We are all…Sardinians?

By Razib Khan | March 18, 2012 2:54 pm

The title is tongue in cheek. But I have been noting with interest Dienekes’ trial runs with TreeMix. With it he has discovered a very peculiar admixture event, at least as determined by the software. The results are below, with my clarifying labels:

Basically the software seems to be implying that there has been gene flow into the Yoruba of West Africa from from populations related to the Basque & Sardinians. The most plausible explanation for this would be that this migration event was ancient, so that later admixtures of European populations are not reflected. Whether you believe that the Basque/Sardinians are Neolithic or pre-Neolithic (i.e., that they derive from Holocene settlers, or derive from Paleolithic European substrate), it is probably true that they have a relatively long period of residence in Western Europe.

First, do not take these results at face value. Results generated by abstract methods sometimes make no sense. Often they’re only as good as their assumptions, and the common sense of the humans utilizing the methods. As a classic case in point Allan Wilson, whose group later propounded the mitochondrial Eve theory, originally reported to some journalists in the early 1980s that modern humans may have derived from Australia! There was a reason for this, and that had to do with some problems in the phylogenetic methods they were using (or at least the way they were using it). Later on common sense, and all of archaeology, realigned with the phylogenetics when the research group went over their methods. Until lots of people use TreeMix and start to find similar results I would be very cautious about taking any given counter-intuitive result as plausible.

With all that being said, I think that these results may point to a broader truth which is only starting to shake out: the prehistory of Africa has been much more complex than we had assumed, and the ‘Out of Africa’ model has blinded us to this. One finding which seems reasonably robust is that there is a large difference between the hunter-gatherer populations of the continent, and other Sub-Saharan Africans. Depending on the method you use the most reasonable clade may even be African hunter-gatherhers vs. everyone else (including agriculturalist Africans). There has been massive genetic-demographic change in Southeast Asia and South Asia, so why would Africa be any different? In fact, we know it’s not any different, the Bantu expansion has totally reshaped Africa over the past 3,000 years. And why should we stop at that? Is it too implausible that there was back migration from Eurasia during the Neolithic Subpluvial, during which there was a “Green Sahara”? Actually, I would bet that the back migration was a little more ancient, but that is neither here nor there. If the past is as complex as I think it might be then the idea that ‘We are all Africans’ may still be true, but it may also be trivial….

MORE ABOUT: Human Evolution
  • http://dispatchesfromturtleisland.blogspot.com ohwilleke

    The Neolithic Subpluvial population is a plausible source for the Chadic peoples of the African Sahel, but a rather implausible progenitor for most Niger-Congo language speaking West Africans (e.g. the Yoruba).

    Y-DNA R1b-V88 in Chadic populations is suggestive that this ethno-linguistic group is particularly associated with back migration from Eurasia. See Cruciani (2010) and a caveat on the conclusions there by Lancaster (2010). Cerny (2009) supports a consistent timeline on the mtDNA side with the distinctively Chadic mtDNA having Cushitic East African roots at around the same time. Eurasian Y-DNA and mtDNA haplogroups are almost absent from the Yoruba and the only Niger-Congo speaking population within non-trivial percentages of these Eurasian haplogroups are the Fulbe right on the boundary of the Niger-Congo linguistic area in Northern Cameroon.

    The back migration of mtDNA haplogroups M1 and U6, commonly found in Afro-Asiatic linguistic populations, in contrast, is frequently estimated to have occurred in the early Holocene, prior to the arid Sahara era that preceded the Neolithic Subpluvial. But, traces of this back migration are rare in West African Niger-Congo speakers. The back migration associated with the Ethio-Semitic languages, in turn, was likely later than the Neolithic Subpluvial, perhaps in the Bronze Age.

    A Holocene gene flow in the opposite direction (from Africa to French Basque/Sardinia) seems quite a bit more plausible historically if there really is a link, even though this is not what TreeMix is concluding. Many Southern European populations have low percentage levels of Y-DNA haplogroup E and mtDNA haplogroup L, split between Eastern and Western distributions, and the apparent link could come from that secondary African source migration. The large number of European groups in his TreeMix model and limited number of African out groups in his TreeMix model could make the European examples of African DNA traces look more diverse than the African sample, and hence look like a source rather than a sink.

    A later post by Dienekes estimates the non-San component of various African populations in a very low K Admixture model with extreme outgroups (Neanderthal, Denisovia, and Chimpanzee). It could be that these models are simply lumping everything that is non-African hunter-gatherer as part of the same component and would decompose into Eurasian and intra-African expansion components with more detail. There has been genetic evidence suggesting a proto-Khoisan/Pygmy population from other Africans around the time of the Out of Africa event. So any genetic traits other Africans and Eurasians had in common could be younger than the divide between the San and West Africans (or West African dervived Bantus).

