Reconstructing genetic ripples in time and space

By Razib Khan | July 31, 2013 5:04 am

The inimitable Joe Pickrell has dropped his Khoisan-are-part-Italian preprint onto arXiv, Ancient west Eurasian ancestry in southern and eastern Africa. I’m being glib in my characterization of the paper’s core conclusion, but there’s a reason for such a flip response: the inferences that he seems to draw from the genetic data strike me as verging on crazy. But that’s OK, what genetics is telling us is that history was a whole lot crazier than we had imagined.

Let’s back up for a moment here. For several decades now geneticists have assumed that the Bushmen of the Kalahari, the Khoisan-qua-Khoisan, Africa’s last hunter-gatherers who retain their ancestral language along with the Hadza, are the ur-humans. The basal lineage that first diverged from the rest of mankind at the cusp of the Out of Africa event. This is evident in Y chromosomal and mtDNA phylogenies, where the Bushmen and their kin harbor variants which coalesce deeply in time with those of others. And, a few years ago another group revealed the likelihood that Bushmen also are products of an admixture event in the last ~50,000 years with a distinct hominin lineage which diverged ~1 million years before the present from the main line which led up to anatomically modern humanity. Now Pickrell et al. present us with a twist which is perhaps even more astringent than a lime: in their genomes the Bushmen and their Khoisan kin, the Khoe herders, reflect an ancient admixture event with East Africans, who themselves were the outcomes of hybridizations between West Eurasians and indigenous African populations. More relevantly for my concise summation of the conclusion, the West Eurasian component does not necessarily reflect modern Middle Eastern populations, so much as Southern Europeans!

How did they infer such bizarre results? Magic? No. Basically the authors looked at patterns of linkage disequilibrium. Got it? Probably not. If you are curious, confused, and intent upon understanding the thrust of their methods in your bones, you probably need to read Loh et al. Barring that trust in the great hive-mind that is the Reich lab, or attempt to swallow my trite condensation.

If you consider a short to medium length sequence of the genome, there are genetic variants, alleles, segregating across that sequence. The frequency of these alleles vary across populations. And, there are on occasion correlations of allelic combinations, seen together across a single sequence than would be likely if the alleles across the loci assorted at random. A concrete example would be a population which is the product of a recent admixture event between Africans and Europeans. Recombination would take many generations to break apart all the associations between alleles which are diagnostic and distinctive of African and European ancestry, so long blocks of ancestry tracts could be inferred simply by phasing the genome on the individual level (i.e., you know the sequence of each homolog inherited from each parent, instead of just genotype values). There would be linkage disequilibrium within the population because particular variants would be associated with others across loci due to recent distinct ancestry at the genomic level. If you noticed that SNP 1 had an African allele, then SNP 2 located nearby in the locus is also more likely to have an African allele than expectation, until the point that linkage equilibrium is attained.

As I noted above, these associations are broken apart over time in a regular fashion by genetic recombination. Therefore, the decay in linkage disequilibrium across the genome can allow you to infer time since a putative admixture event. This works at various time depths. African Americans have long range LD because the admixture was relatively recent. To date older admixture events one must be more cunning, as the LD decays and becomes exceedingly faint as recombination hacks apart previous distinctive associations as two genetic backgrounds merge. But what about multiple admixture events and the consequent linkage disequilibrium patterns? What the authors did in the above paper was to test the fit of the data to a composite of LD curves in scenarios where it seems likely that there were two possible admixture events. And, they found multiple populations which did fit this model.

Dispensing with the technicalities, here are the results of admixture events as inferred from the LD decay curves:

The most parsimonious model that Pickrell et al. propose is simple as it is crazy.

