Breeders often talk about inbreeding and outcrossing as though they were the only possibilities -- and generally with negative comments about the latter. There are other possibilities, and I have long been a proponent of assortative mating. It is not a theoretical concept that doesn't work in practice; I know several breeders who do it and achieve good results. This essay will attempt to explain why it is a good idea, but first I need to define the alternatives.

Random Mating
Though random mating is not a common breeding practice, understanding what this implies is important. Random mating is exactly what the name implies: mates are chosen with no regard for similarity or relatedness. (If the population is inbred to some extent, randomly-selected mates may be related.)

Random mating is one of the assumptions behind the Hardy-Weinberg formula, which allows one to calculate the frequency of heterozygous carriers from the frequency of individuals expressing some recessive trait in a population. Because inbreeding among purebred dogs and in other small populations decreases the frequency of heterozygotes, these estimates may be higher than the actual incidence.

Inbreeding and Linebreeding
Inbreeding is the practice of breeding two animals that are related (i.e., have one or more common ancestors). The degree of inbreeding may be assigned a value between 0 and 1, called the inbreeding coefficient, where 0 indicates that the animals have no common ancestors. Because the number of ancestors potentially doubles with every generation you go back in a pedigree, you eventually get to a point, even in a very large population, where there are simply not enough ancestors. Thus, all populations are inbred to some degree, and a true outcross (the term generally used when two animals are "unrelated") is not really possible. The term is generally misused to describe a cross between two animals with different phenotypes. In a population with a limited number of founders, a maximum number of ancestors -- the effective population size -- is reached in some past generation. This number will be governed by various factors, such as the total population size, how far individuals travel during their lifetime, and whether there are inbreeding taboos or other mechanisms that reduce the likelihood of close relatives mating.

Inbreeding does not change allele frequencies directly, but it does increase the proportion of homozygotes. Individuals homozygous for deleterious genes are likely to be removed from the breeding pool by natural selection (if they do not survive to reproductive age) or by man.

Linebreeding is merely a term used for a particular type of inbreeding that often focusses on one ancestor who was considered exceptional. Particularly if it is a male, this exceptional ancestor may end up as grandfather and great-grandfather -- sometimes more than once -- in the same pedigree. Father-daughter, mother-son, and some other combinations also result in a disproportionate number of genes coming from a single ancestor. This type of close inbreeding is less common. [In contrast, the mating of full sibs or first cousins doubles up on two ancestors equally.]

As the result of several common practices, most pure-bred domestic animals are more inbred than they really need to be. One is that some breeders own a small number of animals and breed only within their own group. A second is that many breeders have the idea that outstanding animals can be produced by inbreeding -- by doubling up on the good alleles while somehow avoiding the bad. Even if you were to point out that this is a gamble, such breeders might respond that they are simply helping natural selection.

Beyond the conventional close-relative inbreeding, there is another practice that has much the same effect, namely the popular sire phenomenon (generally over-use of a well-promoted champion). In fact, many who breed to such a dog believe they are doing a "good thing," as they will be increasing the frequency of occurrence of the genes that made him a champion. What they may not realize is that they are increasing the frequency of all genes carried by this animal -- whether they are good, bad, or innocuous -- and that champions, like any other animal, carry a number of undesirable recessive alleles (the genetic load) that are masked by wild-type alleles. The result of the popular sire phenomenon is that almost all members of the breed will carry a little bit of Jake Hugelberg, and any undesirable trait carried by Jake will no longer be rare. Finding a safe, unrelated mate then becomes an exercise in futility.

If we lived in a world where all the genes followed the simple rule that there may only be good alleles, which are dominant, and bad alleles, which are recessive, then inbreeding could be an effective tool for improving a breed. However, during the past 25 years, geneticists have been directly measuring genetic diversity in populations by looking at the DNA or proteins, rather than at the phenotype. They have found that many individuals who cannot easily be distinguished by their phenotypic appearance nevertheless have considerable differences in their genotype. Some of these alternative alleles (termed neutral isoalleles) are functionally equivalent. Others have lost only a small portion of their normal function.

Suppose we have a "mutant" allele that has lost only 5-10% of its normal function. In many cases, this would not produce a noticeable effect. If you made an individual homozygous for this allele, you would not even be aware that you had done so. Now consider that the same fate may befall a number of genes during an inbreeding program. Eventually, you will have an individual that is considerably less fit than one carrying the normal alleles for all (or even most of) these genes. There is no magic formula for regaining what you have lost. You must start again.

[Sometimes mutant alleles result in an even more dramatic loss of function, but remain undiscovered under normal conditions. A good example is vWD in Dobermans.]