  • Halvorson

    For what it’s worth, from p. 123 of Coon’s Living Races of Man:

    “Meanwhile we may note that in detailed analysis of 571 modern Negro crania, made by advanced mathematical techniques, has shown that these crania gravitate between two poles, a Mediterranean Caucasoid and a Pygmy one. The former type is again divisible into an ordinary Mediterranean and a Western Asian type, which suggests more than a single northern point of origin for the Caucasoid element. ….The Negroes resemble the Caucasoids closely in a number of genetic traits that are inherited in a simple fashion. Examples of these are fingerprints, types of earwax, and the major blood groups. The Negroes also have some of the same local, predominantly African, blood types as the Pygmies.”

  • Dwight E. Howell

    It may be worth noting than Otzi matches up best with Sardinians. Since the locales don’t you can get some sort of feel for how much gene replacement has occurred in the region.

  • http://forwhattheywereweare.blogspot.com/ Maju

    It must be a bug: the weight of the arrow is 68% and that means that Yorubas are only true Africans by 1-0.68=0.32=32%.

    I can believe that Dienekes deceives himself into believing that because he also believes that Y-DNA E, the main male lineage of Africa, is a back-migrant from Eurasia. But nobody ever before found this kind of thing – and the very least in such cases is to make alternative tests with different samples or different programs before believing what is so hard to believe.

    Exceptional claims require exceptional evidence, and this is not it at all.

    It must be a bug of the software.

  • http://www.genomesunzipped.org J Pickrell

    The default in TreeMix (which I’m coming to realize was very poorly chosen) is to to treat all SNPs as independent (more or less). In genome-wide data, this is obviously not the case. The very small scale bar on the TreeMix plot in this post suggests to me that the model is being overconfident in its estimation of the parameters due to this problematic default setting.

    My suggestion (which I’ve relayed to Dienekes, but is worth mentioning on this post as well) is to run TreeMix with a different -k setting (to group together SNPs in linkage disequilibrium). In 650K data, -k 1000 should be approximately appropriate.

    I’m not sure this will affect the result, but it may.


  • http://dienekes.blogspot.com Dienekes

    It does seem that Yoruba, Mandenka, and Bantu are shifted relative to San in a West Eurasian direction:


    This is in agreement with the results of McEvoy et al. that Africans and West Eurasians share more ancestry than Africans and East Asians do:


    It is the case that most human Y-chromosomes belong to the CT group, and of the subgroups of this DE is found in Eurasia and Africa and CF in Eurasia. And, indeed, of D and E, they are both found in Eurasia, and only E is found in Africa.

    It is often said that Sub-Saharan Africans are the most ancient population because of their high effective population size/great genetic diversity. It is peculiar that _they_ (the high size population) would be so affected by drift as to lose all CT descendant nodes and end up with a twig (E), whereas the same thing did not happen in Eurasia.

    It is doubly peculiar, since this apparent “drift” works in peculiar ways, sparing the A’s and B’s and wiping only CT descendants selectively.

    There really is no case for substantial population continuity in Africa. The archaic finds in Nigeria and Central Africa are nothing like modern Negroid populations; Hofmeyr and other pre-Holocene skulls are nothing like the modern San.

    The ancestors of modern humans certainly originated in Africa (due to the total Y-chromosome and mtDNA phylogeny). However, that does not mean that Africa was the engine that always kept spewing out people but never received them.

    By 100ka we see evidence of modern humans from the Levant to Southern Arabia. There are earlier traces of modern humans (with archaic features) at the fringes of the Sahara in North and East Africa. Moreover, the root of the Y-chromosome phylogeny is estimated as being ~140ka using SNP based methods, with a root in Northwest Africa.

    So, I have hypothesized that modern humans were evolving in parts of Africa pre-100ka, especially in a “Green Sahara”, slowly achieving anatomical modernity. The de-greening of the Sahara would have pushed some of them into other parts of Africa, as well as into the Near East, were we first find them c. 100ka.

    However, the AMH who lived c. 100ka still showed no signs of behavioral modernity or the ability to outcompete other hominins. Where did they fully evolve into the highly expansive species? My guess is Arabia during its own “green phase” pre-70ka.

    As Arabia turned back to desert post-70ka Eurasians experienced their own bottleneck (which is conveniently ascribed by geneticists to a 70ka “Out-of-Africa” evident that has absolutely no archaeological evidence for it). But, it was ultimately, the UP Revolution where they apparently acquired the technological behavioral toolkit to ultimately spread to the four corners of the world, and we see them from Europe to East Asia to Australia.

    And, we do have some intriguing clues from the same era: the Hofmeyr skull, similar to Eurasians of the same times turns up in Africa. And, more recently, Hammer et al. looking at small regions of DNA found a ~36ky admixture event in Africa. And, Y-haplogroup E is dated to about 50ka as well.

