The ethnicity calculations used by ancestry.com and many other commercial DNA testers are toxic hokum. Wonderful marketing tools precisely because they appeal to the lizard back-brain in all of us, they gloss over the fact that there is no such thing as “ethnicity”. Peoples and differences and communities there are aplenty, but ahistorical essences that define groups as this ethnicity or that? Puh-lease.
Ancestry’s assertion that “the ethnicity estimate provides a distant picture of a customer’s genetic origins, perhaps hundreds or thousands of years ago”, is just plain wrong. What you get is a comparison of your test results with a (pretty paltry) reference panel, reasonably accurate to four generations, less accurate to five, sometimes useful to six and almost always worthless before then. A distant picture it is not. (For more on the flaws of this stuff, see UCL’s “Debunking Genetic Astrology“.)
So when ancestry.com announced “Genetic Communities” as a new feature of its DNA service, I was sceptical, to say the least. Then I saw the map produced by their analysis of my own test, and I was blown away. None of my family tree is on ancestry, so it was produced purely by DNA analysis. And they hit the bulls-eye on the detailed North Connacht and Galway origins of all 16 of my 3 X great-grandparents.
How could they do this, working purely from the DNA? According to the white paper accompanying the new service, a “genetic community” is simply a group of people from more or less the same place who married each other over multiple generations, a nice, loose target, and much more sensible than “ethnicity”. They arrived at their more-than-300 communities by detailed meta-analysis of the DNA matches in more than 2 million samples. Instead of just comparing my test with all the others and seeing to whom I was most related, they took all those to whom I was related and examined who they were related to. And so on and so on.
It was then possible with the aid of an algorithm for detecting densely connected sub-networks within large datasets (the “Louvain Method“, if you must know) to identify the groups most closely related to each other. They then went on to use their own online trees to associate these groups with particular locations, and then ran the whole process again and again to zero in on sub-sub-groups. The granularity of the results is truly extraordinary. In Ireland alone, there are (so far) seventeen different subgroups, ranging from East Donegal to West Cork to North Connacht to Connemara. Each group is presented alongside a series of good short histories explaining the history of the area over the past two centuries and its outmigrations to the US.
Ancestry has used the critical mass of its huge collection of DNA test results to provide a genuine, scientifically-grounded genetic atlas of the past 200 years, no less.
I still have my quibbles (to misquote Charlton Heston, “They’ll prise the quibbles from my cold, dead hands”.) There are unexplained sciency-looking variations in the size of the location circles on the map: what do they represent? Are they unique to each test analysis, or generic? How were the precise-looking boundaries of the communities arrived at? Above all, why is it not possible to see the data underlying the maps by clicking through?
But quibbles they remain. The whole thing is nothing short of deadly.