23andme’s new African breakdown put to the test

My first DNA test ever was with 23andme. Nine years ago already! In January 2010 I was thrilled but soon afterwards also quite underwhelmed to receive my very basic admixture results. The only distinction being made back then was between African, Asian and European DNA. Native American DNA did not even have a separate category yet 🙂 As I am of Cape Verdean descent I was actually most anxious to have my Upper Guinean lineage confirmed. Instead my African score just pointed towards the entire continent! One of my immediate reactions at that time therefore was:

“I hope that one day 23andme’s Ancestry Reports will be helpful in finding out where to locate my ancestry regionally and not just on a continental scale.”

After a (very) long wait it seems that this day has finally arrived! Last month 23andme rolled out an updated version (3.0) of Ancestry Composition to all their customers. Regardless of when they originally took the test. This update has actually been on release since September 2018 for 23andme’s most recent customers. But to its credit 23andme also made this update available to its earliest customers, like myself. Over the years I have been through more than one update on 23andme already. But this is the first time I can say that finally a meaningful African breakdown is being provided! For more details see:

• New African & East Asian Details in 23andMe’s Latest Ancestry Composition Update (23andme blog, 2018)
• 23andMe Adds 1000+ More Regions and 30+ New Reports for Our Most Refined View of Ancestry To-Date (23andme blog, 2019)

***

Figure 1 (click to enlarge)

Updated 23andme results from across the African continent. A small but representative sample. Highlighting how 23andme’s new African regions appear to be quite predictive, for native Africans themselves. Unrealistic expectations about “100% accuracy” as well as counter-productive obsessing about regional labeling should be avoided. Instead take note of how the expected regions (circled in red by myself) reach levels of over 70% reaching into 98%! Taking a macro-regional perspective (combining overlapping regions from within West Africa versus Central/Southern Africa versus Northeast Africa) these results are usually in line as well. Also the additional ancestral locations appearing below the regional scores are on point!

***

I have always believed that the best way to find out about the predictive accuracy of any particular DNA test or update is to look at the results of people who actually know their (recent) origins. In order to improve correct interpretation I have therefore started a survey among African DNA testers (n=173). Using their group averages as some sort of rudimentary benchmarks so to speak. Similar to the survey I conducted among African AncestryDNA testers in previous years (see this page). Of course also some basic knowledge about DNA testing (in particular 23andme’s reference populations and methodology) as well as historical context will remain essential to really get the most out of your admixture results!¹

Main topics if you continue reading:

1. Survey findings for 173 African 23andme testers from 31 countries (incl. 25 Cape Verdeans)
2. Maps showing the geographical distribution of the new African regions on 23andme (based on my survey findings)
3. Implications for Afro-Diasporans
4. Examples to illustrate how regional admixture DOES matter!

Twelve new African regions on 23andme

***

Figure 2 (click to enlarge)

Nine specific regions and 3 broadly macro-regions are now available on 23andme’s Ancestry Composition to describe your origins across the African continent (aside from “Broadly Sub-Saharan African” and “North African & Arabian”). The country name labeling is not to be taken too literally, as always. But it is actually quite indicative if you simply take it as a proxy and also take into account surrounding countries. Despite being less specific it will still also be helpful to distinguish between macro-regional areas within Africa: “West African” versus “Central & Southern East Africa” versus “Northern East Africa”.

***

In the past I have been quite critical about how 23andme neglected its customers who tested to learn more about their ancestry rather than for 23andme’s health reports.² Despite announcements being made already in 2013 (over FIVE years ago! see this link) the promised specification of African DNA kept being postponed. Seemingly as a low priority on 23andme’s list. Even when AncestryDNA demonstrated that such an endeavour was perfectly feasible in that very same year of 2013 (see this link).

The former African breakdown (2013-2018) on 23andme consisted of a mere three regions: “West African”, “Central & South African” and “East African”. Arguably only the distinction being made between Northeast African DNA and Niger-Congo DNA could be considered useful. The misleadingly labeled “Central & South African” region only serving to single out Hunter-Gatherer DNA (Pygmy & San). Which is usually minimal for Afro-Diasporans, except South African Coloureds. Otherwise this former African breakdown on 23andme was utterly uninformative for Afro-Diasporans. And in fact also for many Africans seeking a regional specification of their origins. Going beyond any bland “West African” score. Which to add insult to injury was actually also measuring Bantu origins from Central & Southern Africa! For more detailed discussion see:

• Ancestry Composition (2013-2018)

But I have to admit that several exciting developments have been taking place on 23andme since 2018. Improvements in 23andme’s admixture reports as well as new features such as recent ancestral locations could very well be beneficial to anyone wanting to learn more about their ancestry. In upcoming blog posts I intend to explore in greater detail these new opportunities arising on 23andme. To keep things focused this blog post will however only deal with the following question: how well does 23andme’s describe the ancestry of native Africans after its recent update? 

