Friday, 2 February 2018

My admixture results: A comparison between three companies


In this blog post, I will be discussing and reviewing my admixture results from three of the "big four" genealogical DNA testing companies: 23andMe, Family Tree DNA (FTDNA), and MyHeritage. (I haven't yet tested with Ancestry). Admixture is often talked about as something to do "for fun". I've wanted to write a piece on admixture for a long time, because I think admixture really is fun, because it's a popular and controversial topic within genetic genealogy, and because I wish to help other people understand how they can be interpret their admixture in light of genealogical paper trails as well as other family members' DNA results. A lot has been written about this topic from an American point of view, while voices from the rest of the world have been relatively silent, except perhaps on YouTube. Through this post, I will be presenting a European and Scandinavian perspective on admixture interpretation.

Some people, perhaps especially in Scandinavia, have a tendency to dismiss every unexpected admixture result as statistical noise. A few even go as far as dismissing the concept of admixture estimates altogether. Personally I think admixture results can be extremely helpful to family history research if they're investigated and interpreted properly. And because of what I know about history and human beings, and because of what I've seen of the world, I never dismiss out of hand the possibility of human migration or movement anywhere in the world.

Note: Many people, including many genetic genealogists, seem to understand the term "ethnicity" as a synonym for "race". As a social anthropologist, I disagree with this use of the term. Coming from the tradition of the anthropologist Fredrik Barth, I see ethnicity as a social phenomenon, amounting mainly to a combination of self-identification with, and acceptance into, a social group that sees itself as different from others. In my blog, I will try to stick to the terms "admixture" and "biogeographical ancestry" when talking about "ethnicity" in the biological sense.

For the record, here is a diagram of my known/probable ethnic ancestry since the year 1500, which includes any ethnic category - in the non-biological sense - that can be tied to specific ancestors in my tree. The diagram is based on all available sources, including evidence from genetic matching, and likely NPEs have been taken into account (see my blog post about this).




My 23andMe results (Ancestry Composition)

I will start off by discussing my 23andMe results (at 50% confidence level, the default view), which I find to be both detailed, interesting and accurate:















I am the only one in my family who has tested with 23andMe so far, but several others have tested with FTDNA. For this reason, I will be mentioning their FTDNA results in relation to my 23andMe results, since that is the only data I've got.

According to 23andMe, their admixture estimate goes back approximately 500 years. This is convenient, since this time period largely falls within a "genealogical timeframe", which makes it easier to compare the results to my known ancestry according to the paper trail. I've spent many years researching my family tree using traditional records, and this knowledge has been extremely helpful to me - in fact invaluable - in interpreting my admixture estimates and other results of my genealogical DNA tests. The more you know about your "paper genealogy", the better.

I come from Norway, and the vast bulk of my known ancestry is ethnically Norwegian, with Swedish as a distant second. The size of the Scandinavian and Finnish percentages at 23andMe are very consistent with my paper genealogy. The British & Irish is also very accurate, especially if I really do have an English 3ggf (and that particular family story does seem to be true Yes, it did seem so, but it has been proven wrong). The Broadly Northwestern European probably represents a mixture of my documented German, Danish, Frisian, Dutch/Flemish and Northern French ancestry; it's also possible that some Scandinavian and Finnish might be hiding in this category. The Eastern European and Southern European categories are consistent with my paper trail, but the percentages are a bit higher than expected. My mother gets 5% Eastern European at FTDNA, which indicates that my 0.6% at 23andMe is real and not statistical noise even though it's a small amount. This DNA might well be connected to our family's story of Russian ancestry.

In addition to the expected European percentages, 23andMe also gave me some trace categories that came as an exotic surprise to someone like me, who is not from a region (such as the USA) that is known for being highly admixed. The first surprising category is the Native American. In certain admixture calculators, notably the ones at GEDmatch, "Native American" will frequently show up in people of Sámi, Finnish and Central/Eastern European descent as a misreading for ancient Siberian or Central Asian admixture (see Roberta Estes' blog post about Finding Native American Ethnic Results in Germanic People). 23andMe, however, does not usually show Native American erroneously, even in small amounts.

