How complete are DNA tests at determining ethnic / racial background?

March 4, 2012 9:01 AM

How complete are DNA tests at determining ethnic / racial background?

I am researching the phenomena of "Black Indians"--African American people who also have Native American ancestry. One of the dilemmas that has faced many Black people in recent years is whether to take and how to interpret the results of genetic tests, used to demonstrate ethnic and racial heritage. How complete are these tests? Is there a significant margin of error, due to incomplete profiling?

Tina Turner believed that she had significant Navajo and Cherokee ancestry, but genetic testing revealed that she was only 1% Native. Is that 1% likely to be a truly accurate number?

I have heard that the tests often only look at part of one's ancestry. For example, only the paternal or maternal line. Could Turner thus actually have a lot of Native ancestry, perhaps 10% or more, despite being told this small number, because only some ancestors are revealed by the testing?

Any insights into rates of error, or advice for interpreting such tests, would be greatly appreciated.

posted by mortaddams to science & nature (10 answers total)

I can't speak to the race ethnicity question except to say that I view the socially constructed categories unhelpful at best.

But I can tell you that Tina Turner can only test for her mother's lineage. She would need a man in her family to do the test and test for the paternal line. (a brother or her father..obviously not her mother's brother). Men can test for either side-- but they're two separate tests.
posted by vitabellosi at 9:12 AM on March 4, 2012 [1 favorite]

I think vitabellosi is talking about mitochondrial DNA and paternal Y-chromosome haplogroups but I don't think thats what you're asking about.

You might want to check out the HapMap project which uses reference populations to try to determine global genetic variation. African, for example, is compared to a reference population of Yoruba in Nigeria.

Usually, if you sign up for sites such 23andme.com, they'll also try and determine how much your DNA matches pre-colonial distributions (gotten from the reference populations) to try and determine your ancestry. Mine, as a for example, comes to about 2/3 European and 1/3 Asian/Native-American. I'm Mexican and that mix is about right for Latin Americans.
posted by vacapinta at 9:35 AM on March 4, 2012

New autosomal tests can test all your chromosomes, from all your ancestors, regardless of their (or your) gender. Up until recently you could only test your direct maternal line (mother's mother's mother's...mother) through mitochondrial DNA, or males could test their direct paternal line (father's father's father's...father) through y-chromosome testing. Those tests are still very useful for genealogy purposes, but there are more options now. Examples of the new autosomal tests would be FamilyTreeDNA.com's "Family Finder" test, or 23andMe.com's v2 or v3 tests.

In terms of testing ethnicity, though, the new autosomal tests can pretty much only pick up on the last six to eight generations. This will get better in time, but not yet.

The other downside of using genetic genealogy to determine ethnicity is that many smaller and minority populations are still vastly under-sampled in the comparison DNA databases. This too will get better in time, but not yet.

Finally, most commercial DNA firms are not sampling the whole genome yet, just a lot of it.

I would compare the whole situation to something like an early digital camera with limited megapixels; you can tell that the photo it takes is of a tree, but it's still sort of blurry, and what kind of tree is it? Still hard to say, but promising for the future.
posted by Asparagirl at 9:41 AM on March 4, 2012 [7 favorites]

Part of the problem with any kind of genetic test is that people tend to think reductionistically about genetics. There is no gene for African or Native American or Chinese or whatever. There are isotypes that are more common among people from different places (and rare or absent in people from others) and you can figure things out statistically to some level, but after a while things fall into the noise as Asparagirl says.

You get 1.5% of your DNA from each great great great great great grandparent, but if the 1.5% you got from you ancestor who was, oh, let's say Eskimo–Aleut, doesn't contain any of the genes that are particularly concentrated in the Eskimo–Aleut population it's not going to be very noticeable using current technology.
posted by Kid Charlemagne at 9:55 AM on March 4, 2012 [14 favorites]

I recently had my DNA analyzed by 23andMe.

First, yes, they can tell you both your maternal (mitochondrial DNA) and paternal (y-chromosome DNA) lines. These haplogroups give you a pretty good idea where those two lines of ancestors lived thousands of years ago. Because Native Americans likely migrated to the American continents > 10,000 years ago (before the end of the last Ice Age), you can fairly accurately pick out the genetic lines that identify them.

But those are only two lines of ancestors, when of course you have many more.

The second way they can try to pin down your ancestry is by comparing your DNA to other people who have some knowledge of their ancestry. They keep a database in which people enter the known ancestry of their grandparents, and they can match up bits of your DNA with other peoples' DNA. They then tell you what those other people report as their ancestry. For example, for me, 70% of people whose DNA is like mine identified their relatives as coming from Ireland, 50% identified France, and so forth.

That's probably somewhat less accurate, but at least in my case, it coincides with what I know about my own ancestry.
posted by mikeand1 at 10:27 AM on March 4, 2012

Nthing that your understanding of "ethnicity" and ethnic background is far more nuanced, socially constructed, and complicated than the questions most DNA testing can really answer.

23andMe was able to figure out that I was likely related to a bunch (okay, twenty or thirty) of its users - but the sorting of that list improved significantly when I told the site places my recent ancestors had lived, and gave it surnames I know they had - up to that point, it was sorted by haplogroup only. I know the site is using information from outside the "figure out your ethnicity" test, and/or guessing, because I'm female and it's turning up people I'd be related to on my father's side. On the ethnicity page, I still look exactly like my mother's mother's mother - it hasn't yet told me I have any Jewish ancestry, or Eastern European ancestry, something it'd need paternal y-chromosone stuff to know about.