As shown above in figures 1 & 2 the number of African regions in 23andme’s Ancestry Composition has increased significantly. From only three to twelve (leaving aside “broadly Sub-Saharan African” and “North African & Arabian”). With all main areas of Africa (south of the Sahara) being represented. In my evaluation below I will disregard unrealistic expectations about “100% accuracy”. Also counter-productive obsessing about regional labeling should be avoided. Instead I will focus on the informational value to be gained despite shortcomings. There is much to be learnt from the updated 23andme results being reported for native Africans. Despite being less specific it will still also be helpful to distinguish between macro-regional areas within Africa: “West African” versus “Central & Southern East Africa” versus “Northern East Africa”. This is actually an additional regional framework I also applied during this survey³.

Some people might object against the inclusion of the “broadly West African”, “broadly Northern East African” and “broadly Congolese & Southern East African” categories in 23andme’s new African breakdown. As many people (unrealistically) prefer to have their origins specified from A to Z. Only being content with a seemingly exact ethnic designation. But personally I think it is a more honest approach for DNA testing companies to not be more specific than their data supports. As after all unspecified or “broadly” results will ultimately be better than false results. You simply cannot always have clear delineation on all fronts. Things are bound to get blurry at some point. Due to recombination your genetic make-up wil often not fully align with genealogy. Also the impact of ancient migrations, inter-ethnic intermingling etc. is too often underestimated as a primary cause of widespread genetic similarity between neighbouring populations.

This (inevitable) circumstance should however not prevent you from taking advantage of the major improvements carried out by 23andme! West Africa now has 3 more specific sub-regions: “Senegambian & Guinean”, “Ghanaian, Liberian and Sierra Leonean” (initially labeled “Coastal West African”) and “Nigerian”. A very useful start in narrowing down West African ancestry to more meaningful areas. For the first time 23andme is also making a genuine attempt to distinguish between West African & Central African DNA. A crucial pre-condition to understand the African roots of Atlantic Afro-Diasporans! Central & Southern East Africa now consisting of three sub-regions. No longer only the marginal “Hunter-Gatherer” region but also “Congolese” and “Southern East African”. Northeast African DNA was already well defined in the previous version. And amazingly it is now being further refined into: “Ethiopian & Eritrean”, “Somali” and “Sudanese”.

African breakdown for 173 African 23andme testers from 31 countries

Table 1 (click to enlarge)

This table contains my main survey findings: the group averages for 173 Africans from 31 countries. It illustrates how 23andme’s new African breakdown is performing for people of known background. Click on this link for an expanded & up-to-date version of this table. Also including group averages for “Broadly West African”, “Broadly Congolese & Southern East African” and “Broadly Northern East African”. These scores have been left out of this overview merely because of lack of space.

***

Over 170 samples with backgrounds from at least 31 different African countries have been used for my main survey findings featured above.⁴ This seems like a reasonably robust number and a wide enough array to pick up on some preliminary patterns. Even when for most of the separate nationalities I was only able to obtain a minimal sample size. Obviously these findings are not intended to reflect any fictional national or ethnic averages! The main purpose of this overview is to give an approximate idea of what to expect when wondering about how 23andme’s update has affected the results of their African customers. See also this spreadsheet which contains all the individual results I used for my survey findings:

• 23andme results of 173 Africans after update

For screenshots of the individual results and more detailed discussion see also:

• 23andme results from West Africa (part 1, Senegal to Ivory Coast)
• 23andme results from West Africa (part 2, Ghana to Nigeria)
• 23andme results from Central & Southern Africa
• 23andme results from East Africa

I should first of all point out that due to space restrictions I did not include group averages for “Broadly West African”, “Broadly Congolese & Southern East African” and “Broadly Northern East African” in Table 1 (follow this link to see these scores specified as well). However you can combine all colour-coded regions (either green, yellow or red) for additional insight. You will obtain the total West African, total Central & Southern African and total Northeast African scores (labeled “subtotal + broadly”). And despite being less specific I find that this macro-regional framework actually reinforces the predictive accuracy of 23andme’s new African breakdown!

As one might reasonably expect West Africans (from Senegal to Nigeria) are described by 23andme as being around 95% “West African”. Central & Southeast Africans (from Congo to Mozambique to Kenya) are now (for the first time!) correctly described as being predominantly “Congolese & South Eastern African”, around 80%. And most impressively Northeast Africans (from Sudan to Somalia) are around 98-99% “North Eastern African” according to 23andme. Admittedly countries from intermediate zones (like Mali, Cameroon, Ruanda, Uganda etc.) fall somewhat in between. But such outcomes still make sense if you keep in mind overlapping geography & genetics and ancient migrations across the continent.

