Finding out who your "ancestors" were via DNA

Newsweek has an entertaining story which highlights the recent penetration of science into the venerable enterprise of genealogy. The good:

...Adopted at birth, Royer knew nothing about her biological parents. But certain physical traits-wide nose, dark skin-led people to guess that she was Iranian or even Cambodian. "I always wondered," she says. Two hundred dollars and a swab of her cheek gave her an answer: Royer's maternal ancestors were most likely Native American. The knowledge, she says, "makes you feel more of a person."

The dumb:

DNA testing is forcing some people to rethink their identities. Phil Goff, 42, of Naperville, Ill., thought his heritage was pure English, but a Y chromosome test matched him at least partially to Scandinavia. Now he wonders if he has any Viking blood in him.

Note the contrast here. In the first case, we have an extreme and unfortunate situation where an adoptee doesn't know the identity of her biological parents and whose appearence bespeaks an "exotic" lineage. The conditional probabilities, that is, take the lack of knowledge about antecedants and physical clues into account, work out so that an mtDNA test can be very illuminating. In the second case we have a situation where you are attempting to make inferences far back into the past and resolve distinctions that are relatively minor in comparison to the first scenario. Not only are Scandinavians and the English relatively affinal populations in comparison to Native Americans and Europeans, but there is plenty of evidence of gene flow between Scandinavia, northern German and England. Additionally, the historical literature (as well as linguistic clues) tells us that the "Danelaw" of the north and east of England was actually settled by a fair number of Danes ~1000. Drawing too many inferences from a uniparental lineage can be very misleading, all you are gleaning is the coalescent for that one gene. Oxford geneticist Bryan Sykes actually found that half the people with his surname share the same Y chromosomal lineage. That doesn't mean they are particularly closely related at all, they simply have a patriline in common. Similarly mtDNA lineages are tracing a sequence of mothers and daughters.These tests don't speaking to the overwhelming majority of your ancestral paths which are "broken" by mother-son or father-daughter steps.1 Another problem is that many of the haplotypes might be misclassified. Or consider an individual who hails from southern India but was told by Spencer Wells (pretaped video) about "ancestors who left Central Asia for Europe" after his DNA was sequenced by the Genographic Project. The article also highlights autosomal methods bandied by DNAprint Genomics which give you a measure of your various ancestries. And they also have problems.2 When friends ask me if they should shell out a few hundred dollars I usually say no, because I don't think that these tests tell you anything you don't know (there are exceptional cases, such as the one above).

Nevertheless, scientists like Spencer Wells, Bryan Sykes and companies like DNAPrint Genomics will keep pushing and hyping these tests as answers to deep rooted existential questions relating to "where we came from." There are several issues here. First, money. Sykes has a company, Oxford Ancestors, he has an incentive to oversell the relevance of these methodologies. Myself, I have seen referrals coming in from financial message boards in relation to DNAPrint Genomics on my other weblog, there is money to be invested, so you don't want to screw up your Power Point presentation with qualifications. Wells has managed to convince thousands of people to pay to give him a swarm of data. Who needs grant applications?

These scientists and companies are preying on human needs and intuitions which are being inappropriately triggered by these findings (part of it is semantic, part of it is reasonable extrapolation based on what I would call "folk genetics"). The reality is that our ancestry is extremely reticulated, it does not extend indefinitely outward back into genetic space in an every expanding mushroom cloud of twigs, it begins to wrap back onto itself and twists into a ball of spaghetti which eventually becomes attenuated via founder events (the "Out of Africa" one being the most prominent). A Y chromosomal or mtDNA lineage will tell you only so much, and more so on the populational level (the soup of alleles can perhaps give you clues to the migrations between populations, at least that's the theory). Even the more nuanced autosomal tests, which sample many loci on the genome instead of just a lineage through one nonrecombining marker, are problematic in their framing. South Asians routinely get back results that suggest they are admixed, but that's only because the tests are geared toward Europeans, East Asians and Africans. The preponderance of the evidence is that South Asians are not admixtures at all, but their own populational cluster, with various relationships to other geographic races. The further and further we move back in generations the less and less our conventional intuitions matter, who we are as individuals becomes diffused into thousands of humans who carry bits and pieces of information which will eventually make it into us. But, selection also matters, alleles, genetic variants, may sweep across populations without an concomitant replacement of other regions of the genome. In the deep past pedigrees are irrelevant, bodies are nothing but way stations where collections of genes get together for a generation.

In short, it's complicated. But the scientists who oversell this work and the companies selling you ancestry-in-a-test-tube would not benefit much from admitting this.

Via Dienekes.

1 - Because there is a variance in the number of sons and daughters a couple will have around the expectation of 0.5 the vast majority of male and female lineages will be broken at some point and go extinct as a mother has only sons or a father only has daughters. This is why we can have one mitochondrial Eve, one mtDNA mother, but be confident there were many other breeding females around 100,000 years BP. The persistence of one lineage is simply how the odds work out. A similar problem crops up in monogamous societies which are based on patrilineages, for example, Augustan Rome. Male lines simply kept dying off because the likelihood of having many successive generations of sons who live to adulthood and have their own sons was very low.

2 - I do think this method is kosher for forensic work if don't know anything about a suspect or victim. But most people have much more knowledge about their background than this and I am extremely skeptical that any real value is added.

Tags

More like this

Part of the problem is that most non-scientists don't understand that the majority of genetic variation has no effect on phenotype. If you find out that you share the same mitochondial haplotype as the founder of some particular female lineage, most people think that must really MEAN something, in terms of their own innate physiological or psychological qualities. Also, people don't realize how large the human genome is, and many probably assume that a strech of DNA a few kilobases long is a significant portion.

jacob, perhaps. it seems that one problem is, to be glib, people conflat phylogeny with morphology. they don't conceive of the power of selection to reshape phenotype over a few thousand years.

I have another, less scientific hypothesis:
what if Goff's mom had a brief fling with an infamous Scandanavian named Hjalmar Andersen at Oslo in 1952? Goff is 46, so you just never know...*wink*

I paid for the Family Tree DNA mtDNA HV1 and HV2 tests back when they were more expensive.

a) I don't think Family Tree DNA intentionally promises anything other than what it delivers. The problem is that a lot of people are scientifically illiterate, not that Bennett Greenspan gets up and says all women in the K haplogroup look like Queen Katrina of 50,000 B.C., or anything terrible like that.

b) There is something really, really touching about getting my M* (or whatever) results back and realizing that I got my mtDNA pattern -- basically, a genetic bracelet -- from some woman who must have lived in Asia 50,000 years ago. And, in turn, that I got the mtDNA itself from some little microbe that was crawling around hundreds of millions of years ago. Of course, I got all of my other DNA from other people, but "M Mom" is, in effect, the one I have the picture of.

c) In some cases, the results may tell an interesting story. Example: my matrilineal ancestors were Jews from Belarus, but my HV1 type seems to be Central Asian. There are a handful of other people with Jewish ancestry who've been tested and have the same haplotype. As far as I can tell, the only other people who share our HV1 type and have their results on the Web are Chinese. One explanation I've come up with is that maybe I'm descended from a Chinese woman who somehow went west on the Silk Road with a Jewish trader around the time of Genghis Khan. Maybe the real story is more boring, but a girl can dream.

d) The more people get tested and put their results on the Web, the better testing will work.

Of course, the draw back is that someone mean will use all this to breed an army of mutant cannibal clones who will eat all of us, but that's life.