Unlikely cousins on 23andme

October 30, 2016 7:57 PM

My knowledge of biology is minimal, so bear with me. I'm on 23andme, which shows me a list of relatives. From the names and biographical information, most look accurate (same South Indian ethnicity as mine). However, there are 2 people on the list who are white American and don't appear to have any Indian connection, who share about 0.10% of my DNA. I don't have white ancestors either, AFAIK. Is this a sequencing error? A coincidental common mutation? Or an actual shared ancestry?
posted by redlines to Science & Nature (16 answers total)

.1% is 4th or 5th cousin range, which means you may share a great-great-great-great-grandparent, which you have 64 of - it is unlikely you know the ethnicity of all 64, but even if you do, perhaps one of their kids who is not a direct ancestor of you moved to America and had kids with a white person, so they would be 1/32nd South Asian. It seems pretty likely that that's a real cousin connection.
posted by brainmouse at 8:12 PM on October 30, 2016 [12 favorites]

They might be descended from South Indian people who worked overseas as lascars or in some other capacity and married a white European whose descendants moved to America. Definitely a lot of moving around, but not impossible.
posted by Small Dollar at 8:25 PM on October 30, 2016 [2 favorites]

At that generational range, you are also looking at British occupation of India and the possibility of an Anglo-Indian relationship which then later moved to America.
posted by Thella at 8:44 PM on October 30, 2016 [10 favorites]

.1% is 4th or 5th cousin range, which means you may share a great-great-great-great-grandparent,

Note also that the hypothetical common ancestor would have been born about 180 years before you, or in the early part of the 18th century. There were a good number of Europeans in the subcontinent at that time. And even if you don't have a European among your "official" ancestors, it's all too believable that a white European had a dalliance with one of those ancestors—either willingly or (shudder) unwillingly—and chose to keep it secret.
posted by Johnny Assay at 8:45 PM on October 30, 2016 [8 favorites]

You may not have a European in your tree, but they can have one of your ancestors in theirs.
posted by cecic at 8:46 PM on October 30, 2016 [16 favorites]

You really don't have to go back that many generations before absolutely everyone is related to you (or no one). It's something like 9th or 10th century for Europeans, not as many more as you would expect for literally everyone alive right now.

0.1% overlap means nothing, it's random chance, particularly given 23andme only measures a list of polymorphisms rather than full sequence anyway.
posted by shelleycat at 10:28 PM on October 30, 2016 [5 favorites]

The other thing to remember is that you and I might not have ancestors in common since the time our ancestors followed some mammoths out of Africa, but we're going to have DNA in common just because we both kept the ability to digest starch and make hemoglobin. In this case, of that 0.1% that matches, there's no guarantee it all came down one branch of their (or your) family tree.

It's not like Europe and India were unknown to one another until modernity. The Roman empire had pretty robust trade with India.
posted by Kid Charlemagne at 11:47 PM on October 30, 2016

you and I might not have ancestors in common since the time our ancestors followed some mammoths out of Africa

Oh you absolutely do, everybody does. The most recent common ancestor is far more recent than that. More like 2000-3000 years ago. Yay for globalisation.

Also people usually have less than 64 great-great-great-great-grandparents. It's the only way the numbers work out. Human ancestry is not neat and straightforward and we are all very inter-related.

23andme results need to be taken with a giant pinch of salt for these and so many other reasons. Trying to read into 0.1% of your results is a waste of time. Focus on the big numbers in your results which have at least some chance of being meaningful.
posted by shelleycat at 12:13 AM on October 31, 2016 [10 favorites]

Oh you absolutely do, everybody does. The most recent common ancestor is far more recent than that. More like 2000-3000 years ago. Yay for globalisation.

The citations I'm finding on Wikipedia for figures like this, such as this one, appear to be based on computer modeling of populations of Sims-like virtual analogues to humans, rather than actual real-world comprehensive analysis of genetic diversity in modern humans.

It seems counterintuitive, because for example this would mean that a member of an Amazonian uncontacted people would somehow have a common ancestor with present-day Africans, Europeans, Asians, and Australians, and the living members of other groups around the world who were only contacted within the last century; an ancestor who lived during a relatively recent time when the Americas had no recorded or archaeologically-corroborated contact with those other continents.

The Wikipedia entries for Y-MRCA and mt-MRCA give dates much further in the past for a common ancestor.
posted by XMLicious at 1:18 AM on October 31, 2016 [3 favorites]

Whereas I was reading articles in Nature etc rather than bothering with Wikipedia or other low-quality sources. There are multiple strands of evidence at this point, hence the wide time range. Yes, there is uncertainty over uncontacted populations, but that's not at all relevant for anyone using 23andme.
posted by shelleycat at 2:13 AM on October 31, 2016 [2 favorites]

It's just random chance. The technology doesn't exist to determine whether people are related with that degree of fidelity, and certainly not at 23andme prices.
posted by miyabo at 6:07 AM on October 31, 2016 [1 favorite]

My understanding is that 23 and me relative-matches (and ancestry divining) are based on a small portion of DNA. not the entire starch-digesting, 2-arm-having, hemoglobin-making DNA. If they used that we would all be 99% similar. Instead they just look at the bits that vary much more widely. So if they say 10% similar, that means that the parts that people have that are different from one another are 10% similar for you. If they compared everything then you'd be 99% similar to them, just like you're 99% similar to everyone else and 97% similar to other random primates.

So I wouldn't just assume this is random we're-all-human similarity. I would assume they're real distant cousins, and assume further that you don't know all your ancestors back that far, nor what became of every single one of their descendants: After all, you didn't know about the cousins you find more plausible, right?
posted by If only I had a penguin... at 7:46 AM on October 31, 2016 [3 favorites]

Whereas I was reading articles in Nature etc rather than bothering with Wikipedia or other low-quality sources.

Could you give references? The only peer reviewed paper I've read with the low estimate (probably the same Nautre paper linked above) was pretty much what XMLicious said--a computer model with assumptions about mobility between groups.

It was interesting in that under plausible models you didn't need many people intermigrating to come up with that estimate, but it's not empirical.
posted by mark k at 8:13 AM on October 31, 2016 [5 favorites]

The databases these commercial companies use are not shared. It's possible you'd get different results if you ran the same DNA through a different company's database.

23 and me did a promotional video with buzzfeed. The rep told one half-Asian, half-Black woman that her heritage was 100% African. Just saying
posted by Neekee at 9:11 AM on October 31, 2016

That 2004 Rohde/Olson/Chang paper is not a good basis for asserting that the most recent common global ancestor lived at a specific date range. That's purely computational modelling, with almost no foundation in population genetics or anthropology. It's an interesting piece of work, but it's in the "Letters to Nature" section, not a fully formed article. Being published in a famous journal doesn't make one's curiosity project into settled, consensus material.
posted by late afternoon dreaming hotel at 1:03 PM on October 31, 2016 [2 favorites]

I'm sorry for getting sidetracked. The most recent global ancestor is not relevant here, because we are talking about two populations that are already definitely intertwined much more than the entire globe. Please drop it and focus on answering the question.

If you have data to show why such a small overlap in 23andme results is meaningful then have at it. But given how they do their studies and given how much of what they report is spurious anyway, I see no reason to read anything at all into such a tiny number (as a scientist with a PhD in a relevant area).
posted by shelleycat at 10:55 PM on October 31, 2016