One of my main qualms about 23andme in the past has been the lack of clarifying context. They certainly have made some attempts to improve themselves in this regard. But overall I have to say the information provided on their website to help you make more sense of your results is still inadequate. In particular the potentially misleading country name labeling of most of the new regions is not being rectified by mentioning how these regions are in fact almost always also descriptive of DNA found across borders!

This is why I will provide self-made maps below displaying the wider geographic distribution of each region, according to my preliminary survey findings. I will only leave out the “Hunter-Gatherer” region because basically it was hardly showing up among my survey participants. Including also for Central Africans and Southern Africans, who previously received much more noticeable scores for this region.⁵

It is much to take in and therefore I will only focus on the main tendencies for each part of the continent. At times I will draw comparisons with the former African breakdown on AncestryDNA. Unless stated otherwise I will always be referring to the old version of AncestryDNA, current between 2013-2018! As I believe that Ancestry’s last update of September 2018, unlike 23andme’s current update, has not been beneficial for Africans and Afro-descendants. See also:

• African breakdown on AncestryDNA for Africans (2013-2018)
• Did Ancestry kill their African breakdown? (part 1)

“Senegambian & Guinean”: good proxy for Upper Guinean DNA 

Map 1 (click to enlarge)

***

Quite accurate indicator of Upper Guinean DNA. Including also for Cape Verdeans! I will publish a separate page with my Cape Verdean survey findings eventually. Similar to the “Senegal” region on AncestryDNA (prior to the 2018 update!). But probably somewhat more predictive, especially for Guineans. Extending also into northern Sierra Leone and western Mali.

Not shown in the map but actually all my survey participants of Fula descent in Nigeria, Niger and Sudan also received significant “Senegambian & Guinean” scores. Confirming their western origins, from the historical Fula heartlands of Futa Tooro in Senegal & Futa Djallon in Guinea Conakry (see this page for several maps on Fula migrations).

***

“Ghanaian, Liberian & Sierra Leonean”: less predictive for eastern Ghanaians?

Map 2 (click to enlarge)

***

Seems to be somewhat less accurate for Ghanaians than for Sierra Leoneans and Liberians, sofar. Which might be correlated with ethnic background. Still many times quite impressive predictions, also for Ghanaians of Akan descent. Especially when combined with the ancestral location feature. Goes beyond the three countries mentioned in the labeling. As in fact this region is also very descriptive of DNA found in Burkina Faso and naturally the Ivory Coast. Because of the inclusion of Sierra Leonean samples overlapping into Upper Guinea as well. But also to the east reaching considerable scores in Benin (30%) and even among two of my northern Nigerian samples (17%). Somewhat unsatisfactory grouping therefore. Especially given the need of distinction between origins from the Gold Coast, Wind Coast and Bight of Benin for Afro-Diasporans. When properly interpreted still useful though.

***

“Nigerian” region also covers Ghana, Benin and Cameroon

Map 3 (click to enlarge)

***

Impressive coverage of southern Nigerian DNA for Nigerians themselves. Group averages of around 90% for my Bini/Edo, Yoruba and Igbo samples! With little variation in between. Only northern Nigerians clearly receiving lower “Nigerian” scores. Much better than on AncestryDNA! However considerable overlap also with DNA found to the west and the east of Nigeria. Native Ghanaians and Cameroonians without any recent Nigerian lineage will still often show “Nigeria” scores in excess of 30%. Ironically reversing the situation on Ancestry. Not a perfect outcome therefore. But it might be preferable for many Afro-Diasporans to have their Nigerian lineage overestimated rather than confusingly mislabeled by Ancestry’s “Benin/Togo” and “Cameroon/Congo” regions.

Not shown in this map but during my survey I found that substantial amounts of “Nigeria” can be useful also to indicate Krio and Americo-Liberian background. As well as related Aku lineage among Gambians. Due to the absorption of Recaptive Africans from southern Nigeria but also other places (such as the Congo) leaving a distinctive genetic imprint among certain population segments of Sierra Leone, Liberia and Gambia.

***

“Congolese” region describes DNA across Central & Southern Africa

Map 4 (click to enlarge)

***

The new socalled “Congolese region is quite predictive, but not restricted to Congolese origins only! In reality it is measuring genetic similarity among Bantu-speaking populations who are dispersed over a far greater territory! As my survey findings clearly demonstrate this “Congolese” region can be found as far south as Mozambique and South Africa! To the north it also has a substantial presence in Cameroon. But more so among certain ethnic groups closely related to Bantu populations. Given the absence of Cameroonian samples on 23andme (unlike Ancestry!) this creates considerable genetic overlap with “Nigerian”, for understandable reasons. To the east there is also some overlap with “Southern East African”. Especially among Bantu populations of Kenya and Tanzania which have a lower degree of additional Nilo-Saharan or Cushitic lineage. Not unlike how the former “Cameroon/Congo” region on AncestryDNA also expanded further south than expected (based on the labeling).

