23andMe Makes More Improvements to Ancestry Features

23andMe’s popular Ancestry Composition just got a little better.

The feature estimates what percentage of your DNA comes from populations around the world, breaking down the Asian, African or European ancestry to give you insight into the geographic origins of your ancestors going back many generations.

African American Example

An example of an African American’s customer’s Ancestry Composition.

Customers should start seeing the update to their Ancestry Composition over the next several weeks. It will offer people with African or Asian ancestry greater detail than they had before. We added several additional reference populations — for a total of 31.

Customer’s ancestry results are broken down by region, and then subregions and then even finer detail. So for example, we may breakdown the results of someone of African Ancestry by first identifying the percentage of “Sub-Saharan African Ancestry.” That in turn might be broken down between percentages of west, African, central and south African.

The update allows 23andMe to offer individuals with African ancestry more detail into whether their ancestors came from eastern Africa, western Africa, southern Africa or central Africa. This is similar to how 23andMe reports out data for people of European ancestry, but 23andMe is not yet able to break down African ancestry by country or tribe. We are currently working on an African Ancestry Project, which we hope will ultimately improve results for people with African ancestry.

An example of an Asian American's ancestry results.

An example of an Asian American’s ancestry results.

Customers with Asian ancestry will get much more detail into their ancestry, breaking ancestry between eastern Asia, southeastern Asia and south Asia. In addition the update will now provide much more detail for customers with East Asian ancestry by breaking down that ancestry among five countries — China, Japan, Korea, Mongolia and eastern Russia (people with Yakut ancestry).

23andMe is conservative in making ancestry assignments, so in many cases we may list a certain percentage of a customer’s results as “non-specific” to a particular region. A customer’s results may include a certain percent of “Non-Specific Northern European” ancestry, for example. That simply means that while 23andMe has identified the broader region of their ancestry, we do not have enough information to be able to predict the country of origin for that ancestry. We do offer customers the opportunity to select a more “speculative estimate,” which can sometimes give them more detail.






  • http://hartford.tumblr.com/ Tiffany H

    Anything more specific for South Asians this time around?

  • Sb

    Why do some of the options have question marks next to them for me? For instance, all of the Sub-Saharan African categories, the East Asian subcategories, and the Southeast Asian category? They’re where the percentage would be on the left side. I don’t know if that just means they need to be loaded for me in time, or if there’s something I should be seeing or what? Thanks for the help.

    • Sb

      Never mind, someone pointed out if I hovered over the question marks (I had tried to click on them, but not hover) that it said “Not computed yet” :)

  • Michael Belgrove

    I was wondering why Ashkenazi is now green instead of blue like the rest of Europe

    • 23blog

      So it is more easily identifiable.

      • Michael Belgrove

        Any particular reason why you would want to make it more identifiable among Europe? Are Ashkenazi that different?

        • Jean-Paul

          Yes, a Middle Eastern group with a smaller than surrounding populations admixes (more homogenous group), living in Central and Eastern Europe. This will mean that they definitely have a distinct genetic makeup.

  • Gerald Alexander Lopez Castell

    By using the Biaka/Mbuti and the San for “Central” Africa is very much useless for new worlders as none of us have direct ancestry from them, but rather indirect via the congolese/angolan, and i nkow you guys dont have these samples, but essentially all of us new world african descendants will come out like almost exclusively West African, with small bits of Pygmy which you guys are calling central-south african, and small bits of East African which is really Masai/Ethiopian, using the publicly available dataset of sudanese nilotes would make a nice extra east african category, i applaud the new technology and intiative but honestly, none of us are going to get much of a breakdown like this…. It would have been much better to use South-African bantu’s and the Luhya in Kenya for a “Bantu” cluster, or two bantu clusters, ive done this myself on admixture using only like 100k snp’s im sure you guys can do it bette, i dotn understand why this wasn’t done instead of using the pygmy populations which none of us come from.

  • puddytoes

    I was wondering that too.

  • Nabil

    Why are southwest asian populations like the Bedouins, Palestinians, Saudis and Yemenis grouped under *North Africa*? A *South Middle-East* or *Arabian* component wouldn’t be more appropriate for these populations ? The actual *Middle-East* could be a *North Middle-East* component and the *North Africa* will only refer to real north African populations like Tunisians, Mozabites(Algerians), Algerians and Moroccans

    • hernandayoleary

      Because the majority of people living in north africa today moved there during the colonial period from parts of the Ottoman empire in west asia like the Middle East. Remember this test goes back just 500 years.

