The Canine Diversity Project

Population Genetics and Breeding

by John Armstrong

Early genetics

When Mendel's work was rediscovered at the beginning of the twentieth century, the new field of Genetics went in several directions. The T. H. Morgan (1) school quickly got tired of crossing green to yellow peas and moved on to discovering white-eyed fruit flies, linkage and genetic maps. The Garrod (2) school started trying to figure out how genes controlled metabolism... and eventually everything else. The mathematically inclined, fearing to get their hands dirty, started thinking about how genes get shuffled around in a population. That led, around 1910, to the famous Hardy-Weinberg formula that relates the frequency of alleles to that of genotypes. For those unfamiliar with it, the basic formula assumes two alleles for a gene and sets the frequency of the first allele as p and that of the second one as q where p + q = 1. In a population that meets certain conditions, the relative frequencies of the three genotypes AA, Aa and aa will be p-squared, 2pq and q-squared. The main conditions are random mating, a large population, and no forces acting to change the allele frequencies.

This does not imply an endorsement of random mating, but was simply a starting point for the eventual development of equations to describe other situations. Most natural populations do not follow these rules. In nature, selection is often harsh, and most animals do not practice random mating. In many species that live in packs or herds, only the dominant male may breed and competition for that spot may be intense. Otherwise, the most common practice is probably assortative mating, where mates are chosen that have similar qualities (size, temperament, etc.) or are not closely related (negative assortative mating). How they decide on the latter is still being determined, but recognition of relatives probably depends on pheromones to a large extent.

Genetics without color

In the beginning, all geneticists held to much the same beliefs, or "model" that there was one, and only one, good (or "wild-type") version of each gene. There were also a few nasty recessive mutants that would occasionally surface. They didn't really expect to find a large amount of diversity for most genes. They lived in a black-and-white world where genes were like light switches either on or off, no in-between. As most of the bad mutations appeared to be recessive, good breeding was reduced to finding ways of efficiently identifying those carrying "degeneracies". Faith in inbreeding as a method for breeding the perfect individual was reinforced by various authors:
"Inbreeding... is a method of holding fast to that which is good and of casting out that which is bad. It establishes homozygous purity..." Onstott (1946)
The morgan and garrod geneticists wanted nice "clean" mutations that were easily distinguished from the wild-type, and a population geneticist would never stoop to thinking about what a gene actually does! Their main concern was to figure out equations that would describe more complex situations involving selection, migration, and mutation, and explain what would happen to a new mutation give certain assumptions. Morphological variants, such as green and yellow peas, were not even really thought of in the same way. I mean, can you really think of a green pea as a "mutant"? (Or is yellow the mutant? How can that be if it is dominant?)

However, by the early 60s, if not before, those on the front lines were certainly aware of mutations that retained partial function ("leaky mutants"), even if they didn't want to work with them. By the 70s, the population geneticists actually started going out into the field and measuring the diversity in populations. They went in with the expectation of finding little difference between most individuals in a population and discovered far more than they had anticipated. The dust still hasn't settled completely. Logic suggested that if there was a considerable amount of genetic diversity, then there should be some reason for it. In a large population, the rare recessive mutation has little chance of gaining a toe-hold, and if it gets to the level where there are a noticeable number of homozygous mutants, selection will do its best to push it down again.

Explaining diversity

Several theories were proposed, but somewhere along the line, the realization dawned that many populations are actually a loose collection of small populations that are semi-isolated. In a small population, random events take over and the frequencies of particular alleles may change dramatically just by chance ("genetic drift"). Given enough time, these random fluctuations generally eliminate all but one allele, which is said to be "fixed". How quickly this happens depends on how small the population is. Unequal use of individuals in the population increases the rate of allele loss because it decreases the effective population size.

Alleles with dramatic effects on viability are still generally selected against, but if the population includes several alleles of a particular gene, the "best" choice will not always be the winner. Sometimes an allele that reduces fitness by a small amount will take over. Over time, a small population may accumulate enough of these sub-optimal mutations for the impact to be noticeable.

Small populations also tend to become unintentionally inbred simply because there are not enough ancestors for each member of the current population to have a unique set (3). Neither intentional nor unintentional inbreeding lead to changes in allele frequencies, unless combined with selection, but they do lead to loss of heterozygosity. The decrease in fitness that results from accumulation of suboptimal recessives in the homozygous state is what we generally call "inbreeding depression".

If there is an exchange of genes by individuals crossing over into another population's territory, the reduction in fitness due to gene loss will be reduced. The populations that we see in difficulty have often been cut off from other populations preventing this essential migration of genes. Canine examples include the Ethiopian and Mexican wolves, and the grey wolves on Isle Royale.