Through autosomal cousin-matching combined with detailed chromosome painting comparisons at GEDmatch, I have come to the conclusion that a certain 4th great-grandfather of mine was probably the illegitimate son of a man from Arcahaie (Lakayè) in Haiti, who was born in 1788 as a free person of mixed African and European descent. This has been proven wrong, but the Native American DNA might still be real, and it might even be Taíno. My grandmother's brother (from the same ancestral line) has tested with FTDNA and shows up with a trace of <1% South American. There are no known Sámi or Finnish or Eastern European ancestors on this side of my family. I believe that our Native American DNA is real and that it comes from a very distant ancestor - perhaps living in the 1500s - who was a member of the Taíno people, the indigenous people of Haiti and other islands in the Caribbean. In other words, our Haitian ancestor's background seems to have been not just biracial, but in fact triracial. The Taíno and their language originated in South America (either in the Andes mountains or the Amazon basin), which explains why our Native American DNA is read as specifically South American.

Edit: My Native American segment disappeared in the April 2018 update. It seems that it was really an East Asian segment, unrelated to my grandmother's brother's South American (which I still believe is real).

Next, the East Asian. This percentage probably is a sign of Siberian ancestry more than 1000 years ago, and other family members' results show that it comes specifically from my paternal grandfather. He has tested at FTDNA and gets traces of West Middle East, Northeast Asia, Southeast Asia as well as Oceania. My grandfather is almost 100% Norwegian, except for one Finnish 3ggf (born before 1790), one 4ggm (born 1769) who was most likely a South Sámi woman, and a few very distant ancestors from Denmark, Germany and Scotland (1500s and 1600s). In addition, since almost all of his ethnically Norwegian ancestry is from the South Sámi homeland (Saepmie) or from areas formerly populated by South Sámi people, my grandfather probably has a much higher degree of ancient Sámi ancestry than is evident from the paper trail. The most likely explanation for my grandfather's Asian DNA is Siberian admixture through his Sámi and/or Finnish ancestors, and from him this DNA has been passed down to my father and myself. I assess it as being real but ancient, which is not the same as "noise".

The final surprise was the East African. This is certainly not something that commonly pops up in the 23andMe results of Norwegians. Could it be random noise? Or a misreading for something else? In fact, when I compare my chromosome paintings with my mother in GEDmatch, she shows up with East African DNA in exactly the same spot as me, which shows that (A) my East African DNA is real, and (B) that it comes from my mother's side of the family. There also seems to be some East Asian, or possibly Oceanian, mixed in with the East African segment. TL Dixon, the eminent owner of the blog Roots & Recombinant DNA, believes that the segment probably comes from a very distant ancestor from Madagascar, whose inhabitants, the Malagasy, descend from Austronesians from Borneo and Bantu peoples from Southeast Africa. Europeans landed in Madagascar for the first time in the year 1500, and the island was part of the Trans-Atlantic slave trade as well as the Indian Ocean slave trade. My mother probably descends from a Malagasy slave who was taken to Europe sometime in the 1500s or 1600s, or from a Malagasy descendant who travelled to Europe for unknown reasons, perhaps as a sailor. My mother's African DNA has nothing to do with my Haitian connection, which is on my father's side.

Edit: In the April 2018 update, my East African segment disappeared and is now read as "No Data Available". However, based on the GEDmatch verification and TL Dixon's advice, I still believe that the segment is there and that it is real.

My Family Tree DNA results (myOrigins)



These results are rather coarse-grained compared to 23andMe, but still quite accurate. My main region is still Scandinavia (91%). Anything else would have raised my suspicion at once! The most obvious difference between 23andMe and FTDNA is that FTDNA does not report as many trace regions as 23andMe does. The one trace region I do get, <1% Northeast Asia, is consistent with 23andMe's <0.1% Broadly East Asian.

FTDNA gives me twice as much British as 23andMe does. This might seem strange at first, but different regions within Europe are actually quite difficult to distinguish genetically, and some of the "British Isles" might actually be Scandinavian, Northern German, etc. Or it might be real British DNA, but from an earlier point in time than 23andMe covers (i.e. more than 500 years). People have been moving between Norway and the British Isles since the Viking Age and probably even earlier. I am not worrying too much about discrepancies in the sub-regional percentages; these tests are most accurate on the continental level (e.g. Africa versus Europe).