Leaving terminology aside I do think it counts as a true improvement that almost all of my survey participants from this part of Africa (excl. Cameroon, Madagascar & South African Coloureds) are indeed showing up as above 80% Central & Southeast African (combining “Congolese” with “Southern East Africa”, “Hunter-Gatherer” & “Broadly C. & SEA” scores, see this table)

“Southern East African” centered in Swahili countries 

Map 5 (click to enlarge)

***

Good prediction accuracy for Kenyans and to a lesser degree Tanzanians, but otherwise quite variable. The labeling of this region may be referring to the southeast of Africa. But the real focus is on DNA from the Swahili countries: Tanzania, Kenya, Ruanda and Uganda. The reference populations used by 23andme are also from these countries (see this link). The Swahili countries are actually located more so in the central part of East Africa. A bit confusing but 23andme probably just wanted to make a distinction with their Northeast African regions, based on the socalled Horn countries.

For Afro-Diasporans it is worthwhile to know that Malagasy DNA is also described by this region. But not convincingly so (23.4% on average sofar and 20.3% for “Congolese”). The same goes for South African Coloureds, among whom I found a group average of 40%.  Among my survey participants from Zimbabwe and Zambia “Southern East African” scores are mostly in between 10-20%. But it can get as low as 4% for one of my Mozambican samples! Not an equivalent of the former “Southeast Bantu” region on Ancestry therefore. Perhaps more similar in scope to Ancestry’s new “Eastern African” region. Except it does not really go northwards beyond Kenya and Uganda.

***

“Sudanese” also proxy of Nilotic DNA further south? 

Map 6 (click to enlarge)

***

Very predictive of Sudanese origins. Going by the results of my survey participants from both South Sudan & Sudan around 90%! Impressive also the near lack of overlap with neighbouring regions of “Ethiopian & Eritrean” and “Somali”. Still descriptive as well of possibly Nilotic(-like) lineage among a selected group of people in Uganda, Kenya and Ruanda.

Not shown in the map but small amounts of “Sudanese” (<5%) have also been appearing in other parts of Africa. In particular for northern Nigerians. Given Nigeria’s geographical location as well as ancient migrations (Chadic) quite understandable. But still good to be aware that these genetic components have been native to Nigeria for centuries or even millennia already. In particular for Afro-Diasporans wondering about a minuscule score (<1%) of “Sudanese” for example. If truly genuine, it will most likely have been inherited by way of a Sahelian West African ancestor!

***

“Ethiopian & Eritrean” also proxy of Cushitic DNA further south? 

Map 7 (click to enlarge)

***

Very predictive of Ethiopian & Eritrean origins. Going by the results of my survey participants from both these countries around 96%! Impressive also the near lack of overlap with the neighbouring region of “Somali”.  But still additionally appearing in minor amounts among my Sudanese results (around 5%) and for my Kenyan survey group (max. score so far 16%). Even more pronounced “Ethiopian & Eritrean” scores were reported for my Ugandan and Ruandan samples (max. scores around 30%). Obviously the country name labeling is not to be taken literally in these cases! Interesting that sofar this region seems to be appearing as a primary indicator of South Cushitic lineage and not “Somali”. Possibly to do with the more diverse composition of the Ethiopian and Eritrean customer samples as compared with the Somali ones?

***

“Somali” most exclusive ethnic region? 

Map 8 (click to enlarge)

***

Very predictive of ethnic Somali origins. Going by the results of my Somali survey participants around 97%! Impressive also the near lack of overlap with the neighbouring regions of “Ethiopian & Eritrean” and “Sudanese”.  Unlike these two latter regions I have not yet found any noticeable reporting (>2%) of “Somali” for my survey participants from other East African countries. Safe for one person of possibly Bajuni descent, an ethnic minority from southern Somalia. His rather subdued score of only 23.1% “Somali” seemingly confirming the quite exclusive character of this region. Probably caused by 23andme’s selection of ethnic Somali customer samples as Reference Populations with a high degree of genetic homogeneity. Should be interesting to see how 23andme handles describing the origins of mixed Somali persons with known lineage. For example a half Somali, half Eritrean person or a Somali with 1 known great-grandparent from Yemen.

***

 Implications for Afro-Diasporans

Overall speaking 23andme did a decent job with its newly updated African breakdown in my opinion. Perhaps not a perfect analogy 😉 But imagine your African breakdown (scaled to 100%) being worth $100. Each regional score of 1% being $1, a dime representing a mere 0.1% of your African DNA. You will get more added value from focusing on the bills rather than the coins, no matter how shiny they may appear! Keep your eye on the main prize and don’t get distracted by the small change! In my upcoming surveys for various Afro-Diasporan groups I will once again focus on group averages and the primary regional scores. As these can be considered to have the highest reliability at this stage and might also be confirmed independently by historical sources. Of course this is not to deny the potential value to be gained from smaller regional scores. However you do run a greater risk then of only obtaining virtual bitcoins 😉

***

As always customers should really invest more time in informing themselves properly about this update on 23andme. In order to avoid being left confused or mislead by their new results. Sadly I do not believe that 23andme currently puts in enough effort at providing helpful sections/pages offering guidance and context. Which is why I have conducted this survey and also posted the regional maps above. In the past I have attempted to extract general lessons to be learnt from the AncestryDNA results being reported for Africans (see this page for an overview). Wishing to improve correct interpretation of AncestryDNA’s African breakdown for Afro-Diasporans as well as Africans themselves. I intend to do the same based on this newly updated African breakdown on 23andme.