1) An ancient initial admixture event in the environs of the Horn of Africa between a proto-West Eurasian population and a proto-Sudanic population

2) A second admixture event which occurs when a population derived downstream from event 1 encounters the ancestors of the Khoisan

Pickrell et al. infer a ~3,000 year old admixture event between West Eurasians and Africans for the Semitic populations of the Ethiopian plateau in keeping with Pagani et al.’s only marginally less crazy results. Then you have step 2, with an admixture between proto-Bushmen/proto-Khoe and the hybrid East Africans ~1,500 years ago. Let us accept these genetic results on the face of it. What they bring home to me is the power of culture. Though vastly diminished today, groups such as the Khoe Nama managed to preserve their integrity and independence down to the period of European colonialism (only being truly decimated in Namibia in the early 20th century by the Germans). A wave of Bantu farmers overwhelmed most of southern Africa, but select groups of Khoisan managed to maintain zones of habitation where they persisted with their unique cultural traditions and perpetuated their language. Some of this surely was ecology, as the vast Karoo region is not particularly amenable to the Bantu cultural toolkit. But, I also suspect that institutional and economic (e.g. cattle culture) influences that the East Africans had upon the Khoe, and perhaps even indirectly the Bushmen, also made these populations more robust to the Bantu expansion than otherwise would have been the case.

Being a preprint on arXiv, the paper of which I speak here is free to you, and copiously explained in loving detail in the supplements in terms of method and madness. I am not particularly enthusiastic about having long discussions about how these results are crazy and can not be right. They are crazy. But I know enough about the methodology here to understand the logic, and accept that the authors are grasping at something very strange and true, even if their particular interpretation and specific results may be disputable. Let me quote the paper at this point:

The hypothesis that west Eurasian ancestry entered eastern Africa through Arabia must be reconciled with the observation that the best modern proxies for this ancestry are often found in southern Europe rather than the Middle East (Supplementary Table 4). This observation can be interpreted in the context of ancient DNA work in Europe, which has shown that, approximately 5,000 years ago, people genetically closely related to modern southern Europeans were present as far north as Scandinavia [Keller et al., 2012; Skoglund et al., 2012]. We thus find it plausible that the people living in the Middle East today are not representative of the people who were living the Middle East 3,000 years ago. Indeed, even in historical times, there have been extensive population movements from and to the Middle East [Davies, 1997; Kennedy, 2008].

Think on that. If Pickrell et. al. are right do you think that the Middle East is particularly special in this regard? I will say that it comes to mind that the high consanguinity may result in strange outcomes if one is not careful with the sampling strategy (I’m thinking of the Samaritans I see in their data), though I doubt that this is an incautious group. But I do think it is plausible that some European populations are better proxies for the ancient Levantines than the modern Levantines because the latter have been washed over by multiple demographic waves (though I want to see more comparisons with Christian Arab* samples).

A second bombshell dropped by Pickrell et. al.:

We note that we have interpreted admixture signals in terms of large-scale movements of people. An alternative frame for interpreting these results might instead propose an isolation-by-distance model in which populations primarily remain in a single location but individuals choose mates from within some relatively small radius. In principle, this sort of model could introduce west Eurasian ancestry into southern Africa via a “diffusion-like” process. Two observations argue against this possibility. First, the gene ow we observe is asymmetric: while some eastern African populations have up to 50% west Eurasian ancestry, levels of sub-Saharan African ancestry in the Middle East and Europe are considerably lower than this (maximum of 15% [Moorjani et al., 2011]) and do not appear to consist of ancestry related to the Khoisan. Second, the signal of west Eurasian ancestry is present in southern Africa but absent from central Africa, despite the fact that central Africa is geographically closer to the putative source of the ancestry. These geographically-specific and asymmetric dispersal patterns are most parsimoniously explained by migration from west Eurasia into eastern Africa, and then from eastern to southern Africa.

Isolation-by-distance is alluded to implicitly when we speak of human genetic variation as clinal. And it’s not totally lacking in utility as a null model. But I think we need to add another layer of complexity upon this parsimonious elegance of human clans eternally exchanging mates in monotonous step-wise fashion. Multiple populations over the past 10,000 years (and likely earlier!) were rocked massive demographic turmoil, as foreigners from afar amalgamated themselves upon the local substrate, and abolished the old to bring forth something new. The author of this post is himself a product of such an event. The genetic story of mankind is not just one of continuous and diffuse gene flow gradually over a landscape of small-sale societies. No, this placid background condition was periodically perturbed by an explosion of translocating peoples, likely triggered by a technological or cultural revolution of some sort. The genetic impact in many cases is too great to be anything but a folk wandering.