The one weird - and probably wrong - part about FTDNA's analysis is that I get absolutely no Finnish. Finnish DNA has been found to diverge significantly from other Europeans, and should be detectable by this test. In the old version of FTDNA myOrigins, I used to have 10% "Finland and Northern Siberia". I have documented Finnish ancestors (1600s and 1700s) as well as a high number of Finnish cousin matches at FTDNA and elsewhere. In other words, there should be some Finnish showing up. Most likely it's being misread as Scandinavian (note that my Scandinavian percentage is significantly higher here than at 23andMe), or as Scandinavian plus some Northeast Asian.

My MyHeritageDNA results


















MyHeritage is a well-known company that has been in the family tree business for years, but their DNA testing service is rather new. MyHeritageDNA is very much an up-and-coming phenomenon in the world of genetic genealogy, especially among first-time testers.

While not inaccurate, my admixture estimate from MyHeritage is just as coarse-grained as FTDNA's, and includes only my main Northern European regions. Interestingly, MyHeritage gives me Finnish at a very significant level similar to what I got in the old version of FTDNA myOrigins. 10.5% is probably a bit too high, but at least the Finnish is there, as it should be.

I haven't actually tested with MyHeritage; I've only uploaded my FTDNA raw data. It is possible that my results at MyHeritage would be more fine-grained and interesting if I actually did their test. Also, their admixture analysis is still in Beta and will certainly improve in the future.

What I really like about MyHeritage is the nice layout of the admixture map. The musical presentation is also nice, not least since they actually managed to include one of my favourite songs in their medley :)

Conclusion


All three companies agree on my main regions of genetic ancestry: I am mainly Scandinavian, with the British Isles as my second region. This strong Northwestern European genetic signal is fully expected and compatible with my paper trail.

I like 23andMe best. Their analysis gives me the most high-resolution results, detecting trace amounts of admixture that seem to be real even though they're very small. 23andMe also gives me the most accurate percentages of European regions as compared with my paper trail. In addition, 23andMe has very good tools for interpreting your results; most notably, you can change the confidence level in order to remove the more speculative percentages.

FTDNA and MyHeritage have more or less the same level of resolution. FTDNA detects my trace of East Asian DNA, but overlooks my Finnish. MyHeritage detects significant Finnish but overlooks the trace of East Asian. None of them report my trace amounts of Eastern European, Southern European, Native American or African, although these trace amounts have been verified through other family members' results, as well as by other means.

I plan to test with Ancestry one day. I look forward to comparing their analysis to the three that I've been discussing here.

2 comments:

  1. Hi Miriam---I tested with Ancestry and got 25% Western Europe, 21% Finnish/Northwest Russia, 18% Irish, 12% British, 12% Scandinavia, 10% Iberian Peninsula, and trace amounts of Eastern Europe, Italy, and Greece.

    I uploaded my DNA to FTDNA and before they changed their MyOrigins recently I got British Isles 31%, Western and Central Europe 28%, Finland and Northern Siberia 25%, Scandinavia 9%, and Eastern Europe 7%. All fairly close to Ancestry. Then, they changed MyOrigins and now I get Scandinavia 37%, British Isles 26%, Southeast Europe 19%, Finland 18%, and trace results of Eastern Europe. No central Europe at all! I have documented Dutch ancestors who founded New Amsterdam (New York City) and documented German/French ancestors who came to the U.S. in the 1700's. It seems they all have disappeared. :)

    I recently did my mtDNA haplogroup on FTDNA and that came back U3a. My maternal great-grandmother was from Våler, Hedmark, Norway.

    My Heritage shows no Scandinavian at all. North and West Europe 83%, Finnish 14.7%, English 2.3%. No trace amounts of anything. DNAland has me at 66% Northern European, Finnish 31%, Southwestern European 2.7%.

    I've run my results through Gedmatch, and that has been fun and interesting. I guess a lot depends on which reference group these sites use.

    ReplyDelete
  2. Very fascinating reading! I'm sure your Scandinavian at MyHeritage is somehow hiding in the North and West Europe and/or Finnish.

    It's interesting that your Dutch and German/French ancestors "disappeared" in FTDNA myOrigins. The same happened to mine. I've got well documented German, French, Frisian and Dutch/Flemish ancestors in the 1600s and 1700s. Their DNA should be in the "Western and Central Europe" category, of which I get 0%. However, they amount to only 2.5% of my total ancestral background, which means their DNA is probably easily misread as something else, perhaps "British Isles" or "Scandinavia".

    ReplyDelete