Gene Expression

Uyghurs are hybrids

i-170fcf8504355f8a267e1657a55b6f5a-uyg1.jpgAnother paper is out which falls under the category of using genetics to understand human history; Analysis of Genomic Admixture in Uyghur and Its Implication in Mapping Strategy:

The Uyghur (UIG) population, settled in Xinjiang, China, is a population presenting a typical admixture of Eastern and Western anthropometric traits. We dissected its genomic structure at population level, individual level, and chromosome level by using 20,177 SNPs spanning nearly the entire chromosome 21. Our results showed that UIG was formed by two-way admixture, with 60% European ancestry and 40% East Asian ancestry…Both the magnitude of LD and fragmentary ancestral chromosome segments indicated a long history of Uyghur. Under the assumption of a hybrid isolation (HI) model, we estimated that the admixture event of UIG occurred about 126 [107∼146] generations ago, or 2520 [2140∼2920] years ago assuming 20 years per generation. In spite of the long history and short LD of Uyghur compared with recent admixture populations such as the African-American population, we suggest that mapping by admixture LD (MALD) is still applicable in the Uyghur population but ∼10-fold AIMs are necessary for a whole-genome scan.

The rationale for these sorts of admixture studies is medical; you need to know the genetic background of a population to make sure that you don’t infer spurious assocations due to population substructure. Imagine that you have a study group and allele B on gene 1 is strongly associated with a higher incidence of disease X. This is important information which is going to be relevant for targeting and treatment. But what if you find out later that your study group consisted of two historically distinct populations, and allele B is only found in one of these. Additionally, within this population allele B is not associated with disease X. One can infer then that the original finding was simply due to the population substructure, and the correlation between allele B and disease X was due to the higher frequency of both these characters within a population as opposed to any causal relationship. Sometimes the substructure is going to be cryptic, and in other cases researchers need to be clear about the level of granularity necessary for any particular question.

i-a566b3803afa7ee2ede4797b3e152ca4-blondug.jpgThat’s all great; I love medicine as much as the next person. But what does this tell us abut human history? First, note that the samples in this study were collected from Khotan, an ancient city along the southern edge of the Tarim Basin in Xinjiang. The authors of this study note that it was imperative to obtain samples from these southern Uyghur populations because they are likely to exhibit the least recent admixture from Han Chinese populations. Personally, I suspect that these recent exogenous genetic signatures in northern Xinjiang populations are likely to be from intermarriage with the Hui, ethnic Chinese speaking Muslims who are called Dungans in Central Asia. In any case, before Khotan was Uyghur, it was Tocharian, and this region of Central Asia produced the famous Tarim mummies. As you likely know the mummies were described as being surprisingly European in appearance. Why surprising? Well, it is Asia, and in Asia people look “Asian,” which generally excludes individuals with blonde hair and blue eyes.

I think it is important to move beyond the constraint and shock of labels. Terms like “Europe” and “Asia” are rather ill-defined, and their boundaries tend to shift with the geopolitical winds. For example, Peter the Great of Russia famously lobbied cartographers to place the official boundary of Europe at the Urals. During the previous centuries propagandists in the service of Poland-Lithuania depicted Russia as outpost of Asiatic Tartary. But long before the modern era people that looked “European” (read: white) have been resident outside of the geographical continent of Europe, taking advantage of steppe highways of Inner Asia (in other words, all people who lived in Europe might have looked European, but all people who looked European did not live in Europe).

Or have they always? I have a suspicion that the rise of agriculture and nomadism changed things when it came to long distance travel, making it relatively more common than it had been in the period before the past 10,000 years. I have posted before why I think the Uyghurs’ genetic profile suggests a very particular shape to the past which excludes some scenarios. In this paper the authors indicate that the genesis of the Uyghurs lay within the past few thousand years. The relatively low within population variance in ancestral components and the extent of linkage disequilibrium places a limit in how recently the hybridization event occurred. Recombination has had enough time to chop up ancestrally specific genomic regions while many generations of random mating have distributed the disparate ancestries across all lineages. But what about the upper boundary of the origin of the Uyghurs? The authors state:

…However, this result could be underestimated due to the assumption of a hybrid isolation (HI) model. In this model, we assumed that Uyghur was formed by a single event of admixture during a short period of time, which might not be true of the real history of the Uyghur. Considering the geographical location where the Uyghur settled, continuous gene flow from populations of European and Asian descent was very likely….

On a priori grounds I too consider continuous gene flow likely. But I’m not sure so at this point, and the reason is one specific gene, SLC24A5, which is the second in rank among ancestry-informative markers. This is because the variant of SLC24A5 found in Europeans is rarely found in people of African or East Asian ancestry. The European variant is also the derived one; that is, relatively recently in evolutionary history a mutant polymorphism was subject to a powerful selective event which swept through European (and West and South Asian) populations. Recent work implies that the sweep might have occurred as recent as 6,000 years before the present (or as early as 20,000 years ago).

i-27e75596d87f388b515bbd4b0dafd923-nonblondeug.jpgWhat does that have to do with Uyghurs? If Uyghurs, or their ancestors, were mediators for gene flow from east to west and vice versa for thousands of years why is it that East and West Eurasians have different genetic profiles for the loci recently selected for skin color? Strongly selected alleles should sweep across populations assuming even minimal levels of gene flow. SLC24A5‘s derived variant has swept as far as Sri Lanka. The fact that it isn’t extant at appreciable frequencies in East Asia suggests to me that Uyghurs did not serve as a mediator population between the east and west. Rather, Uyghurs may very well be one of those populations where the hybrid-isolation model is actually a rather good approximation of reality.