In this blogpost the following has already been established, based on my survey findings for 173 Africans from 31 countries:

• Regional admixture DOES matter! Given correct interpretation it can be very useful indeed in many cases. The 12 new African regions on 23andme are reasonably predictive for Africans themselves. Many times even impressively so! With group averages for expected main regions surpassing the 90% level even! At times the results do require some additional reasoning. Especially for countries which have not been included within 23andme’s Reference Populations. Given a future expansion of 23andme’s African sample database this should eventually be remedied though (hopefully not taking another five years…)

• Labeling of ancestral categories should not be taken too literally. Rather regard them as proxies. My survey findings and regional maps may be used as guidance for greater understanding. This circumstance of blurry & mislabeled regional categories is not restricted to 23andme btw. But seems to be inherent to admixture analysis (at this stage). It will be self-defeating to allow such imperfections to deprive yourself of valuable insight! The way I see it the glass is certainly more than half full right now 😉

• Minimal trace regions are often least informative (when <5%) or even liable to be mere “noise” (when<1%). During this African survey I came across many seemingly outlandish regional scores.⁶ For example minor “Sudanese”  and “Southern East African” scores being reported for northern Nigerians or minor West African scores for South Africans. Given either geography or historical plausibility they did not make much sense. However being aware of genetic overlap, imperfect algorithms and known bugs I have concluded that usually these trivial scores can safely be disregarded. Unless you have additional clues and corroborating evidence it will usually not be worth your time to investigate any further. Which is not say that small regional amounts are per definition without informational value! However time spent in backing up your primary regional scores (DNA matches, historical context, genealogy etc.)  might often deliver greater and more reliable results.

Additional survey findings for Afro-Diasporans are needed however to see if 23andme’s new African breakdown has truly passed the test. Although based only on my African survey findings 23andme certainly did a good job already! For a follow-up evaluation such Afro-Diasporan survey findings need to then also be contrasted with historical plausibility. In particular slave trade patterns which can be verified from the invaluable Slave Voyages Database (which incidentally also had an update very recently!).

For Cape Verdeans, admittedly a special case of Afro-Diasporans, I am already quite confident that the new “Senegambian & Guinean” region is a HUGE improvement! I find it very reassuring to see how this region is not being reported above trace amounts for any of my African survey participants beyond West Africa. Unless there is a historically plausible reason for it (Fula migrations). I have already seen several other hopeful results from across the Afro-Diaspora. For Brazilians for example. Featuring a predominant “Congolese” score, as expected given their greater Central African lineage, on average.  Also many Hispanic results featuring a primary “Senegambian & Guinean” score. In line with the Upper Guinean founding effect I already uncovered during my AncestryDNA survey. I am curious to see how much 23andme’s new African breakdown will prove to be in alignment for Jamaicans and Haitians as well.

For African Americans it seems that sofar “Nigerian” will very frequently (but not always!) turn up as primary region. More consistently so than during my AncestryDNA survey for African Americans (n=350) in 2015. But back then I already speculated that the true proportion of Nigerian lineage among African Americans was underestimated on AncestryDNA, due to regional overlap with “Benin/Togo” and “Cameroon/Congo” (see section 4). Ironically it appears a reversed situation may now be at hand on 23andme.  This is something I aim to investigate further.  Also in light of some questionable statements made by 23andme, claiming that “As much as two thirds of African-Americans’ Sub-Saharan DNA may trace back to Nigerian ancestors” and ” Today, around 20% of African American ancestry is from this regio [“Senegambian & Guinean”]

At this moment I am naturally not able yet to determine the accuracy of  the current “Nigerian” scores on 23andme for African Americans and other Afro-Diasporans. Perhaps an over-correction has indeed taken place. Given the potential overlap with DNA from (eastern) Ghana, Togo and Benin to the west and Cameroon to the east. With quite likely “Congolese” also being underestimated to some degree. This will all be context-dependent though. To be judged on a case-by case basis. Also involving additional clues such as African DNA matches, local historical context etc..