Unlike isolation-by-distance these patterns do not flow linearly across space, but exhibit discordant lashing patterns through ecologically fertile terrain. Rather than a mist gliding across the plains, imaging a flash flood scouring a ravine. A more gentle analogy would be that these are demographic ripples, which expand outward, temporarily distorting the calm surface of isolation-by-distance dynamics, and eventually fading back into the background and becoming the new normal. But once the ripple has faded how do we know that it was once? That is a difficult thing indeed, and these results indicate the problems inherent. It may be that the echoes of the ripple that Pickrell et al. detect issue from a source which no longer exists. Are the scions of the first farmers of the ancient Levant hidden away in the valleys of Tuscany and the plains of Tanzania? A crazy proposition also, but not necessarily a false one.

Citation: arXiv:1307.8014v1 [q-bio.PE]

* I know some Christian Arabs do not want to be called Arabs.

CATEGORIZED UNDER: Human Genetics, Human Genomics

Comments (30)

  1. andrew oh-willeke

    To be clear, there is nothing terribly new about proposals of East African admixture with Khoisan, or with proposals that there were at least a couple of waves of West Eurasian admixture with East Africans (aka back migration), some of which may have been in place by the time of East African-Khoisan admixture.

    The novel parts of the proposal are the Southern European affinity of the West Eurasian component and the details of the timing.

    “I know some Christian Arabs do not want to be called Arabs.”

    I hadn’t been aware of that. Thanks for the tip.

  2. razibkhan

    if you read the paper the author makes it clear to not confuse modern distributions of populations to ancient ones. so they weren’t southern europeans as such, but the common ancestors of southern europeans and this ‘west eurasian’ element.

    • facefault

      I did say *ancestors* of modern southern Europeans. I’m curious if we can figure out what historical culture they most likely belonged to.

  3. razibkhan

    it makes more sense (after a fashion) if you read the preprint. if you don’t have time, click through to the PDF and control-f “yemenite jews.” they are in there. that was my initial though, though i know this group well enough to have assumed they’d at least check that again since that’s a public data set.

    • pgbk87

      I don’t buy it… Just like Pagani, this guy’s conclusions are flawed, and partially, unscientific. I honestly feel like some “Meditid Race” theology is going on here.

      On amateur runs like Harappa, Eurogenes, etc… Sardinians are modal for “Mediterranean” and Yemeni Jews are modal for “Southwest Asian”. Horners generally don’t have “Mediterranean”-like admixture, they (almost exclusively) have a shared “Southwest Asian” ancestry with Arabians, Berbers and Egyptians. This, IMO is “Afro-Asiatic”.

      Uniparental markers (Y-DNA J*, J1, T) (mtDNA M1, N1a, R0a, etc…) are basically screaming Red Sea/Arabian

      • razibkhan

        Just like Pagani, this guy’s conclusions are flawed, and partially, unscientific. I honestly feel like some “Meditid Race” theology is going on here.

        you undermine yourself by imputing a conspiracy. or ulterior motive. i know pickrell somewhat personally and can vouch that he doesn’t adhere to/care about/even know about the sort of things you are alluding to.

        On amateur runs like Harappa, Eurogenes, etc… Sardinians are modal for “Mediterranean” and Yemeni Jews are modal for “Southwest Asian”. Horners generally don’t have “Mediterranean”-like admixture, they (almost exclusively) have a shared “Southwest Asian” ancestry with Arabians, Berbers and Egyptians. This, IMO is “Afro-Asiatic”.

        i get the same too…ON ADMIXTURE. they do not use ADMIXTURE, but methods which more explicitly attempt to extract out the eurasian allele frequencies. the patterns you refer to make sense if that if you specify a finite set of models the program which converge upon the most plausible. but that doesn’t mean it’s the true model.

        one way in which you might align these results with the ADMIXTURE results is that the ‘yemenite/arabian’ cluster is itself a synthesis, and that that population is a hybrid. a similar admixture event in ethiopia might be producing an analogous allele frequency mix. but, if you decompose the elements of both ‘yemenite’ components you might find some sort of eurasian similar to what they find.

