Sunday, March 18, 2012

Dienekes' Argues For West African Admixture With Eurasians

A couple of posts at Dienekes' Anthropology blog doing various kinds of analysis of publicly available autosomal genome data argue that various ancestrally West African populations have signficiant (up to about 20%) Eurasian admixture. Specifically, he notes that "it appears that Yoruba are 82.9% of the San-centered component, BantuKenya 78.3%, BantuSouthAfrica 86.9%, BiakaPygmy 94.7%, Mandenka 80.9%, MbutiPygmy 97.9%."

In an earlier post, he argues that the admixture seems to have the closest connection to the Sardinian and Basque populations of Europe, which are among those with the least Indo-European impact genetically.

He also reproduces recent Denisovian and Neanderthal admixture results that were reached by direct comparisons to ancient DNA evidence with principle component analysis methods.

Dienekes argues that the absence of Eurasian admixture seen in prior studies flow from using populations like the West African Yoruba population, rather than more definitively indigenously African populations like the San, as poles of diversity measures, and because prior studies looked at subsets of genes tuned to Eurasian diversity rather than the larger set of genes showing diversity within Africa.

Uniparental genetic variation does not tend to confirm this conclusion, unless one subscribed to the controversial proposition that Y-DNA haplogroup E is a back migration of Eurasia rather than an African sourced haplogroup.

It seems premature to jump to major paradigm shaking conclusions from this single amateur data run, which is at odds with most of the prior literature on the subject, but it is interesting and worth further attention. The methodology is reasonably respectable, even if some nuances of it, like the use of TreeMix software to look at models with lateral gene sharing between populations, or the incorporation of ancient archaic hominin and chimpanzee genese into PCA analysis, are relatively new and may have quirks that are not yet terribly well understood.

The TreeMix software itself has disclaimers about the limitations of its methods in admixture events that don't closely approximate the unidirection, effectively instantaneous, and reasonable recent (by population genetics standard) assumption built into its models. For example, it could be that what Dienekes is seeing is not admixture, but a shared common origin in a proto-Eurasian population in Africa that post-dates the divergence of San and Pygmy populations from other populations in Africa, one branch of which expanded into East African and another branch of which expanded into West Africa.


Maju said...

I did not comment (because it's tedious to post where you have no email feedback and comment moderation, never mind that D. is not open-minded at all) but it's obviously an effect of the low number of components and K=5 and the concentration of the "sub-saharan" among Pygmies and Bushmen (San), what makes it actually a "paleo-African" component (in Dienekes terminology).

So the algorithm, when having to place other Africans, of which there are not so many, between the "paleo-African" and the Eurasian components, says: they are somewhat in between.

He's just hallucinating without drugs.

Andrew Oh-Willeke said...

I have a comment at Gene Expression that fleshes out what could be going on statistically a bit better than in the post here (particularly in the Basque-Sardinian examples to which there is a link, the only Southern Europeans in his model).

Researching that comment also increased my confidence in the time and place of the ethnogenesis of the people who speak the Chadic languages now (in the Neolithic Subpluvial around Lake Chad via the general area where Cushitic languages are spoken now), but this still leaves the source of these peoples prior to ethnogenesis somewhat cryptic, except that they ultimately made their way to Lake Chad probably via the Nile Valley and went up the White Nile instead of the Blue Nile. How R1b-V88 goat herders wound up in Pre-Dynastic Neolithic Egypt is harder to say.

Maju said...

If there was some X% Eurasian among Chadic speakers but not the Mandenka or the Yoruba, then it would be a reasonable possibility (would need some double or triple check but my hunch that probably correct, at least if X%<20%). That would fall within reasonable expectations.

But in this exercise D. is finding Eurasian admixture all around, including the Mandenka, Yoruba and what not! And that's obviously an error of the analysis strategy - Dienekes' error.

Double error because he believes it to be true when he should be systematically doubting these unusual and unexpected findings.

I think that Eurasian back-flow into Africa is already more or less well known: North Africans are mostly of West Eurasian origin, Horners have some 20% of West Asian admixture and the rest very few (Maasai, Fulani, possibly Chadic speakers) or nothing (Mandenka, Yoruba and all the rest).

But then D. finds overnight 68% of Yoruba blood being "Basque"! Sorry but it doesn't make any sense.