      For the majority of north africans whose ancestors come from the middle east, your dna is going to come back as middle eastern so the 2 are just lumped together.

      Many of the modern North African people are genetically indistinguishable from Palestinians, saudis, yemenis, and people from the former ottoman empire aka “arab” countries. hence it actually makes sense. For instance, I am shown as having mozabite dna but its not listed as arab or north african but rather west african. Yet mozabites are algerian berbers, I guess its just the way they chose to seperate out indigenous algerians from ones with recent immigrant ancestry.

  • Anthony

    I’m h
    aving some doubts that 23andme’s African Ancestry Project will get a decent turnout. I’ve talked to probably over 50 West African residents on Facebook and only like 3 of them said they’d check it out, which doesn’t mean that they will check it out or that they will participate in it. Some others said flat out that they are not interested, and the rest just ignored it. I don’t think there will be a good turnout for the project. I hope I’m wrong.

  • Anthony

    African Americans will still get the low single digit “unassigned” and most will still get the less than 2% “East Asian & Native American” which is most likely statistical noise?

  • Yashna

    Are there any plans to expand on South Asian ancestry? Because all my analysis shows is South Asian and I would really like more details on that.

    • 23blog

      Yashna, We are continually updating the ancestry features. The team that worked on those features created a sort of “learning algorithm” so that as we get more data we can return more results. We will continue to make improvements and improve the results customers get.

  • 23blog

    Just to clarify, 2 percent is not likely statistical noise. Under 1 percent perhaps but over 1 percent while low is significant.

  • 23blog

    Anthony, we’ve actually been very fortunate to have a lot of individuals respond and have gotten a lot more individuals in a short amount of time than we thought. The sign up is still open at the moment.

  • April Harkness

    Uh, Just East Asian? This HAlf Filipina is DISAPPOINTED. There’s more to Asia than Korea, Japan and China- like all of SOUTHEAST ASIA. How about Vietnam, Cambodia, Indonesia and heck the Philippines..etc. Let me guess. I’ll be lumped into China….

  • Sucheta Potnis

    Hi. I am Indian and would like to participate. I am guessing this isnt a problem due to the DNA sample traveling over the distance? Also have some specific worry about a medical condition from my son in law’s family being passed on to my grand child. Would the DNA analysis of both parents help with predicting what are the chances for this? Thanks and look forward to hear your advice.

  • 23blog

    It depends on the medical condition. Here is the link to the health reports we do: https://www.23andme.com/health/all/.

  • Jean-Paul Armache

    Hi,
    I recently realized that my ancestry composition changed dramatically. From less than 1% estimate of one of the populations, now it indicates 13.5. Is it due to a growing number of participants and more detailed information, or could it something else ?

    • Scott23H

      We have added in additional reference populations and added more individuals to the database. Using that information we’ve been able to recalculate Ancestry Composition so some individuals are seeing changes. More often they are smaller changes, but in your case it looks much more dramatic. What was less than 1 percent that is now 13.5 percent? I’m assuming this was reassigned from a nonspecific category.

  • Scott23H

    Jean-Paul,
    I’ve got a comment from a guy that said his AC went from 12 percent Italian now that’s not indicated but it says he has 1 percent Middle Eastern.
    We’ve updated reference populations and also added more people to our database so that we are more able to accurate predict ancestry composition. As for the change in your AC some of this is a function of people with South European ancestry. Particularly people with Italian ancestry. Those predictions are a bit more difficult because of known similarities to Middle Eastern and North Africans. That said the Ancestry Composition we now return to customers is more accurate.

  • Scott23H

    Jean-Paul,
    We do now have additional Asian and African reference populations, but we also have a lot more customers. The prediction for AC is improved with that larger dataset.

  • Scott23H

    Jean-Paul,
    I’d say in general yes. The more numbers we get the more precision we can get in prediction. The predictive model built by our researchers improves as the database grows. Regarding your last question, are you asking about your Central Asian ancestry results or asking about our reference populations?

  • Stan

    I am quite disappointed with the updated to AC. I am not sure how to trust the new information as it is so different from the statistics that I used to see before the upgrade. It just shows that the level of “accurateness” it so flexible that with “another” upgrade you fill find yourself in a completely different category.

    • Scott23H

      Stan,
      I’m not sure about what was changed, so it would be hard for me to respond to your criticism. A more common experience is that the new assignments are more specific.