Regardless of how this may turn out to be, based on my African survey findings one thing may already be fairly certain. Given a predominant “Nigerian” score (let’s say >50% of your scaled African breakdown) the odds of such a big amount being mistaken for “Senegambian & Guinean” or any of the non-West African regions are negligible. In such cases it may already be ruled out that a majority of your African DNA hails from either Upper Guinea or Central, Southern or East Africa. Instead the likelihood of a majority Lower Guinean background is very big. Even when there might indeed be some unknown overlap with “Ghanaian, Liberian & Sierra Leonean” and “Congolese” a genuinely significant ancestral connection with Nigeria will usually be very plausible.

***

A few examples to illustrate the informational value of regional admixture

According to many pundits only continental admixture is to be taken seriously in DNA testing. Sub-continental, a.k.a. ethnicity estimates, a.k.a. regional admixture only being fit for entertainment purposes… I myself have never taken this stance, preferring to judge each case on its own merits. Attempting to maximize informational value despite imperfections and avoiding source snobbery. Which is why I have conducted my AncestryDNA surveys among Africans and Afro-descendants in the past. Applying an additional macro-regional framework for extra insight (Upper Guinea, Lower Guinea, Central/Southeast Africa). Based on my promising African survey findings described above I will perform something similar now also for 23andme.

Below are some examples which I find particularly illustrative of the informational value to be gained from regional admixture analysis on 23andme. I will also feature screenshots of survey participants of recently mixed background.⁷ Because I actually find that especially in these mixed cases 23andme’s update really shows it added value. And such outcomes may also (cautiously) be seen as encouraging for Afro-Diasporans. Although naturally their multi-generationally mixed African origins are much more complex and probably also more tricky to disentangle in DNA testing. Obviously it is a mere selection and I am not implying that such accuracy is obtained in all cases! I have tried to outline several limitations of 23andme’s analysis already in this blog post and I will continue to do so in future blog posts.

For more screenshots of the individual results as well as more detailed discussion see:

• 23andme results from West Africa (part 1, Senegal to Ivory Coast)
• 23andme results from West Africa (part 2, Ghana to Nigeria)
• 23andme results from Central & Southern Africa
• 23andme results from East Africa

***

CAPE VERDE (Santiago)

***

These are the updated results of a DNA cousin of mine. My 25 Cape Verdean samples are over 70% “Senegambian & Guinean”, on average. As usual I have applied a scaling formula in my survey whereby the African breakdown of mixed persons counts as 100%. This outcome is once more confirming the overwhelmingly Upper Guinean roots for Cape Verdeans! I intend to expand my Cape Verdean survey group to 100 samples at which time I will publish my findings in greater detail. Follow this link to see my spreadsheet with the individual Cape Verdean 23andme results. As well as this link to my previous analysis based on 100 Ancestry results (group average of 62% “Senegal” & 13% “Mali”). Also take note of how “Cabo Verde” is accurately pinpointed as ancestral location. However a far more fitting place would be below “Senegambian & Guinean” rather than under “Broadly West African”!

***

GUINÉ BISSAU (1/2 Russian)

***

Very impressive result for this person with one parent from Guinea Bissau and one parent from Russia. Perfect illustration of how regional admixture CAN be very useful both for the African and European breakdown. Notice also the correct ancestral location of Russia being assigned! Hopefully with greater expansion of 23andme’s database eventually also Guiné Bissau will be specified for this person’s African side.

***

SUDAN (Fulani a.k.a. Fellatah)

***

Very special results. Awesome testimony of how 23andme’s new African breakdown is indeed able to disentangle the intricate regional origins of this Sudanese Fulani person, a.k.a. Fellata. The new “Senegambian & Guinean” region is clearly functioning as a marker of Fula lineage for him as well as Nigerian Hausa-Fulani. An outcome which was formerly also obtained on Ancestry (see this spreadsheet). Confirmation of eastbound Fula migrations from Guinea across the Sahel into northern Nigeria and then onwards to Sudan! Along the way inter-ethnic unions did occur as reflected by the substantial “Sudanese” as well minor “Nigerian” scores. Even more amazing how this is further corroborated by the appearance of “Guinea” and “Nigeria” as ancestral locations! For more details on the distinctive genetics of Sudanese Fula see also this fascinating study.

***

1/2 NIGERIA (Igbo) & 1/2  LIBERIAN

***

Extraordinary how well this person’s mixed African background is described by 23andme! As can be seen from his own profile details this person is half Nigerian (Igbo going by his surname) and half Liberian. And amazingly this is also pretty much confirmed by his regional scores which are >45% for both “Nigerian” and “Ghanaian, Liberian & Siera Leonean”. The minor remaining scores may be disregarded as irrelevant. This outcome for a mixed person from different parts of West Africa can be seen as encouraging also for Afro-Diasporans. Although naturally their multi-generationally mixed African origins are much more complex and probably also more tricky to disentangle in DNA testing.

***

1/2 NIGERIAN (Yoruba & 1/2 Ashkenazi)

***

Very impressive result for this person with one Yoruba parent from Nigeria and one Ashkenazi Jewish parent. Perfect illustration of how regional admixture CAN be very useful both for the African and European breakdown. Imagine this person had been adopted and was not aware of his origins. Surely such results would help him tremendously in his ancestral quest!