    It is peculiar that those who cling to the traditional Out-of-Africa theory hypothesize that Out-of-Africans took a very short time to replace every archaic hominin in Eurasia, but supposedly took about 180,000 years to replace archaic hominins in Africa itself (from the time of Omo I to the time of Iwo Eleru). Or, those who think that Hofmeyr is ancestral to modern humans in Eurasia postulate that his kin spread all over the world and replaced all other archaic hominins, but apparently waited around for 20-30 thousand years to replace archaic hominins in Africa itself.

    The replacement of archaic hominins in Africa was probably a complex process; there is simply no evidence of a smooth march from archaicity to modernity in the anthropological record. We can discern at least three stages in it:

    1 – Emergence of AMH 200-150ka, at first limited to the Sahara and its fringes
    2.- Maturation of AMH and final “exodus” during the de-greening of the Sahara, to the Near East and the rest of Africa, but with no technological behavioral advantage over other hominins of the time
    3.- Back-migration to Africa of DE-bearers post-50ka with the new package that allowed modern humans to become such a successful expansive species: this must have brought them far and wide in Africa, although they may very well have avoided areas for which they were maladapted.
    4.- Final push during the Holocene and secondarily post-agriculture, where we see clear signals of Y-chromosome expansions and the archaic forms complete disappear from the record.

  • Charles Nydorf

    Linguists like Heinrich Wagner have long pointed out what they interpret as typological similarities between Western European and African languages. On a more personal level, my Engish wife and I were watching a documentary on a West African traditional dance. She experienced a shock of recognition at seeing very particular moves that she knows from English Morris dancing.

  • http://www.zackvision.com/weblog/ Zack

    I ran TreeMix on the Reich et al Indian Cline dataset (alongwith CEU, Yoruba and Onge). And I found a 57% migration edge from the Onge to Yoruba. [sarcasm]Obviously, there was back-migration from India to Africa.[/sarcasm]

  • http://forwhattheywereweare.blogspot.com/ Maju

    #6 Obviously the E problem is a real issue but it’s an issue only affecting Africa apparently (somehow E drifted out all other CF’DE (or pre-E) Y clades an became hegemonic in all the mtDNA L2″6 clan). This does not imply any back-migration from Asia, much less one that would cover all Africa but rather exceptional survival of pre-E clades (mostly CF but also pre-D) in the NE outskirts of early humankind, in the OoA route across the Red Sea.

    The slight greater affinity of West Eurasians and Africans is caused by geographic proximity and some admixture events, including the migration of E1b (Afroasiatic-speakers?) on North Africa and then across the Mediterranean (but also the re-admixture with OoA remnants in Persian Gulf, South Arabia, North Africa and maybe even the Levant).

    #8 “[sarcasm]Obviously, there was back-migration from India to Africa.[/sarcasm]”

    Nice. Sarcasm and counter-info appreciated, really.

    I’d even go a step further and say back-migration “from SE Asia” because even if Andaman belongs to the state of India geographically it is part of SE Asia and the nearest mainland is Burma.

  • http://dienekes.blogspot.com Dienekes

    Re: choice of K

    My analysis cited by Razib in this post was based on 166,770 SNPs.

    Following Joe Pickrell’s proposal of using -k 1000 for 650K data, I have repeated the experiment with -k 250.

    The first migration edge added was again: (Basque,Sardinian)-to-Yoruba (64%)

    Therefore, it does _not_ appear that the choice of -k is responsible for this inference of substantial back-migration.

    It would be a good idea to have some framework to determine the appropriate choice of -k.

    In particular, if there is an old admixture event between divergent populations, then recombination will have resulted in modern sequence that is a pastiche of short archaic segments interspersed with longer modern ones (see here: http://dienekes.blogspot.com/2012/01/archaic-dna-data-mining-for-dummies.html)

    By using a very large value of -k we are essentially bundling up archaic with modern sequence, making it more difficult to detect admixture stemming from such an admixture event. But, obviously this was not the case for this particular example with different -k, probably because the detected archaic admixture in Yoruba is so extensive.

    Re: Onge

    It is not entirely clear where the back-migration that brought Y-haplogroup DE chromosomes to Africa originated. The likeliest place is somewhere between East Africa and Tibet/Andamans, since that is the geographical space between the now-disjoint D and E worlds.

  • Lank

    Zack, that’s quite interesting actually, although I agree it doesn’t signify actual back-migration. In Tishkoff et al. (2009), which remains the only study with a truly diverse African dataset, the African affinity of Oceanians is specifically East African. The Onge would be the closest thing to an Oceanian group in your dataset, I presume. Could this be some sort of OoA remnant? A shared affinity between Africans and Oceanians (and partially South Asians), whose ancestors migrated along the southern route.