  • Scott23H

    Jean-Paul, Yes as we grow our database we will likely improve our ability to make Ancestry Composition predictions. And yes if we get additional reference populations — not just in Asia but in Africa as well — we will be able to return more detailed results to customers with ancestry from those regions.

  • Diana

    Are there any plans to divide Southeast Asia into specific countries? I just found out that I am 5% Southeast Asian and would like more details. Thanks!

  • Derek T Pomaranski

    I have a question. About a year or two ago my 2nd cousin on moms side took your test. He showed up as 99.4% European and 0.5% sub-saharan african. We share his mothers paternal side. His fathers parents are from Poland and his mother claims she has traced her mothers line back to the 1000AD. Her fathers line we have traced back in some cases to the mid 1700s. I do geneology as we’ll and neither she nor I have come across anythjng to make us suspect any african ancestry. In fact we are all very European looking. My mother and brother tan fairly we’ll, but her father is half Hungarian and he tans we’ll. these results do not reflect HIS lineage. What I am curious about is could this be a mistake, just “noise” as I have read, or somethjng else?

    • hernandayoleary

      Adjust ancestry composition to conservative, if it is still showing .5% sub-saharan then you with 90% or more confidence have sub-saharan dna. That is not surprising, the dna can go very far back particularly in the neolithic and ancient period there were many sub-saharan migrations into europe. .5% african is not enough to affect your appearance. There have been people with as high as 20-35% african dna and appear to be normal looking white people. One famous example is robin thicke’s son who is 25% african dna and appears to be blonde hair white.

      And although you may be able to trace your ancestry back in Poland to 1000 A.D. how do you know that some of those ancestors were not black? Polish doesn’t always equal white.

      Władysław Franciszek Jabłonowski was a black polish brigadier-general who fought with Tadeusz Kościuszko in the Kościuszko Uprising against imperial russia and Prussia. He was the leader of the 6000 man danube legion there were other black Polish in the 1700s of prominence like George Bridgetower famous classical musician. And this was in the 1700s. The only reason we know he was black is because he is mentioned in a letter by Napolean and Davout who were his classmates. Had he not been of prominence no one would ever know he was black and he would just be assumed to be Polish and therefore white.

      You might find there was actually a sizeable presence of sub-saharan people in north/central europe during the late middle ages, some of them like dumas, pushkin, king medici, gannibal and queen charlotte, Chevalier de Saint-George even rose to prominence as aristocrats, kings, queens, generals, world renown writers and commanders in the army. Some are more controversial like louis xiv as some documents describe him as black, he was calld the sun king and he has a daughter whose portrait appears black Louise Marie-Therese, the black Nun of Moret but his portraits appear to be white. Europe has been multiracial and multiethnic for a very long time. I’d be more surprised by someone who was only white than someone with small 5 or 10% african ancestry at this point. Just read around the board, there are few of us here who have no african ancestry regardless how we look.

  • ULGHAN

    Why 23andMe ignored Central Asian populations in its AC composition?
    They put some Central Asian like Uygurs and Hazaras in wrong compositions (south Asian) and put aside others like Kazakhs, kyrghyz, Uzbecks…
    Is any plan to put them in AC in the future or not?

    • Scott23H

      Ulghan, Central Asian populations are not being ignored, but we still need to add to our reference populations to be able to more accurately make Ancestry Composition estimates. We are constantly updating and improving those estimates and hope to have more to offer in the future.

  • Rick Nuthman

    How does ‘speculative’ work? Is this just an educated guess as to where the unspecified regions represent?

    • Scott23H

      Hi Rick,
      For each section of your DNA, Ancestry Composition assigns a probability to each reference population for a given segment of DNA. We have our own threshold where we have a very high confidence in the estimate we report to you, but we give customers the ability to adjust that threshold using either a conservative or speculative estimate.

      The default Standard Estimate corresponds to 75% confidence. You have the option to make the estimate more strict (the Conservative Estimate is at 90% confidence) or more lenient (the Speculative Estimate is at 50% confidence).

      You can read more about how the confidence threshold works in the “Aggregation & Reporting” section of the Ancestry Composition Guide.

  • Was wondering if your Native American results will get better in time. I downloaded my raw data and sent it to Dr. Doug McDonald who confirmed, in his words, a “very real” 1.3%-1.4% Native American. But, 23andMe does not show it. If I put it on Conservative then it shows 1.4% Unassigned. I don’t feel that is a coincidence. So, in time if that 1.4% Unassigned is indeed Native American will it show eventually?

Return to top