June 15, 2014

Mexico's Native American diversity

Interesting study on Mexico's Native American diversity:

Andrés Moreno Estrada et al., The genetics of Mexico recapitulates Native American substructure and affects biomedical traits. Science 2014. Freely available with registrationLINK [doi:10.1126/science.1251688]

Mexico harbors great cultural and ethnic diversity, yet fine-scale patterns of human genome-wide variation from this region remain largely uncharacterized. We studied genomic variation within Mexico from over 1000 individuals representing 20 indigenous and 11 mestizo populations. We found striking genetic stratification among indigenous populations within Mexico at varying degrees of geographic isolation. Some groups were as differentiated as Europeans are from East Asians. Pre-Columbian genetic substructure is recapitulated in the indigenous ancestry of admixed mestizo individuals across the country. Furthermore, two independently phenotyped cohorts of Mexicans and Mexican Americans showed a significant association between subcontinental ancestry and lung function. Thus, accounting for fine-scale ancestry patterns is critical for medical and population genetic studies within Mexico, in Mexican-descent populations, and likely in many other populations worldwide.

Fig. 1-D
First of all it has to be highlighted that the sentence "some groups were as differentiated as Europeans are from East Asians" is a bit misleading. It refers to the raw FST parameter (Fixation Index) which in these cases is caused by extreme drift, product of isolation and small number endogamy.

Otherwise the Seris (Comcaac), who are the only population affected by the claim, are clearly derived not only from the same root as the rest of Native Americans but more specifically from the ancestor population of the Tarahumaras (Rarámuri), as fig.1-D reflects (right). 

The Seris are a small population of coastal Sonora who add up to less than one thousand people and have remained proudly distinct, not only from the colonial population but also from other fellow Native Americans. In spite of this long extreme isolation that makes the appear "as differentiated as Europeans are from East Asians", it is apparent that they must derive from the Uto-Aztecan populations of NW Mexico (and maybe also across the border). 

K=9 (fig. 2-B-part)
Other very isolated and heavily drifted populations are the Lacandon and Tojolabal Mayas. Again, in spite of their radical isolation, they seem related to other Mayas by origin. In these cases their languages are recognized as members of the Maya family, while the Seri language is considered an isolate. 

Actually the extreme FST scores only apply between these extremely drifted populations: FST{Seri-Lacandon}=0.136, FST{Seri-Tojolabal}=0.121. 

This reference is interesting because it explains how subcontinental levels of differentiation can happen in relatively short time if the founder populations are small and isolated for some 20 Ka. It is a warning call against reaching to too many conclusions based only on populations with a long history of isolation.

Otherwise the Seri FST scores are high but more normal: 0.087 to 0.096.  See table S-4 for further details. 

The tree is interesting also because it suggest a main division separating the Nahuas from the rest of the Uto-Aztecan meta-population (Saris included). The Nahuas, who approximately correspond to the the ancient Aztecs, are actually divided in several groups, which seem rather akin to their immediate neighbors and not so much among them or their linguistic relatives. 

This implies that, as the ancestors of the Nahuas migrated southwards, they assimilated so many locales that they largely lost their distinctiveness. In the ADMIXTURE graph to the left, we see that they do keep a variably small fraction of Uto-Aztecan affinity (not just them, also the Purepecha and Totonac, whose languages are distinct). 

Otherwise Mexican Natives have two main components at K=9: the main Mexican one (blue) and the Maya one (orange). The Maya division is also apparent in the tree. 

However it must be mentioned that the ADMIXTURE run available in the supp. materials (fig. S-10) reaches down to K=20, showing further differentiation between the various Mesoamerican populations dominated by the blue components at K=9. 

For comparison, in the European segment only the Basque component shows up as distinct in all those runs (since K=10). So we are talking about a fairly diverse population compared with European relative homogeneity.