  • http://www.genomesunzipped.org J Pickrell


    That’s quite interesting. What happens if you include a “better” outgroup (e.g., San) in the analysis?

  • pconroy

    Charles, we have Mummers in Ireland, and the Mummers dance is similar to Morris dancing.

    Then there is the Bodhrán – which is similar to drums from Morocco to the Middle East:

    Then you have Sean-nós (literally “Old Style” ) singing – which is similar to North African and even Pakistani singing – as I pointed out previously:

    Then you have Shorthorn cattle found in South West Ireland, North Africa and down the coast to Senegal and inland to the Central Sahel.

  • Roberto Congiu

    Interesting…I guess it would have been really easy to go from southern italy to africa and vice versa during the last ice age: http://moblog.net/media/m/i/n/minushabens/15000-years-ago-italy-during-latest-ice-age.jpg

  • Lank

    One finding which seems reasonably robust is that there is a large difference between the hunter-gatherer populations of the continent, and other Sub-Saharan Africans.

    If you ask me, the most significant distinction between hunter-gatherers and other Africans is not the recent success of Y-DNA E-PN2 in the latter group. Y-DNA E is estimated to be 15-20 thousand years older than CT, which is not much.

    The mtDNA phylogeny reveals a remarkable structure. L0, diverging at the root of the human mtDNA phylogeny, is the predominant (apparently near 100% in some groups) haplogroup in the Khoisan. L1, which was next to split off, has a frequency near 100% in western Pygmy populations. It is not without reason that hunter-gatherer populations are considered the oldest human populations in the world.

    West Africans are a mixture of L1, L2 and L3(xM,N). L0 is generally not found there. The influence of back-migrating DE Y-chromosomes certainly cannot be excluded here, but addressing the mtDNA structure (which is a more reliable sign of continuity than Y-DNA, judging by aDNA) should be a priority in this discussion.

  • http://dienekes.blogspot.com Dienekes

    Re: #12

    Using the HGDP San as outgroup there is a 63% migration edge from (Basque, Sardinian) to San


    Further evidence for continent-wide Eurasian admixture in Africa that had been previously undetected due to ascertainment bias that neglected African variation:


  • J8

    It is premature to describe this evidence as widespread in non foraging groups.
    As I posted on Dieneke’s blog:

    It would be interesting to know what subgroups of Yoruba and Mandenka were used. If Eurasian admixture truly is present here, perhaps known circumstances can explain it. The Mandenka live in the Sahel-Savannah, and some samples might be affected by Maghrebian and/or Fulani-like admixture. Many Kenyan Bantu tribes are known to be admixed with Cushitic peoples etc., and parts of Northern Yoruba land were repeatedly invaded by the Fulani in the 19th century (One also might consider dealings in parts of the region with, not too far away Chadic speakers. Perhaps also, larger sample of African groups should be used, including some with less known history of such contact.
    Regarding Zulu: Nguni and many S. African Bantu subgroups are a branch of the Eastern Bantu language family, which may have coalesced near the Great lakes. Cushitic peoples inhabited the Eastern parts of this region, and some southern Bantu, as do Bantu groups in much of the East may have a small Cushitic admixture.

  • pconroy

    @ Zack said:
    I ran TreeMix on the Reich et al Indian Cline dataset (alongwith CEU, Yoruba and Onge). And I found a 57% migration edge from the Onge to Yoruba. [sarcasm]Obviously, there was back-migration from India to Africa.[/sarcasm]

    All joking aside, I would love to see you actually attempt this TreeMix run if you haven’t already done so. Why, because sub-Saharan African cattle are by-and-large African maternally and Zebu paternally, so some group introduced Zebu cattle to Africa.

    I speculated years ago that an Elamite-like group from Southern Iran or Pakistan might have been responsible for this.

    I’ve also speculated that Austronesian ancestry will be found in Southern India, and possibly Yemen or the Gulf States and almost certainly East Africa.


Discover's Newsletter

Sign up to get the latest science news delivered weekly right to your inbox!

Gene Expression

This blog is about evolution, genetics, genomics and their interstices. Please beware that comments are aggressively moderated. Uncivil or churlish comments will likely get you banned immediately, so make any contribution count!

About Razib Khan

I have degrees in biology and biochemistry, a passion for genetics, history, and philosophy, and shrimp is my favorite food. In relation to nationality I'm a American Northwesterner, in politics I'm a reactionary, and as for religion I have none (I'm an atheist). If you want to know more, see the links at http://www.razib.com


See More


RSS Razib’s Pinboard

Edifying books

Collapse bottom bar

Login to your Account

E-mail address:
Remember me
Forgot your password?
No problem. Click here to have it e-mailed to you.

Not Registered Yet?

Register now for FREE. Registration only takes a few minutes to complete. Register now »