A Breed Apart I

well_read_border_collieThose of us in Border Collies have long known that they’re in a class by themselves in the arenas of intelligence, agility, drive, and trainability. They’re different from other dogs and that difference is often wide. And we know that the difference is in their genes.

When you map the genome you can start to measure the evolutionary distance between these genes just like you can measure the distance between cities on a map, and you can also view the paths between these genomic cities just like the interstate highway system.

By looking at specific sets of genes and individual traits you can compare how likely it is that an individual gene controls the expression of a trait by comparing the frequency the gene would appear in a random individual to how often it appears in individuals with the desired trait. Because correlation is necessary to prove causation, looking at highly correlated gene-trait pairs expedites the process of proving gene-trait causation.

One group of scientists is using this statistical inference method (often called Bayesian Analysis) to study and lure out the genes that are responsible for complex traits in dogs such as noise phobia. The central breed in this study is the humble Border Collie.

A tertiary goal in their study is to investigate “whether or not there is a “split” between dogs bred primarily for working, and dogs bred primarily for show.” And apparently there is.

To ferret out the genes responsible for specific traits found in the Border Collie and not in other breeds, the researchers need to construct a neighborhood map of dogs which places highly correlated dogs closer together and uncorrelated dogs further apart.


If you’ve ever seen a program on evolution, you’ll recognize these maps–called Phylogenetic trees–as they are often used to show the “tree of life” or the descent of man from primitive life down to our present state, with various branches separating us first from other human species, and earlier from the other forms of life present and past.

These graphs are rooted in an ancient common ancestor, branches often denote a separation of species or some predetermined degree of distinction, and they are directed. An undirected graph would be “is related to” such that A-B means that A is related to B and B is related to A. In a directed graph, A-B does not imply B-A, for example “is an ancestor of” would be a directed graph because if A is an ancestor of B, B is not and can not be an ancestor of A.

Since we know the causative method of correlation in dogs (familiar relation, i.e. breeding and inbreeding) the branches of this graph can make a good approximation for estimating common ancestry in rooted graphs. Unrooted graphs can provide a means of distinguishing relatedness without assumptions of common ancestry.

One such graph was created for the study of Border Collies and it includes a sampling of other herding breeds:



Unrooted phylogenetic network constructed by Bayesian analysis, based on 4200 SNPs spread evenly across the canine genome. Blue = German Shepherd Dogs, purple = Portuguese Water Dogs, green = Australian Shepherds, pink = show Border Collies, red = working Border Collies.

To conduct proper studies of association between genes and behavior, we must first check for “stratification” (population substructure) within our breed samples. This is a question of immediate concern in breeds that are “split,” or contain subpopulations that are bred for very different purposes. If we do not account for such structure before conducting association analyses, it is possible to obtain spurious associations between genotypes and behavior that reflect breed splits (such as show vs. working) rather than actual functional significance.

We included a small number of kennel club registered show Border Collies (primarily of Australasian breeding) in our Border Collie sample for genotyping, the remainder of which was made up of ISDS and ABCA (working registries) registered dogs. Our phylogenetic, clustering, and principal components analyses all suggest a genetic split within the breed between working and show Border Collies that is probably as large as the genetic distances between some breeds. We hope to collect samples from more geographical regions, and from different populations of Border Collies (working, show, and sport), to further explore these findings.


OMFG! Barbie Collies really are another breed! The Great Schism of Border Collies has happened. Sheeple everywhere are giddy with delight that the difference between their “coyotes” and the “barbie collies” isn’t just apparent visually, but you can even see it on the molecular level, too.

From reading the report, you’d suspect that there’s some problem or that this analysis proves something vital. There isn’t and it doesn’t. In the next part, I’ll get real and discuss what this chart is really telling us.

* * *
Comments and disagreements are welcome, but be sure to read the Comment Policy. If this post made you think and you'd like to read more like it, consider a donation to my 4 Border Collies' Treat and Toy Fund. They'll be glad you did. You can subscribe to the feed or enter your e-mail in the field on the left to receive notice of new content. You can also like BorderWars on Facebook for more frequent musings and curiosities.
* * *

Related Posts Plugin for WordPress, Blogger...

About Christopher

Christopher Landauer is a fifth generation Colorado native and second generation Border Collie enthusiast. Border Collies have been the Landauer family dogs since the 1960s and Christopher got his first one as a toddler. He began his own modest breeding program with the purchase of Dublin and Celeste in 2006 and currently shares his home with their children Mercury and Gemma as well. His interest in genetics began in AP Chemistry and AP Biology and was honed at Stanford University.