by Alex Wagner
If your DNA shows certain markers appearing with certain frequency, then it suggests your people may have belonged to certain ancestral populations. But all that conditional gobbledygook signals just how contested those markers and frequencies and sample groups actually are. As with anything having to do with ethnicity and biology, autosomal markers are subject to vicious debate.
This is how it works: Each testing company samples DNA from what they consider “Old World” populations. These are the “reference populations” I had been so confused about, and it’s not that actual Winnebago are at the lab, sitting around comparing their spit to yours. Instead, scientists and community groups and for-profit companies travel the world to find groups of people who have been mostly untouched by genetic mixing, whether through economic, social, or geographic isolation. Members of these populations usually have to prove that both of their grandparents have uniform ancestry markers. With that established, these people are considered to have DNA that is therefore relatively pure.*2 When the scientists, or doctors, or whoever is doing the cheek swabbing and spit testing and blood sampling gather the DNA from what they consider to be a sufficient number of individuals from an unsullied (easy now, I’m using that term ironically) group, it forms what’s known as a “reference population” for a specific ancestral group. It is, in this respect, a virtual population.
Geneticists scan the reference populations to determine if there are unique DNA markers in each group—and these are known as ancestry-informative markers, or AIMs. These reference populations—basically large(ish) DNA databases—are often then purchased by the big testing companies, including 23andMe, AncestryDNA, and Family Tree DNA. This is what that disclaimer on the 23andMe website was referring to with its caveat:
This report CANNOT tell you ancestry estimates for populations for which we do not have sufficient data.
Which raises the question: Okay, which populations have “sufficient data”? As it turns out, mostly geneticists have amassed general reference populations for the regions of the world circa 1492, when the Old World met the New—which is to say, when there were distinct groups of humans who had not yet theoretically intermarried or interbred. There are reference populations for European, African, Asian, and Native American blood.
When people like myself want to find out who their ancestors were, our autosomal DNA is scanned for those AIMs. Depending on what AIMs come up, scientists make a call about your ancestry.
Which is to say: If 22 percent of my autosomal DNA shows the AIMs that correspond to the reference population for Northern European, then, by that math—I’m 22 percent Northern European.
As you can imagine, there are a lot of complications with this process.
First off: There’s the problem with the yardsticks themselves—the DNA in the reference populations. Who’s to say what “pure” means? The very concept of someone—anyone—having homogenous blood seems both arbitrary and impossible. At what point in history was blood considered “pure”—before East Asian blood mixed with Portuguese blood? Or before East Asian blood mixed with other, different East Asian blood? The year 1492 is one line you could draw between the Old World and the New, but there are plenty of other moments in history that came before and after that when a seemingly homogenous population mixed with another, and the DNA of a people was forever changed. The British colonized Burma in the mid-1800s, but the Pyu of Yunnan entered lower Burma in 200 B.C., and the Mon of Indochina arrived around A.D. 1000. At what point was Burmese blood considered “unmixed” and exempt of outside influence? The determination was subjective and largely dependent on what sort of time frame you were looking at: hundreds of years ago…or thousands of years ago. Or millennia.
And then there was the issue of the map: Some DNA was classified using political borders (for example, Irish DNA) while other DNA was defined using regional assignments (for example, South Asian DNA). As it concerned the latter, what countries made up South Asian versus Southeast Asian DNA? As it concerned the former, classification was wholly dependent on the borders you drew—and borders were pretty subjective things in the nineteenth and twentieth centuries (thanks, imperial powers!). A white paper written several years ago by several leading evolutionary geneticists noted:
Many estimations of genetic ancestry are designed to distinguish contributions from reference populations that live in particular geographic regions (e.g., West Africa, Europe, East Asia, and the Americas) that were prominent in colonial-era population movements. This creates a bias that might lead us to define ancestry in reference to particular sociopolitical groups. Moreover, our knowledge of diversity, and hence the genetic contributions to ancestry, of populations in many other parts of the world (e.g., East Africa, South Asia, Arabian Peninsula, and Southeast Asia) is limited.2
Colonialism! It was inescapable in the definition of a people. These scientists rightfully cautioned that we were still classifying our ancestors using the arbitrary lines drawn in the sand by imperial powers. But the British and French knew little to nothing about the genetic contributions of the Mon and the Pyu—hell, they had named the country “Burma” after the Bamah and, in so doing, basically delegitimized the other 135 ethnic groups inside the country’s borders. Employing their estimations for a contemporary understanding of descendants from the region seemed both questionable and unreliable, even if it was inevitable. I was offended for my Burmese people who would have their lives measured against a British yardstick (once again) or whose biological contributions might not even register, due to colonial whimsy or oversight or prejudice (or all three).
As it concerned the questionable metric of 100 percent purity and subsequent fractionalization, even if you could pinpoint some mutually agreed upon moment in time, before which all DNA everywhere was homogenous and afterward was heterogeneous—there was the very practical question of how to find people who still had that DNA. To find men and women who had lived in geographic isolation for such an extended period of time seemed rather difficult, especially these days. In the age of FedEx and TripAdvisor, with all of our blessed commercial and ethnic commingling, who could truly claim to carry genetic markers that precisely matched the genetic markers carried by the Yoruba in the year 1442? Probably no one.