(I am hard at work on my dad and his brother to join, too, for this exact reason.)
posted by SMPA at 10:29 AM on March 4, 2012

Race is a social construct, however, there are many different "alleles" for different areas of the world. It's not "complete", as you can have alleles from many parts of the world that you wouldn't expect or know about. Not only that, but there is a huge difference in the definition of ethnicity and race. In the same way that a person might have a genetic marker from a particular area in the world, they may be from that particular are and thus be considered as a part of an ethnic group in that area, but they may not be from there at all. There is a strong biological component to ethnicity, but it's not necessarily always present. Here is an interesting article available for free that explains it more thoroughly than I could, and it has many citations for you to explore:

Perhaps we should simply recognize that getting past this is really about the dead hand of our cultural past, of vested social and material interests correlated with categorical concepts that did have some biohistorical basis (for example, black slaves were from Africa, white citizens from Europe). We waste our expertise endlessly attempting to adjudicate cultural mores with biological data, debating whether “race” is “real” when we know that the social reality of race, whatever race is, is mediated by the distribution of genetic variation but not determined by it.
Race is a conceptual construct with historical, biological, and anthropological elements. Its mix differs for each person, probably in ways he or she is not even fully aware of. None of these thoughts is new or secret, but for similar sociocultural reasons they seem hard for people, even anthropologists, to accept. The legendary baseball great, Satchel Paige (1906 –1982) (Fig. 3), veteran of the Negro Leagues, once quipped “Don’t look back. Something might be gaining on you!” In the study of race, it’s a shame that even scientists rarely seem to look back, to
realize that we are not putting any distance between ourselves and our past. Paige finally made it to the Major Leagues and never had to look back. Maybe it’s time for us to move for- ward in our race, too.

Weiss KM and Fullerton SM (2005) Racing around, getting nowhere. Evolutionary Anthropology 14: 165–169
posted by two lights above the sea at 11:03 AM on March 4, 2012 [2 favorites]

We did a 13andMe test for both my dad, and my maternal uncle. My dad's family was from Delhi, India (for generations upon generations upon generations). My mom's family was from a very tiny town in the Apennine mountains in central Italy (for generations upon generations upon generations).

The test for our dad was pretty much what we expected - his small paternal line pinpointing him to central Asia (India, Iran, and the Caucasus), and a much larger haplogroup for his maternal line (Indians, Chinese, Tibetans).

For my uncle and my mother's side, there was somewhat a reverse - very small maternal haplogrup, and very large paternal haplogroup. Maternal line matched populations in Poland, Ireland, and Arabs. Paternal line matched Ashkenazi Jews, Sephardic Jews, and Lebanese. I believe we have some very distant Italian Jewish cousins, though I don't know much about them. I know they're all generally very fair featured, and I believe from my maternal grandfather's side. My maternal Italian grandfather had strawberry blonde hair, and generally his side of the family has very light features (light brown or strawberry blonde hair, blue eyes, fair rosy skin) despite having roots to a tiny Italian town for centuries (where you'd think darker, Mediterranean features would dominate, related to trade routes with North Africa & the MidEast). Who knows?

I admit I know very little about this stuff, and there's much to refine, and a lot of historical variables, but it's pretty interesting and somewhat matches our inklings beyond the geographical roots we were aware of.
posted by raztaj at 11:12 AM on March 4, 2012

You might find this episode of Radiolab interesting. It's one hour long but the first part is about testing DNA for race. The conclusion they come to is the ability to test for race is iffy at best.
posted by catwash at 11:53 AM on March 4, 2012 [1 favorite]

There are several kinds of DNA tests. Autosomal DNA testing, like at 23andMe, is a newer kind and does look at all of your DNA from "all of your relatives" (more on this in a sec), unlike Y or mitochondrial tests that only look at your direct paternal (father's father's father...) or maternal (mother's mother's mother...) lines.

However, because of the randomness in passing down DNA from generation to generation, it's hard to use it to get an accurate sense of "how much" you have of a certain kind of ancestry, especially when you're looking at smaller numbers. Think about it this way-- your father has 50% of his DNA from his mother and 50% of his DNA from his father. He passes 50% of his DNA on to you. But which 50% he passes to you is random. So it is theoretically possible (although statistically very unlikely) that he could pass down the full 50% from his mother to you and 0% of the 50% from his father. So theoretically (again, statistically very unlikely but possible) you could have no DNA in common with your own grandmother. The further back you go, the more chances you have to get less than "expected" amounts of DNA from any given ancestor.

And then on top of that there's what Kid Charlemagne mentioned, that it's not like every bit of DNA is somehow clearly marked with what ethnicity it's from, so it depends on whether the part you got passed down is even clearly identifiable as coming from that ethnic group-- plus the more times DNA gets passed down, the more likely that it splits into smaller and smaller "chunks" of DNA interspersed with eachother, which muddies the waters further.

So DNA tests are really better at demonstrating a positive-- "Wow, looks like I really do have Native heritage"-- than proving a negative, and it's always going to be possible that you have a higher percentage of something in your ethnic background than is estimated based on your DNA (although the higher above the estimate you go, the less likely, but nonetheless still possible.)
posted by EmilyClimbs at 3:01 PM on March 5, 2012 [1 favorite]