you lose the early generations of inbreeding, it will look like inbreeding is going down in your breed - which of course it isn't - leading to the erroneous (and oft claimed) conclusion that "breeders are doing a great job at reducing inbreeding". (If you didn't know this about calculating COI, you need to dip into ICBs's FREE online course, COI Bootcamp. You don't know what else you don't know, and you should be able to see how making a simple mistake can result in nonsense when in fact you think you are being careful and responsible. The course is FREE. Just do it!)

We should be using historical bottlenecks as a "starting generation" in calculations of COI from pedigree data. It would be more informative to have pedigree data back to founders, but until the Kennel Clubs of the world decide that the very future of the breeds they register depends on breeders making smart decisions that will prevent further deterioration of gene pools. With a fixed starting generation, we really can see if breeders are adopting breeding strategies that are reducing the rate of inbreeding or, in the case of breeds undergoing genetic rescue, breeders can monitor progress without a large expense for DNA analysis.

Again, until Kennel Clubs step up to provide the data, it is up to breeders to work on creating databases complete back to bottlenecks, but they would gain a valuable tool for making breeding decisions that will reduce the risk of inbreeding depression and protect the health of the gene pool.

So, what is the source of the problem? Animals in closed populations can only breed to their relatives. All breeding to related dogs is inbreeding. Inbreeding produces homozygosity - two copies of the same allele at a locus. This is a good thing for the genes for type. It's a bad thing for genes that are broken. A genetic disorder is not caused by mutation bombs that you can simply remove to restore health. A genetic disorder is what results when a dog does not inherit a copy of the allele necessary for proper function. So, we have some loci that have two copies of a good allele; but for all the loci that have two copies of a broken allele, something function will be broken.

We are trying to produce healthy dogs by throwing mutations out of the gene pool. But it's a closed, finite gene pool; eventually we will throw all of the genes out. In fact, animals in closed populations go extinct.

Again. Animals in closed populations - aka closed gene pools - go extinct.

There is no "breeder magic" that will prevent this. There is no "science magic" that will prevent this. Animals in closed gene pools go extinct. Some sooner, and some later, but inbreeding will relentlessly increase over time, and diversity will decline, until so much stuff is broken that the animals can no longer reproduce and survive.

All of the other things breeders usually discuss really aren't relevant to addressing this overarching, unavoidable problem. Should we worry about hip scores, or is longevity more important? What about an eye problem that has a late onset? What about mutations with only mild effects? There is lots to talk about, and discussions have continued...for years.

But here's the only problem we need to talk about: inbreeding and loss of genetic diversity.

Fix this problem and you will have healthy dogs. If you start with a population of healthy dogs and randomly remove 40% of the alleles the breed started with, you will most certainly break things. This breed's average inbreeding is more than 40%; if half of that (20%) is homozygosity for good genes, then 20% of it is homozygosity for bad genes. That's a lot of stuff that's broken.

You cannot select your way out of this problem; remember, selection removes alleles, and lost alleles are the problem. It might be possible to restore some lost genetic diversity by strategically using less closely related dogs for breeding. Genetic analyses can reveal if this is possible.

We have much better tools to guide breeding decisions now than simply looking at stacks of pedigrees and comparing health issues. At the very least, you should be using those. You should know the heritability of all of the traits and disorders under selection (0.06? 0.33? 0.89? You should know the size of your gene pool (is it 57, 18, or 6?). You should know the effective population size of the breed (504? 92? 4?). You should know the pairwise kinship of the breeding dogs in the population; the inbreeding data suggest that the dogs are on average as closely related as what you would get from 3 or 4 consecutive full-sib crosses. Would you ever do 3 or 4 full-sib crosses???? In terms of genetics, that's what you have. You need to know which dogs in the breeding population have the highest genetic value so you can be sure to breed those, and which have the lowest value so they can be retired. You should know how much improvement in all of these things is possible if the existing genetic diversity in the breed is used in the most strategic way. If it turns out that this will not be adequate to restore the breed to health, then you need to evaluate strategies that will.

These are things you won't learn about in 20 or 30 years of breeding. You probably don't know anything about effective population size or kinship coefficients or founder genome equivalents. These are not things you will learn by breeding. These come from the science of population genetics that has been developed over the last 100 years by study of thousands and thousands of breeding programs for both domestic and wild animals. These are the tools used by breeders of other domestic animals. They are used in genetically managed programs for service dog breeding; they work for dogs just as they do for any other animal.

To solve the problems in this breed and in purebred dogs, we will have to correctly identify the cause of the problem (inbreeding and loss of genetic diversity), determine the best strategies for addressing the problem, and design a breeding strategy to effectively and efficiently restore the breed to health.
​
We have the tools and expertise to do this. We could be doing this NOW.

The Importance of Genetic Management

​The goal for every breeder is to leave their breed healthier and stronger for future generations. But without the right tools, achieving these goals can feel like navigating without a map.

The biggest challenge for breeders of purebred dogs is high levels of inbreeding and the consequences for health. We know that inbreeding has multiple deleterious effects in animals. It increases the risk of recessive genetic disorders being expressed; it causes inbreeding depression, which also affects health; and it results in loss of genetic diversity, which ultimately increases the rate of inbreeding.