The importance of breed origins

Some breeds of domestic dog have evolved gradually over hundreds or perhaps even thousands of years. As most were bred for a purpose, some selection must have been involved. If you want a dog to guard your sheep, and it fails to do so, you are not likely to breed it. However, if you have two good herding dogs, you might breed them to each other, irrespective of their relationship. All the herding dogs in one valley may have been fairly similar and closely related, but exchanges would have occurred between neighboring valleys. If the population over a more extended region was bred for a common purpose, they might constitute a recognizable group, or breed, and it might acquire a descriptive name -- the Bavarian Sheepdog for instance. Such naturally-evolved breeds would be unlikely to have suffered major drift losses, even if they became locally inbred, because sufficient diversity would exist in the whole population, and there was no reason not to breed to a good sheepdog from another country. Whether one regards stud books and closed registries as good ideas or not, providing that sufficient numbers were admitted, such a breed should have at least started with sufficient diversity.

In contrast, when a breed is deliberately created from a small number of founders, the creator(s) generally concentrate first on inbreeding and selection to define the qualities they are after, rather than increasing the initial population and subsequently selecting for those that come closest to meeting their goals. Such a beginning generally removes most of the genetic diversity in the first few generations. If you have been unlucky or chosen badly, there may be little you can do.

The same fate may befall a naturally-evolved breed ( "landrace") if there is no recognized registry in the country of origin and too few founders are admitted into the registry somewhere else. At least in these cases, the potential exists of petitioning for reopening the stud book and admitting additional "founders". In those cases where there is no such reserve, the solution might be a merger with a closely related breed, or at least provision for some interbreed crosses. There are a few documented cases where this has been attempted in the last 20-30 years, but they have met considerable resistance.

Don't shoot the messenger

Population genetics is not really a new discipline, it just seems that way because it's generally the last chapter in a genetics text. Population geneticists are neither white knights come to save us all, nor agents of the devil intent on destroying pure breeds. Population genetics is a tool for looking at an entire population or breed. It can tell you what has happened to the genetic diversity, and whether there is any possibility of improving the situation by making appropriate crosses. How this information is used is up to the breed club and individual breeders. Though lessons may be learned from conservation biology, I do not expect breed clubs are going to be in a position to manage the entire breed. However, they may choose to limit certain practices for the overall good of the breed. The prime target, in my opinion, should be overuse of popular stud dogs.

In a managed population of an endangered species, zoo biologists might choose one of several strategies that are generally aimed at conserving the diversity from the wild population from which the captive population is drawn. This makes the assumption that all founders were equally meritorious and that their genes are all equally worthy of preservation. This is essentially a holding action and, in the absence of selection, runs the risk of creating a population that is less well adapted to returning to the wild.

In my view, the best strategy for dog breeders is carefully planned assortative mating combined with an attempt to minimize or at least reduce the inbreeding coefficient. In practice, if I am asked for an opinion on a suitable mate for a Standard Poodle, I suggest that the breeder assemble a list of dogs he/she would consider breeding to, based on conformation, temperament and whatever other criteria are deemed relevant, and I will tell them the inbreeding coefficient for each potential litter and also about the prominent ancestors in the pedigree. My personal criterion is a 10-generation COI under 10%, but I might pick one close to that, or even a bit over, if I liked the other qualities.

The COI has predictive value. I can tell you that an SP inbred to only 5% will, on average, live about 3 years longer than one bred to 35%, and I can tell you that a 10% increase will likely reduce litter size by about 7%. Both these effects are, in my opinion, most likely to result from accumulation of suboptimal alleles with small individual effects. However, inbreeding also increases the probability of doubling up on any obviously deleterious traits carried by a shared ancestor. I understand why breeders inbreed (or linebreed), but I don't agree that it is necessary to produce good dogs (see Inbreeding and Diversity). As to the claim that it can be used to uncover problems in the line, I agree, but I can also give you case histories where the breeder has proceeded to ignore a hereditary problem uncovered this way, and as a result spread it through the breed.

Neither population genetics nor modern DNA technology is going to provide magical solutions to all our problems. However, used together, they may take us through the 21st Century. Continued reliance on the models put forward in the early days of genetics almost certainly will not.


  1. Thomas Hunt Morgan (Nobel Prize, 1933) and his group provided the first evidence that genes were located on chromosomes, and showed how they could be mapped.
  2. The first clearly-described relationship between genotype and metabolic deficiencies is credited to Sir Archibald Garrod, an English physician. (see The Nature of Genetic Disease)
  3. The theoretical number of ancestors increases exponentially: 2 parents, 4 grandparents, 8 greatgrandparents, and so on. By time you have reached the 10th generation, you need 1024 in that generation, or 2046 counting all the previous generations. The highest number I have encountered in a purebred predigree is just under 1000 for all 10 generations. 300-400 is more common and many fall below that in a rare breed.

© John B. Armstrong, July 2000