I called up Dr. Michael Bamshad, professor of pediatrics and adjunct professor of genome sciences at the University of Washington and a leading authority on evolutionary genetics. I wanted to better understand the idea of reference populations—because the whole thing seemed pretty seriously shaky.
“For accuracy in a reference population,” Bamshad explained, “you need a gold standard. And it’s not clear that there is a gold standard.”
This was problematic. How many people were in these reference populations? Were the reference populations for the Bamah as sizable as the ones for, say, the Irish? Because the DNA testing companies often relied on reference populations from outside groups, you had to imagine that the market demand for consumer tests on the part of Irish Americans was greater than the demand from, say, the Burmese. Especially because Burma had languished in poverty and an Internet blackout for most of the early twenty-first century, and the Burmese were (generally) not easily able to drop three hundred bucks for a genetics test that you had to order online. If there wasn’t a lot of demand coming from Rangoon, then my guess is that the testing companies wouldn’t be going out of their way to find and purchase Bamah reference populations.
On its website, 23andMe at one point detailed the information about its DNA database as follows: “We compiled a set of 10,418 people with known ancestry, from within 23andMe and from public sources.” (In case you were wondering, the company notes, “This a big jump over the 210 individuals that powered our original Ancestry Painting feature.”) Most of the reference data sets, according to the company, come from its consumer database: DNA submitted by people who have purchased the kits. The rest come from public data sets, including
the Human Genome Diversity Project, HapMap, and the 1000 Genomes project. So 23andMe wasn’t relying on purchased DNA—it was relying on public databases…and on DNA from its users.
But if a bulk of its reference data was coming from consumers, wouldn’t that inevitably mean an excess of DNA from affluent, educated Americans who could afford to drop a hundred bucks on a lark—which is to say, mostly white folks?
I got on the phone with 23andMe’s senior director of research, Joanna Mountain, and asked her whether the diversity of the data sets was a concern. “That’s something we think about a lot,” Mountain said, and explained that the company was working with geneticists elsewhere in the world—in places including Sierra Leone, Congo, and Angola—as part of their “focused efforts to expand the diversity.” But these “special collections” of data weren’t being used, at present—they were still in the collection phase, according to Mountain. And she didn’t have a time line for when they might be available. “We’re trying to use them in the future,” she said.
With this in mind, I wondered how big (or small) the reference populations happened to be for each region in my ancestry breakdown. For my broadly 44 percent “East Asian and Native American” heritage—which comprised Native American (Colombian, Karitiana, Maya, Pima, Surui), East Asian, Japanese, Korean (South Korean), Yakut, Mongolian (Daur, Hezhen, Mongolian, Oroqen, Tu, Xibo), Chinese (Chinese, Han, Hong Kongese, Taiwanese), and Southeast Asian (Burmese, Cambodian, Indonesian, Lao, Malaysian, Filipino, Thai, Vietnamese) blood—23andMe was using 808 samples from its consumers and 560 from public databases, for a total of 1,368 samples overall.
For my less than .1 percent Sub-Saharan African heritage—West African (Bantu, Cameroonian, Ghanaian, Ivorian, Liberian, Luhya, Mandinka, Nigerian, Sierra Leonean, Yoruba), East African (Eritrean, Ethiopian, Maasai, Somali), Central and South African (Biaka Pygmies, Mbuti Pygmies, San)—the company was using 228 samples from consumer DNA and 393 from public databases, for a total of 621 samples overall.
I asked her how much data was “enough data” to obtain an accurate result, and Mountain offered, “It’s not that we say, ‘We need a minimum of this.’ We look at the data we have and say, ‘How granular can we get?’ ” In a subsequent email, her colleague Andy Kill clarified: “We look at our data and say, ‘What can we tell you?,’ so we’re confident in all of the reference populations we have and the accuracy of the results based on those populations.”
In cases where the reference populations are smaller, for example, the “South Asian” population, Kill wrote, “That forces us to be more general in the result.” He later added, “In cases where we can’t pinpoint a certain country, we will up-level that assignment to a region.” This is why the company labeled those segments of my DNA “Broadly South Asian,” rather than citing specific countries.
On that note, I asked Mountain how the company defined “countries”—given the fact that a place like Burma, and its borders, had changed mightily over the years. She offered a cryptic “It’s not today, but it is not that far back in time.” She added, later, that this meant “before mass intercontinental travel.”
In the months after I spoke to Mountain, 23andMe began offering its users an Ancestry Timeline, which tells you the time period in which you theoretically inherited that French DNA (for me, sometime between 1860 and 1920) or that Chinese DNA (between 1770 and 1860). But that doesn’t take into consideration the fact that China in the late eighteenth century looked different on the map from what it does today—it just means that one of my relatives in that time period gave me what we now consider to be “Chinese DNA.”