Here is what I believe might have occurred. Prior to the emergence of nomadism and various forms of irrigation agriculture Xinjiang was extremley lightly populated. So thin were humans on the ground that Central Asia served as a gene flow barrier between east and west Eurasia. With the development of new cultural technologies in western Eurasia, from the horse to the wheel, Xinjiang came to be settled by peoples from the west. This explains why 4,000 years ago blondes were relatively common in Xinjiang despite its Asian location. Around 2,000 years ago a reverse historical process occurred, the migration of Turkic peoples to the west from their Mongolian urheimat. In some places the migration’s genetic impact was minimal; e.g., Turkey. But in Xinjiang, which is close to the original Turkic homeland, the cultural change was accompanied by a demographic shift as East Asian genes nearly approached parity with the indigenous genetic substrate.

The genetic inferences derivable from the above paper are interesting, but not revolutionary. They don’t overturn our understanding of the history of Xinjiang, rather, they supplement, clarify and quantitize the picture. Similarly, we knew that there was some non-European admixture within the white Argentine population, but now we have a better sense of the extent of that admixture. In a chain of inferences the ability to relate elements of an argument or concept quantitatively is essential to prevent inaccurate projections.

Via Dienekes.


  1. #1 John Emerson
    March 28, 2008

    A point I don’t remember seeing systematically in these discussions is the fact that there are different types of “population”. Some populations are stable (with little admixture from outside) for centuries — Icelanders are a major example, but peasant populations in general are very stable, with very little migration and only the occasional disruptions caused by wars, epidemics, and other factors which lead to resettlement by outsiders.

    Trading people, sailing peoples, and nomads, by contrast, are continually churned. In the case of nomads, besides actual migration you also have admixtures resulting from kidnapping, slaving, adoption, intermarriage with allies, reaffiliation of renegade subtribal groups, etc. And for nomads travel was relatively unrestricted. (In 1700 the range of the Kalmyks included Xinjiang, Tibet, Moscow, and the Crimea; in 1815 their troops even rached Paris, though that was under Russian auspices.)

    The Uighurs are a sedentary people, but trade has always been important to them. Their ancestors were nomads (before 900 AD) and they’re surrounded by nomadic peoples with whom they intermarry. They’ve also intermarried with people passing through on the Silk Road, and with their trading partners from Beijing to Baghdad.

    An additional point specific to the Uighurs is that if the Tokharians have any descendants, they’re probably among the Uighurs. Tokharian languages were in use in Xinjiang at least as late as 800 A.D., and possibly a few centuries later. They were a Caucasian people, and possibly European. (The Uighurs also have Scythian, Persian, and other Iranian ancestors.)

    Besides just pointing out that the Uighurs have a very mixed past, it also can be said that the Icelanders and the Uighurs are two examples of historically recorded nation-formation. The Icelanders are a generic SCandinavian-Irish mix, but because of inbreeding and a number of decimations by plague and famine, they have a very distinct genetic footprint — one of only four European genetic outliers, according to Cavalli-Sforza (the others being the Lapps, the Basques, and the Sardinians.)

    Whereas the Uighurs are a genetic hodgepodge, but this does not prevent language, religion, kinship relations, and geographical factors from giving them a distinct national character.

    In short: 1250 years ago there was no such thing as a Uighur (in our sense) or an Icelander. The Icelandic ancestors were scattered through Norway, Ireland, and probably Scotland and Denmark, and completely mixed in with their neighbors.

    As for the Uighurs, their diverse ancestors were scattered all over Eurasia: Xinjiang (Tokharians), China, the steppe (Iranians and Turks), and West Asia (Persian peoples). And in fact, there was a nomadic pagan or Manichaean Turkish people called “Uighurs” in Mongolia which came to Xinjiang and took over around 900 A.D., but any way you look at it (genetic, cultural, geographical) they’re only one of a number of threads coming together to comprise the present Uighurs.

  2. #2 John Emerson
    March 28, 2008

    My above post was written before reading Razib’s “more”. More alter.

  3. #3 Martin R
    March 28, 2008

    Cool stuff! One suggestion though: if talking to humanities scholars about these things, avoid the word Urheimat. It’s a four-letter word.

  4. #4 razib
    March 28, 2008


  5. #5 IanR
    March 28, 2008

    I love these studies of genetic admixture. I’d love to be able to do one on my own family – my fathers grandfathers were supposed to be Pathans (at least one was Bihari, perhaps both), but my father has an obvious “East Asian” look. But he also a “very light skinned” Hindu great-grandfather, and a very dark skinned Hindu great-grandmother. Then there’s the family who lived across the road who may have been my grandfather’s cousins, or may have been “jahaji” family (kinship bonds developed by spending three months together on a boat from India).