Sequence of further components or distinctions showing at depths greater than K=9:
  • K=12: Tarahumara
  • K=14: Nahua-Purepecha-Totonac
  • K=15: Tepehuan
  • K=16: Purepecha + Jalisco-Nahua
  • K=18: Triqui
  • K=20 Totonac

Mestizo ancestries

An issue worth mentioning, particularly in relation to the so far unconfirmed but quite plausible Canarian origin of a large share of the "European" ancestry in the Caribbean region, is that the European ancestry of Mexicans seems essentially Iberian, as shown in fig. S-14:

I am anyhow awaiting for a sensible geneticist to address this question properly. When dealing with Mexicans and other Latin American populations of complex colonial ancestry, it seems quite apparent that so diverse European samples are in excess and that instead a North African control is surely missing instead.

A more regionalized approach to Iberian ancestry could also be interesting.

Regarding the Native American share of the ancestry, a finding of this study is that there is important regional variation: Yucatan and Campeche Mexicans have clearly strong Maya ancestry, while in Sonora it is something more like Tarahumara and in the core of Mexico it seems Nahua-like or from other "central" populations like the Zapotec or Totonac. See fig. 2A for details.

There is also very minor Tropical African ancestry across the board, somewhat more relevant in Guerrero and Veracruz, states which historically hosted the main port cities of New Spain and still have some small Afrodescendant populations.


  1. Thanks, Luis, for covering this paper.

    Genomes Unzipped has a related post:

    The was an article about this paper in our local paper here is San Francisco which I think, unfortunately tried to put a "personalized medicine" slant on this research. Oversold, I would say.

    Here's the article in sfgate (the online version of the San Francisco Chronicle):


  2. I'd bet dollars to donuts that if we had a complete picture of ancient DNA from the mesolithic/neolithic transition in Europe, there would be a similarly diverse picture, albeit with more external influences and a shorter period for drift to occur. Structure that may have been lost in some of the demographic upheavals that have occurred since.

  3. On the study, main points of interest to me are

    - The contribution of European ancestry in the Mexican populations plotting as Iberian rather than hyper-Iberian. That seems to help confirm as the more parsimonious hypothesis that the effect in the Caribbean paper reflects real ancestry rather than an artefact (so I was wrong when I thought it might be).

    - Native American ancestry in European-Native American Mexicans reflects regional ancestry structure. This seems to present a model whereby Iberian settlers (men) have mixed with the local people (women) in each region. An alternative model, which seems wrong, would've been that Iberian settlers mixed with Native Americans from a particular region, then this mixed group expanded demographically.

    On FST as I understand, is effectively a measure of between population diversity that is scaled to reflect within population diversity.

    The calculation generally used (I think) is FST = (Between population differences - within population differences ) / Between population differences.

    So you can see that as within population differences decline, the calculation approaches (between population differences)/between population differences, i.e. 1.
    FST gets higher as within population differences decrease, even if the absolute size of between population differences stay the same.

    The absolute size of between population differences in American populations is almost certainly much smaller than between Europeans and Chinese, yet the in group diversity is even lower relative to that.

    This also should apply when we compare Africans to one another and to Eurasian groups - the difference increases due to the lower within population diversity within Eurasians. And to a lesser extent when looking at diversity in East Asians populations relative to West Eurasian populations.

    It's the same kind of reason why physical anthropologists apparently (I don't have much knowledge here) find lots of ethnic "phenotypes" in Native Americans relative to the Old World. More of between family variation and individual variation in body form is shared in each population in the Old World, so effectively gets discarded when trying to build population phenotypes. Not so much the case in the New World.

    1. What you say about the Fst is particularly interesting. Your equation shows clearly how as internal diversity drops, Fst approaches the maximum possible value of 1. This can be extremely misleading if taken without due consideration.

      "An alternative model, which seems wrong, would've been that Iberian settlers mixed with Native Americans from a particular region, then this mixed group expanded demographically."

      Actually, if we look carefully at fig.2-A, it seems apparent that the dark blue "Mexica" component in Mestizos is more widespread than it should be on purely local admixture basis. So I'd say that there was also an expansion of the Mexica component within the colonial society. For example the Durango ancestry is much more Mexica and much less Northerner (light blue) than should be on the local admixture model.


Please, be reasonably respectful when making comments. I do not tolerate in particular sexism, racism nor homophobia. Personal attacks, manipulation and trolling are also very much unwelcome here.The author reserves the right to delete any abusive comment.

Preliminary comment moderation is... ON (sorry, too many trolls).