        • pgbk87

          Will you at some point, decompose the elements of both ‘yemenite’ components? Are you supposing some Mediterranean and/or West Asian/East African hybrid creating a ‘southwest Asian’ cluster?

          This is why I believe Meheris, Socotris and other south Arabian tribes are so necessary. ‘ANA’/West Eurasian-like and ‘ASA’/East African-shifted, native Arabian will likely arise, similar to ASI and ANI. mtDNA L6, M1 and unique N subclades come to mind. This is ALL speculation on my part…

          • razibkhan

            decompose the elements of both ‘yemenite’ components?

            i don’t have the time. i assume that in the next year or so something like this will happen. the behar et al. data set has a thinner, and somewhat different, marker set than the others from what i recall.

            Are you supposing some Mediterranean and/or West Asian/East African hybrid creating a ‘southwest Asian’ cluster?

            this is not a crazy proposition. it may not even be a sub-saharan african population. the model more explicit would be:

            (‘southern european-like’ + ‘east african like population A) = yemenite

            (‘southern european-like’ + ‘east african like population B’) = yemenite-like in ethiopian

            note that in the model above the east african populations are distinct. but the hybrids may resemble each other enough in a model with say K = 12 than they’ll be thrown together. IOW, i’m saying that the ‘south arabian’ cluster may be a reasonable artifact of the sampling and explicit model selection method,

          • pgbk87

            You know this only raises more questions than answers right? lol

            So there is possibility that a native Arabian population existed and has been long absorbed by a west-eurasian one? Could have been something intermediate to East Africans and ASI perhaps?

          • razibkhan

            You know this only raises more questions than answers right? lol

            this is not a bad thing for scientists. though perhaps not optimal that science remain in such a state in perpetuity.

            So there is possibility that a native Arabian population existed and has been long absorbed by a west-eurasian one? Could have been something intermediate to East Africans and ASI perhaps?

            if you assume that isolation-by-distance is the null model, why not? you can imagine a scenario where genetic variation was relatively uniform, and over the last 10,000 years punctuated demographic expansions from a few agriculture/technological loci across the world were overlain upon this monotonous substrate.

          • pgbk87

            Yes, this makes perfect sense.

          • Karl Zimmerman

            I’m very interested to she how the South Arabians genetically as well. There are very few pictures of either group online, but the ones I have seen (especially the latter) seem as if they could have been subject to roughly the same amount of SSA admixture as Arabs (at least on average).

            One other possibility is that South Arabia was heavily genetically modified by admixture with South Asia. It’s often been remarked that South Arabians have “Australioid” or “Veddoid” features. Maybe this is not a remnant of the South Arabian indigenous population. The Indus Valley civilization had extensive trading links (arguably going all the way to Africa, based upon relations between Sahel and Indian crops). It’s plausible there were Harappan trading posts on Arabian coast, which could have influenced the region genetically and culturally.

            That said, the timing doesn’t quite mesh with the paper here. Presumably this study suggests that South Arabia was still “pure” (presuming that, and not the Nile, was the locus of admixture) in 1,000 BC, which is well after the decline of the Indus Valley Civilization.

    • anon

      Maybe oetzi the iceman could be the key to the putative west eurasian population contributing to east and southern africa.

      • razibkhan

        not oetzi, but probably ancient DNA. luckily the middle east is dry. some regions are even cold because of elevation.

  4. Lank

    Reading the full paper, I think it makes a lot of sense and confirms previous findings for southern African populations. The West Eurasian source population most closely resembles the West Eurasian component in East Africans, followed by an inferred ancestral Middle Eastern population (see Figure 5). The West Eurasian component in the admixing East African population is estimated at ~25%, roughly similar to many Cushitic pastoralists that live south of the Ethiopian highlands.

    Southern Europeans are only the top match out of modern populations, which I think says more about migrations within West Eurasia than within Africa. Interestingly Sardinians and Basques often pop up as the source populations in the lowest f3 statistics, and both of these populations often lack a recent “West Asian” element in ADMIXTURE that North/East Africans also usually lack, but which has become widespread in modern Europeans and Middle Easterners. African admixture in modern Middle Eastern populations may also play a role.