Purebred dogs as a group have exceptionally high levels of inbreeding, a consequence of closed stud books that prevent the introduction of fresh genetic diversity to restore that lost over the generations. In addition, it is difficult for breeders to implement the types of genetic management strategies that would control inbreeding and loss of diversity because they simply lack the necessary information. Creating a genetic management plan requires information about the "genetic landscape" of the entire breed - not just individual dogs, or average values for things like inbreeding and kinship, but also data for the variation in these values in the population.

What's the Problem?

​Breeders are recognizing that breeding for health requires that they know more about the dogs and the breed than they did in the past. DNA testing and pedigree analysis can now provide data about the genetics of individual dogs that can be used to reduce the risk of genetic disorders in offspring. What breeders lack, however, is information about the genetic landscape of their breed. When breeders identify a potentially genetic problem, they will try to "breed around" it or remove carriers from breeding. These strategies assume that the genetic solutions they need exist in the breed and they can solve the problem by moving the breed in that direction.

Imagine you are a tourist planning a trip in a place you have never been before. Without a map, you can only navigate by guessing, and you could easily end up in the wrong place or, at the very least, waste lots of time and energy taking wrong turns and running into dead ends. If you don't have the right information, breeding decisions work the same way. Without knowledge of your breed's genetic landscape, you can't plan a path forward or determine if it's even possible to get where you want to go. 
​

​The problem here is the assumption that a solution exists within the breed. Remember that recessive mutations cause problems when a dog inherits two copies of the mutation. It's not the presence of the mutation that you have to deal with, but the absence of the "wild type" (normal) version of the allele necessary for the gene to do its job. But breeders focus on mutations, and they diligently try to solve the problem through selective breeding. This takes an ongoing toll on the gene pool, ultimately increasing the risk that some other defective allele will become the next problem in the breed. You can see how this turns into a cycle of genetic whack-a-mole, damaging the gene pool with every cycle while making no progress on the actual problem. This is where we are in most breeds after decades of selective breeding, now guided by DNA tests. We have failed to improve the health of purebred dogs because we have not determined the right path to health. Without a map, we don't see a destination and can't choose a path that will get us there. We invest time, energy, and money, all with fingers crossed, hoping at least that things won't get worse. 

What's the Solution?

The solution to this problem is obvious. Breeders need more information about the genetics of their breed. Not just about one or a few dogs, but about dogs characterizing the breadth of the breed's genetic landscape. Breeders need to be able to determine - before they hit the road - that the chosen breeding path will take them where they want to go. They need also to adopt breeding strategies that will not work against them along the way. Simply removing from the gene pool any dog that has some issue, or restricting breeding to just a small fraction of the dogs produced, will perpetuate the situation we have now and foil attempts to improve it. We need to be able to identify dogs of greatest genetic value so they can be prioritized for breeding. We need to identify sires before they have litter after litter of puppies that will skew the entire gene pool in one direction and flood the breed with his unique assortment of genetic mutations. (See Pox of the Popular Sires) 

Breeders - and breeds - need is a road map of the genetics of the breed across its entire scope, so breeders can design well-planned solutions to problems. We need genetic management plans that identify a path towards a solution, instead of trying to solve problems by trying to run away from them in some random direction.

We have the information we need to create at least a basic plan for most breeds. Pedigree data and individual genotypes can be leveraged to start filling in the blank areas of the genetic landscape with useful information - where to find dogs with useful genetic diversity, where to find outcross candidates for specific dogs, which sires are overproducing at the expense of other genetically valuable dogs, and more. The information breeders need can be extracted from pedigree and DNA data and used to address the questions breeders ask when making breeding decisions. With regular updates, this information resource can display the current genetic status of both the breed and individual. It can also document the progress resulting from breeding strategies designed to reduce inbreeding and protect and improve genetic diversity.

DogsArk: The Genetic Dashboard

The good news is that we CAN get there. We can improve the health of purebred dogs without sacrificing the traits that make each breed unique. And we can do it efficiently and effectively, with tools and expertise that are available to breeders NOW. ​

ICB has built a breeder tool called DogsArk that provides the information breeders need for sound genetic management. Using either pedigree or DNA data, DogsArk provides a "genetic dashboard" that allows you to -
​

• Visualize genetic diversity: Identify where genetic diversity is strong and where it is at risk;
  
  
• Track lineages and traits: discover genetic clusters and understand the distribution of traits and mutations;
  
  
• Plan sustainable breeding programs: Use real-time data to make informed decisions that preserve your breed's genetic health.

For the first time, we now have the DogsArk Breeder Tool, which will provide the information needed by dog breeders to safeguard the genetic future of their breed. With DogsArk, we can start to plan breeding strategies that will improve the health and welfare of dogs. (We are in the process of adding breeds and data, and the site is still under construction - pardon our dust!)

Have a look at DogsArk and check out the tutorial for guidance. If you have any questions or would like to add your breed to the site, just drop me an email: [email protected]

Facebook is an echo chamber of misinformation, myths, and unfounded opinions about the coefficient of inbreeding, which is unfortunate because there is probably no statistic (other than the kinship coefficient, which is related) that is more useful to breeders who want to minimize the risk of health problems in their puppies.

Here's my list of the 10 really essential things you need to understand about the coefficient of inbreeding. 
​