About the only animals that are routinely inbred to a high level are laboratory mice and rats. There, the breeders start breeding many lines simultaneously in the expectation that the majority will die out or will suffer significant inbreeding depression, which generally means that they are smaller, produce fewer offspring, are more susceptible to disease, and have a shorter average lifespan. Dogs are no different. If you can start with enough lines, a few may make it through the genetic bottleneck with acceptable fitness. However, dog breeders generally don't have the resources to start several dozen or more lines simultaneously.

Sometimes two different alleles may be better than one. Consider the major histocompatibility complex (MHC). These genes are responsible for distinguishing "self" from "foreign", and a heterozygous individual can recognize more possibilities than a homozygous one. Having a variety of MHC alleles is even more important to population survival. Not only does this provide better defense against pathogens, but there is growing evidence that parents who carry different MHC haplotypes may have fewer fertility problems. This is not a universally accepted theory, but today one is hard pressed to find a conservation or zoo biologist concerned with preserving an endangered species who would not list maintaining maximum genetic diversity as one of his/her primary goals.

Assortative Mating
Assortative mating is the mating of individuals that are phenotypically similar. It is a normal practice, to some degree, for humans and various other species. Though phenotype is a product of both genotype and environment, such individuals are more likely to carry the same alleles for genes determining morphology. If we are talking about a conformation that is basically sound from the structural point of view, the genes involved will have been subjected to natural selection for thousands of years and will most likely be dominant. The major characteristics that set one breed apart from another will likely have been fixed early in the breed's history. ("Fixed" means that there is only one allele of present in the population. If there is only one allele, the question of dominance does not arise.) Consequently, when you look at a dog, you are looking at his genes. If the conformation (or, for that matter, the temperament, intelligence, or whatever) is not good, then you are very likely looking at a dog or a breed that is homozygous for one or more recessive alleles that you would probably like to get rid of. If it is the dog and not the breed, you may elect not to breed him, or you may look for a mate that covers the problem. If it is the breed, the only solution would be to introduce some genes from another breed. (That would be an outcross!)

Breeding together animals that share dominant good alleles for most of their genes will produce mainly puppies that also carry these genes. Even if the parents are not homozygous for all these good alleles, you should still get many that are suitable. More important, if animals heterozygous for certain genes are more fit, assortative mating will preserve more heterozygosity than inbreeding. However, unlike inbreeding, assortative mating should not result in an increased risk of the parents sharing hidden recessive mutations. Though we might like to eliminate deleterious recessives, everyone carries a few. Trying to find the "perfect dog" without either visible or hidden flaws is like betting on the lottery. There may conceivably be a big winner out there, but they are certainly not common.

The more you try to cover the deficiencies in one dog with good qualities in another, the less the dogs will have in common. If, then, the results are unsatisfactory, they should not be blamed on assortative mating, as that is no longer what you are doing.

The risks involved
Some trait that breeders consider desirable could be the result of homozygosity for a recessive allele for gene A or gene B. Obviously, crossing an AAbb with an aaBB will produce AaBb progeny that will not express this trait.

(However, aside from some of the genes affecting coat color, I can think of no examples.)

If care is not taken to go back far enough in the pedigrees, you may have two animals with similar phenotypes resulting from common ancestry. Whether you are inbreeding unintentionally or intentionally, the consequences are the same. The solution is simple: check the heritage.

Because assortative mating involves selection (you are hopefully mating the best together, and not the worst), you are denying some dogs the opportunity to pass their genes on to the next generation. This is, perhaps, the subtlest of risks, as it does not seem to involve doing anything "wrong." Most would argue that it is merely doing what nature does -- eliminating the least fit. But what if some of these "less-than-best" happen to be the only ones to carry the best allele for some gene? Out goes the good with the bad!

This is primarily a "low-numbers" risk. The larger the population, the less likely we are to find that important alleles are carried by only a few individuals. However, it pays to know where the diversity lies. Do any of you know which, among the current dogs, are most likely to carry the genes of any given founder?

Notes
Inbreeding calculations do not account for the possibility that an allele will become homozygous by "chance," though this, too, can be calculated if the frequency at which an allele occurs in the whole population is known. Most basic Genetics texts explain how. (See, for example, Willis, pp. 293-295, "The Hardy-Weinberg Law.")

I have seen figures of 2500 genetic diseases in man and there are likely to be as many in Canis familiaris, taken as a whole. In man. the vast majority are rare (allele frequencies of < 0.01, which means < 1 in 1000 affected). However, everyone carries three to five "lethal equivalents." This is their "genetic load." Canine breeds are often established with a handful of founders, so we end up with a subset of one or two dozen problems, at frequencies at least 10-fold higher. [If we had five founders, each with a unique set of problems carried as single recessive alleles, the allele frequency of each will initially be ~ 0.1 and ~ 1% will be affected. © John B. Armstrong, 1997, 1998, 2001

Revised July 14, 2001

source: http://www.canine-genetics.com/breed.htm