    It was not the focus of the paper, but the problem as I see it is for those who would interpret the 3 kya admixture signal in modern East Africans as reflecting a migration directly from West Eurasia 3000 years ago, which contributed all of the West Eurasian ancestry in modern East Africans. This is pure speculation as we don’t have any probable ancestral populations (neither African nor West Eurasian) for this inferred admixture event, which we do for southern Africans, otherwise we might also have concluded that Mediterraneans sailed to the coasts of southern Africa 900-1800 years ago.

    But there’s a lot of evidence to suggest that West Eurasian migrations into Africa occurred much earlier, considering the existence of lineages that don’t look like recent inputs into Africa (e.g. mtDNA M1, U6), and the presence of Cushites (and agriculture) in East Africa earlier than 3000 years ago. The 3 kya signal could perhaps be related to the spread of some agricultural innovations, or some unknown population intermixture in the Horn of Africa, but lacking further evidence it is impossible to tell.

    • razibkhan

      the ppl in the reich lab are very smart. but…who knows if some of the parameter values they took for others in their model are exactly correct. i’m confident they are detecting a pulse event admixture. i am less confident about the timing, though i doubt it is off by an order of magnitude. how far back do you think the admixture event would be ignoring the 3,000 year old dating that from the LD decay methods?

      • Lank

        I would guess the bulk of West Eurasian ancestry in East Africa is a result of migrations during Afroasiatic and/or Neolithic expansions. The spread of Cushitic speakers since ~8000 years ago is tied to a Neolithic economy, although some would argue that Afroasiatic as a whole is pre-Neolithic and originated among the early wild grain collectors in Northeast Africa.

        It is in many ways analogous to the situation in India, where you have a recent (Indo-Aryans, corresponding to the Semites of EA) incursion, an older widespread group of Dravidians; also with West Eurasian ancestry, and possibly linked to Neolithic dispersals. I should add that it seems to me like Indo-Aryans had a somewhat wider impact, whereas the influence of Ethiosemites was limited south of Abyssinia.

        The method details in papers from the Reich lab are generally very elaborate, and they seem to lack any apparent flaws. But the case for southern African admixture is much stronger with the additional pieces of genetic (Y-DNA, lactase persistence, ADMIXTURE etc.) and archaeological evidence. In their case, we can be fairly sure about the origin of the admixing populations <2 kya. But the LD decay alone does not strike me as revealing anything concrete about East African population history, although this could change with some ancient DNA and improved dating of uniparental lineages.

        • razibkhan

          high quality comment. but i will add one thing: individuals in this lab often run simulations to test their methods (as in this paper). this is not a panacea, but they are not flying totally seat of their pants.

  5. Dmitry Pruss

    Isolation-by-distance and major population displacement aren’t really mutually exclusive, in fact they could have some synergy. If the neighbors of the Khoi-San who inhabited more pastoralist-friendly habitats got largely displaced by a wave of East African migration, then the remaining native migrants the less welcoming habitats could have received respective genetic inflows though a diffusion-like process? The timing of the diffusion process would correspond to the window of time when the pre-pastoralist populations were gradually moving away from the in-migrants and into their present desert habitats.

  6. MrJones

    “To be a bit facetious, the results imply that a male NE African/ S Arabian population sourced N and NE Mediterranean (or early Anatolian/ N Levantine) wives”

    I’m not sure if this fits the paper but just addressing your points – what if?

    The first farmers spread wherever their agricultural package was viable. Initially they’d leave relatively marginal regions alone in favor of the most viable. The foragers in the adjacent marginal regions pick up the herding part of the farming package leading their population to expand. The pastoralists then over-run the farmers.

    If this was a recurring pattern you’d see the female half of the first farmers merged with the male half of the local foragers turned pastoralists and the female half might be similar across the whole extent of the farmers original expansion.

    So in this case first farmers spreading round the coast of Arabia find a nice spot somewhere around Yemen, create an advanced polity for the region and as a side-effect of that process turn the relatively weak local foragers into relatively strong pastoralists and get conquered by them. That creates your first hybrid population.


    “and then sent off their male offspring, only, in large masses, back to the Horn of Africa.”