  6. #6 razib
    March 28, 2008

    ian, i don’t think the current over the counter kits have the level of granularity you are looking for, but Y lineage tests might be able to confirm pathan ancestry since they’re iranian.

  7. #7 Luis
    March 28, 2008

    “Hybrid”??? A mule is a hybrid, a horse is not. A liger is a hybrid, a lion is not. X Cupressocyparis leylandii (a very common gradening tree) is a hybryd, a cypress like its ancestor the Cupressus macrocarpa is not.

    Coming to humans, maybe some ancient remains, like the Lagar Velho kid, were hybrids of different human species but Uyghurs are 100% Homo sapiens, not “hybrid”. They may be heavily admixed but never “hybrid”: their components are all of the same species.

    Note: it seems the term is also used in artificial breeding of same-species breeds or cultivars. But again this is not the case of Uyghurs, whose admixture wasn’t directed by some mad scientist.

    Otherwise it’s somewhat interesting, specially the estimate date for the main admixture event that seems to coincide with Turkic expansion through Central Asia, some centuries BCE.

  8. #8 razib
    March 28, 2008

    luis, i don’t give a shit about species concepts. ergo, you can infer what i think of a rock-hard definition for “hybrid.”

  9. #9 Luis
    March 28, 2008

    Everyone is “hybrid” then. You know perfectly that no one belongs to any pure stock, right?

    But it’s a wrong choice of words in any case.

  10. #10 Reader
    March 28, 2008

    One thing I never liked was how the word ‘European’ is used instead of white or Caucasian when Tocharians/Uyghurs are discussed. It implies a certain level of ethnocentrism and some fantasy that one’s (read Western Euro’s/American’s) ancestors spread as far east as China. Whereas it is more like the ancestors of Europeans/Uyghurs/Iranians etc spread to many different areas.

  11. #11 pconroy
    March 28, 2008


    Don’t forget that the European Uyghur ancestors (60%) are likely the Tocharians, and there is a wealth of evidence to suggest that they spoke a language akin to Western Indo-European languages, especially Celtic. Their dress – tartan/plaid – and material culture, sheep, draft oxen, wheeled carts for transport, is most like that of Central Europe. Their appearance is like that of Northern Europe. They carry mtDNA found in Western Europe and some like T1 which is found mostly in Northern Europe. I am Irish and mtDNA T1 as a fact.

  12. #12 razib
    March 28, 2008

    But it’s a wrong choice of words in any case.


  13. #13 toto
    March 29, 2008

    SLC24A5′s derived variant has swept as far as Sri Lanka. The fact that it isn’t extant at appreciable frequencies in East Asia suggests to me that Uyghurs did not serve as a mediator population between the east and west.

    Uh… If we follow the conventional wisdom that SLC24A5 is primarily a “pale skin gene”, and selected for this reason, and if we accept that Northern Chinese get their pale skin from another gene, then why should we expect SLC24A5 to sweep through East Asia? The job it does would aready be “taken”…?

  14. #14 razib
    March 29, 2008

    from another gene, then why should we expect SLC24A5 to sweep through East Asia? The job it does would aready be “taken”…?

    well, you obviously didn’t read my links. the key is that if an allele has a 10% selection coefficient its sweep should be powerful enough that the gene flow has to be really (shockingly) low to prevent it from spreading before a new mutant arises which reduces its local s. get it?

  15. #15 Reader
    March 29, 2008

    pconroy, Thanks. I didn’t know that.

  16. #16 Luis
    March 31, 2008

    Not Western Indoeuropean for most linguists, Pconroy but a branch of its own that apparently kept the archaic “centum” feature also present in some Western IE languages (but not Slavic nor Baltic) and other non-affiliated IE languages such as Greek. Centum/Satem is not anymore accepted (in most cases) as a fundamental devision within IE tongues, but just a phonological change that took place in the steppes after the earliest expansion of his prolific ethno-linguistic group.

    In brief: there is Western IE (most European IEs), there is Eastern IE (Indo-Iranian) and there are other IEs, among them Tocharian (but also Hittite, Armenian, Greek, Albanian). Classification trees vary somewhat depending on who you read (is Latin closer to Celtic or Germanic? Are Armenian and Greek related in spite of one being satem and the other centum?) but this basic branching is generally not affected. And Tocharian always ends up branching out near the root.

  17. #17 pconroy
    March 31, 2008


    There are not as many classification problems as you seem to suggest.

    Celtic and Latin are clearly related – I’m fluent in Irish Gaelic and took 5 years Latin, and can attest to this.

    Yes, Greek and Armenian are fairly closely related, but only marginally more than they are to Indo-Aryan. The Armenians language originated with the Phrygians, a Balkan people after all.

    Tocharian branches from Proto-Celto-Italo-Tocharian.

  18. #18 Rähmätjan
    April 9, 2008

    I think I have to read this post carefully.
    but it’s true that we (Uyghurs) are never similar with chinese..