AncestryDNA, meanwhile, informs users that its “reference panel (version 2.0)”*3 contains 3,000 DNA samples from people in 26 global regions, depending on a fraction of the data that 23andMe is using. Given that there are fifty million people in Burma alone—the fact that AncestryDNA has only three thousand samples spanning less than a seventh of the globe seemed…not all that comprehensive.
For the region of “South Asia” (India, Pakistan, Nepal, Bhutan, Bangladesh, Sri Lanka, and Burma), AncestryDNA has 161 reference samples on file.
For “East Asia” (including Russia, China, North Korea, South Korea, Mongolia, Burma, Japan, Taiwan, Philippines, Indonesia, Thailand, Laos, Cambodia, Vietnam, Singapore, Brunei, and Palau), the company has 645 samples.
For Eastern Europe, it has 646. (Just to ballpark the ratio here: If you include just Ukraine, Poland, Romania, the Czech Republic, Hungary, Belarus, Bulgaria, Slovakia, and Moldova, then the population of the region would be somewhere around 150 million people.)
I got on the phone with Cathy Ball, Ancestry’s chief scientific officer, and again asked what she thought was a “good” amount of data for an accurate reference population. Ball gave me kind of a nonanswer:
“You will never get any scientist to say, I don’t need more data,” she said.
But when I asked her whether certain populations would have more comprehensive reference material because of market demand (the Irish versus the Burmese, in my case), Ball was remarkably forthright.
“Absolutely,” she responded. “We ask ourselves, ‘What would be informative to most Americans? What, in theory, would be useful?’ Western Africa, Europe—those people are places where we concentrate on. England, Ireland, Scotland. Part of it is opportunity. Some places we don’t have—and it’s basically practicalities. Maybe [samples] weren’t collected there and we don’t have a ton of customers building family trees [from those places].
“Then there are places like China,” she continued, “where you have multiple ethnicities, and it’s gonna be pretty hard to know who’s Han Chinese and who’s from another ethnic group. We don’t claim to be specialists, but we try our best.
“And then there are places like Egypt,” said Ball, “where people have been going back and forth for years, and what does it mean to be Egyptian? It’s messy but it’s also interesting.”
Indeed, it was interesting, and it was certainly messy. Ball addressed all the concerns I’d had about the accuracy of these tests and confirmed that there were reasons to be concerned. There wasn’t going to be great reference data for more “exotic” ancestry (like mine), and the existing DNA material was often collected and classified according to questionable parameters (social, political, and ethnic divisions were not exactly standards that everyone agreed upon).
What I was learning in these conversations was that if you were a white person from Europe, you might very well get a more accurate result than, say, if you were a brown person from anywhere else in the world, especially countries with not a lot of purchasing power. (What else was new?)
As a case in point: Ancestry had 272 samples on file for Scandinavia (a region comprised of less than 27 million people), but only 161 samples on file for “South Asia” (a region comprised of 1.75 billion people).
Given these statistics, it seemed hard to believe that any of these companies would have enough material—or the right method by which to analyze it—to accurately reflect the DNA that might have been carried by my Burmese ancestors in the mid-1800s, and by “hard,” I meant “damn near impossible.” This information certainly went a long way in explaining why my mother and grandmother were being assessed with such strong Chinese and Mongolian roots: Those were big parts of the world, and presumably there was more reference data for them. I wouldn’t have been surprised if we were some of the very first Burmese people to even take this test!
I appreciated the fact that sociocultural borders and genetic purity were never going to be agreed upon, and that each company’s DNA database was a work in progress. But the fact that there wasn’t at least some sort of limitation imposed upon these ancestry results—something beyond the vaguely worded caveats and “confidence indexes,” a more distinct signal that the regions and time frames were somewhat fungible and arbitrary—seemed disingenuous at best.
>
I reached out to the consumer hotline at Family Tree DNA to sniff out a little more on these ancestry breakdowns and, after several days of calling and being put on hold, was put in touch with the company’s founder and CEO—a charmingly cantankerous entrepreneur named Bennett Greenspan.
“The problem is that the consumer wants a yes or no,” said Greenspan. “They want a blanket answer: 21 percent Irish and 14 percent English. That’s what people want. Science is not that simple.”
I understood this to be a partial cop-out: Greenspan acknowledged that the percentages were kind of BS but insisted that customers wouldn’t have bought the tests if the results were presented as anything less than definitive. It was the same line of argument that the makers of nacho chips offered whenever anyone asked why their food was stuffed with artificial cheese flavor: Sure, we could take out the extra orange cheese product, but nobody would buy our nacho chips if we did! The real stuff (actual cheese, concession of data limitations) wouldn’t sell.
I wanted to know if Greenspan had a theory about why I’d been told that I was alternately nearly 15 percent Scandinavian, with little French or German DNA to speak of…and also not specifically Scandinavian at all, but instead one-fifth French-German?
At first, Greenspan suggested that one of the tests (not his!) might be overcorrecting for Scandinavian DNA because of its underlying database. He referred to it as “the Sorenson repository”—a set of DNA acquired when the company bought the Sorenson Molecular Genealogy Foundation DNA collection. According to one expert I spoke with, a company that used a collection like this could overemphasize the Scandinavian in its customers.