***

ANGOLA (Mestiço)

***

The “Congolese” region also describes Angolan DNA. As clearly demonstrated by this result for an Angolan of mixed background (=Mestiço). Taking his African breakdown as 100% his scaled “Congolese” score is nearly 80%! (51/64.4=79.2%). Take note also btw how 23andme correctly assigns “Portugal” as ancestral location. If this person also had been of partial Cape Verdean descent most likely “Cabo Verde” would have appeared as ancestral location as well. Plus “Senegambian & Guinean” would have been much more elevated. I actually have several Angolan DNA cousins who have partial Cape Verdean background on Ancestry. And they likewise receive considerable “Senegal” amounts as well as the “Portuguese Islander” migration, as indication of their partial Cape Verdean lineage.

***

 1/2 KENYAN? (& 1/2 English?)

***

Very impressive result for this person who has been adopted and therefore has no certainty about her birth parents. A few years ago she was contemplating that she might have an African American or West Indian parent! However this breakdown makes that practically impossible. And instead either a Kenyan or Tanzanian parent seems very likely (based also on her DNA matches!). Taking her African breakdown as 100%, her scaled score for “South Eastern Africa” is 93% (46.4/4), the second-highest such score in my survey! Perfect illustration of how regional admixture CAN be very useful and certainly is not only fit for entertainment…

***

1/2 ETHIOPIAN & 1/2 AFRICAN AMERICAN

***

Fascinating composition for this person with one parent from Ethiopia and one African American parent. The “Ethiopian & Eritrean” score is practically 50%, as expected. Illustrating how 23andme is able to accurately distinguish between East African and Trans-Atlantic Afro-Diasporan origins. The latter being more varied of course. Indicated in this screenshot by several West African regions and also an additional minor European component.

***

___________________________________________________________________________

Notes

1) Sometimes it almost seems that admixture analysis is being considered mere guessology by its fiercest critics. Or only fit for cocktail parties as the saying goes 😉 This has however not been my experience. I do agree that performance among the various DNA testing companies and third party websites is very variable. And obviously even more so going back in time. I have myself only tested with 23andme and Ancestry and I found that with correct interpretation and knowledge of their methodology you can indeed derive useful information from admixture/ethnicity results. Which were of course not just pulled out of a hat.

I strongly believe that when Tracing African Roots most people do not have the luxury to be snobbish about admixture analysis. Instead they will want to maximize informational value from any promising source available, despite shortcomings. Combining with other research findings (DNA matches, genealogy, relevant historical context, other types of DNA testing, etc.). in order to achieve complementarity rather than putting all your eggs in just one basket.

I know of many people who made important discoveries about their genetic ancestry by using their admixture results a.k.a. ethnicity estimates. For example I have heard several stories by West Indians who had unexpected Asian admixture, minor but still substantial (>10%). And this information was really useful to them as it lead them to previously unknown Asian contract labourer ancestors. There are plenty of other ancestral scenarios for Afro-Diasporans which can be illuminated by way of the continental breakdown which is usually quite accurate.

The regional or subcontinental percentages are indeed not to be taken all too literally. But again I know several persons who did rely on distinctive regional scores to make a breakthrough in their ancestral quest. For example I have been told about at least three instances of NPE being confirmed whereby the father turned out to be East African instead of African American or West Indian. In one case indicated by the very predictive “East African” category on 23andme (pre-update) but also by a singular combination of “Southeastern Bantu” and “Middle Eastern” regional scores on AncestryDNA. Many times I have also seen how unexpected partial Cape Verdean lineage could quite reliably be corroborated by  “Senegal” scores on AncestryDNA. Not only for African Americans, but also for Hawaiians (due to whaling connections, see upcoming blog post)!

Such cases are bound to increase now that 23andme’s regional granularity has been significantly improved. All the more reason to resist being overly dismissive about admixture analysis, as this may deprive you of valuable insights! Although naturally this does not imply you should stop informing your self about any inherent limitations or imperfections. Reviewing the results of native Africans provides a good independent measure in my opinion to evaluate the usefulness of 23andme’s newly updated African breakdown. For more discussion see also:

• Did Ancestry kill their African breakdown? (part 3)

2) In fact my strong disappointment about 23andme’s lack of progress and context given in regards to African DNA test results is one of the main reasons which prompted me to start with this blog in December 2014! For example see my discussion of 23andme’s former version of its Ancestry Composition which I originally published in February 2015:

• Ancestry Composition (2013-2018)

For a more recent take on 23andme’s track record see also the section “Appeal for true commitment“, in this blog post:

• Suggestions for improving the African breakdown on AncestryDNA

3) In my AncestryDNA survey for Africans & Afro-Diasporans I have also made good use of a macro-regional format. Which is still sub-continental. Basically combining interrelated & neighbouring genetic regions in order to allow certain regional patterns to show up more clearly. This turned out to be particularly helpful when wanting to explore any rough correlation between aggregated AncestryDNA results and slave trade patterns. Which I indeed established for the most part. But also for European DNA I found that this approach works quite well. For example by making a distinction between Northwest European DNA ((“Great Britain”, “Ireland”, “Europe West” and “Scandinavia”) versus Southwest European DNA (“Iberian Peninsula” and “Europe South”) and East European DNA (“Europe East”, “Finland/Northwest Russia”, “European Jewish”). For more details see:

• Afro-Diaspora AncestryDNA results: A Comparison 
• Update: Afro-Diasporan AncestryDNA Survey (part 1)
• Update: Afro-Diasporan AncestryDNA Survey (part 2)

4) Almost all African 23andme results included in my survey have been shared with me by the DNA testers themselves. Some results were also kindly shared with me by friends. And a few results were collected by me from social media. Naturally I verified the background of each sample to the best of my capabilities but I did not have absolute certainty in all cases. I like to thank all my African survey participants for having tested on 23andme and sharing their results with me so that it may benefit other people as well!

I have been gathering African testresults on 23andme for many years already. Originally to gain a greater understanding of the African categories included in 23andme’s old version of Ancestry Composition when it was being updated in 2012/2013. Thanks to the kind willingness of people to share their results I was able then to compile some sketchy “population averages” in 2013 which I shared on 23andme’s online community at that time. And which can still be seen in this online spreadsheet. The individual results can be seen by clicking on the tabs on the bottom of the sheet. The screenshots of their results have also been featured on these blog pages:

• West African Ancestry Composition Results (2013-2018)
• Central & Southern African Ancestry Composition Results (2013-2018)
• East African Ancestry Composition Results (2013-2018)

Actually my very first survey efforts date back even earlier to 2011 :-). Based on the pioneering African Ancestry Project by Razib Khan.  I shared these findings also on 23andme’s online community at that time. They can still be seen in this online spreadsheet.

In 2018 I blogged about the former Country of Ancestry results being reported for Africans on 23andme (in 2015):

• DNA matches reported by 23andme for 75 Africans

5) Quite ironically it seems that after this update 23andme is no longer able to properly distinguish “Hunter-Gatherer” DNA. The scores for this region have sharply decreased for my survey participants. Especially for Central Africans hardly any Pygmy lineage seems to be picked up anymore. Which might not really be a big loss. Given that in most cases such connections date back to ancient times anyways. And therefore not really relevant for a genealogical time frame (let’s say 500 years).

However the severe underestimation of Khoi-San lineage for South Africans is a more serious flaw. As this type of ancestry is usually to be traced back to more recent times.  It is known from several DNA studies as well as my own previous survey findings that this is a very important and often even principal ancestral component (within the African breakdown) for South African Coloureds especially. However for some reason this is no longer showing up. Similar I guess to how Pygmy ancestry is now mostly submerged for Central Africans and also how North African(-like) DNA has almost disappeared for Fula people as well as Portuguese and Spaniards! See also:

• 23andme results for South African Coloureds before the update
• Hunter-Gatherer scores for South African Coloureds on Ancestry (before update).

6) These unexpected or even outlandish regional scores can generally be divided in very small trace amounts (<1%) and more noticeable amounts of around let’s say <5%. Both within the African breakdown as well as non-African regional scores. They can be verified from within my spreadsheet. Most of my survey participants did not have a fully 100% African breakdown for example. Due to a known bug (already before this update) causing minuscule amounts (0.1%) of non-African admixture. At times the more noticeable amounts may still make more or less sense after some reasoning. But this will require deeper knowledge of population genetics as well as ancient migrations. The average layman will tend to be just confused by these scores though or mesmerized in case they are exotic ancestry seekers 😉 Hence my analogy about resisting the temptation to be distracted by shiny coins! In my opinion it is best to just regard such scores as unavoidable imperfections and instead focus on the main regional breakdown. Unless you come across a clear deviation from the group averages I have calculated. Minor non-African scores being more worthy of follow-up research in my experience. Even when smaller than 1% in some cases.

7) My survey features several people of mixed-race background, aside from Cape Verdeans. Usually 1 parent being from a specific African country and 1 parent being from Europe. But also Africans with 1 European great-grandparent etc.. In order to make the African composition results inter-comparable between all my survey participants I have scaled the African part of 23andme’s breakdown to 100% for people of mixed background. Basically I applied the following formula:

• Scaled amount = % for a given African region divided by % of total African amount

The scaling formula I used is very simple therefore and can be verified from within the spreadsheet by clicking on any cell featuring a regional score and then viewing the calculation in the function bar (fx) in the upper left corner. All other Excel formulas I used throughout the sheet can also be verified in this same way.

Advertisement