    If the Yemen-based polity is initially more advanced and expands across the water into Ethiopia then the male line would likely be dominant. There wouldn’t necessarily need to be large masses of them. For example isn’t the 20-25% white admixture in African-Americans mostly from sugar plantations – or the plantation system in general? I’d expect the percentage would be higher if the system had lasted longer?

    Or alternatively the conquistador example – especially if the source region is polygamous and the Yemeni conquistadors don’t have any women.


    “And later, further south, their male lineage signature vanished, as (apparently?) at this stage, their female offspring were the more successful.”

    Why would they later head south? Gold maybe? And if they were eventually over-run by the neighboring tribes then their surviving mark might be only carried on the surviving female side unless some segment survived the over-run.

    “According to Lemba oral tradition, their male ancestry originally comprised several male “white people from over the sea” who came to southeast Africa from a country which boasted large cities in order to obtain gold.[16][21][22] After becoming established in Africa, at some point, the tribe split into two groups, one
    staying in Ethiopia and the other travelling farther south, along the east coast.”

  7. andrew oh-willeke

    Is it really so crazy to think that East Africans would receive West Eurasian admixture via the historical evidence associated with the emergence of the Ethio-semitic language family from a single source language that was part of the Semitic language family of the Levant/Arabia around 3000 years ago which is roughly when the linguistic and historical evidence points to that happening?

    An earlier Upper Paleolithic back migration from West Eurasia is associated with M1/U6 mtDNA haplogroups in Africa (although the distribution suggests geographically distinct migration patterns for each of these even though they are roughly contemporaneous).

    I would think that there would also be some Neolithic era contribution and possibly one between a Neolithic wave and an Ethio-semitic wave (e.g. Y-DNA T in Somolia almost surely is another back migration from West Eurasia separate from the dispersal of MtDNA M1/U6, but it isn’t at all obvious when it arrived or if it back migrated by sea from Yemen and vicinity or by land via the Blue Nile). But, the two admixture wave estimate is a floor and not a ceiling.

  8. razibkhan

    the contradiction here is rather plain and moderately persuasive to me. perhaps the model needs more complexity…

    • Joe Pickrell

      Yes, I think I understand the point here. I was mostly focused on the east->south Africa part of the chain, but the initial source of west Eurasian ancestry in eastern Africa merits more thought.

      I think the (more recent) African admixture in the Middle East strongly influences analyses using modern populations. A quick analysis with TreeMix suggests that, after correcting for African admixture in the Middle East, the best proxy for the west Eurasian ancestry in east Africa is Bedouins. That might reconcile this a bit.

      • Onur

        But Cypriots, Armenians and Georgians were also tested in this study and they all generally showed up as worse proxies for the West Asian migrants than the tested Southern European populations did. Do they have (more recent) African admixture too? Cypriots may have such admixture, but it should be so slight that it probably has no effect on Cypriots’ proxy status. But Armenians and Georgians do not seem to show any such admixture at all.

        • Joe Pickrell

          Cypriots show up as pretty good proxies for the west Eurasian ancestry in east Africa (see Supplementary Table 4).

          But I think I’m missing something: what’s special about Cypriots, Armenians and Georgians that makes them particularly interesting candidates for the source of this ancestry?

          • razibkhan

            one thing that jumps out about these three groups is that like assyrians they were not part of the cosmopolitanism of the islamic period due to ethno-religious reasons. so presumably a better snapshop of mid-east genetics before 650 AD.

          • Onur

            Razib pretty much explained what is so special about them in his last comment. I would like to add that they are geographically closer by land to Africa than Southern Europeans are, so they are expected to be better proxies for the West Asian migrants in Africa than Southern Europeans are, which is in contrast with your inferences of the proxy population.

  9. MrJones

    If a parent population in the Yemen with limited scope for profitable expansion in Arabia spawns a child population in Ethiopia which over time grows to be much bigger and stronger than the parent e.g. Axum, such that at some later date the child population eventually conquers and maybe even displaces the original parent population would that square any circles?

    (I’m just guessing here, i don’t know if it would or not.)


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


See More


RSS Razib’s Pinboard

Edifying books

Collapse bottom bar