Tuesday, 19 April 2011

Unsupervised Northern Eurasia+new samples part 2

This is part 2 of the unsupervised North Eurasian Run. Just as any unsupervised run including such diverse populations, you can expect a lot of noise, especially when evaluating single individuals versus large n populations.
So don't see meaning in every single small component you may find in one or two individuals. It is most likely just noise. Still you can see some patterns, and I'd like to call attention to the dark blue (peaks in Amerindians) and pink (Nganasan) elements. This is an unsupervised run, all I did was decide on the populations and individuals to be analysed and how many components I wanted to ADMIXTURE to figure out. In an unsupervised run I can't tell ADMIXTURE "find this versus that" as in supervised. It finds what it does on its own. And yet results are globally pretty similar to these.

FRNA has French and North African ancestry.
FRMA French and Madagascar/South Asian
Tigrinya is from Eritrea.

Some considerations
Unsupervised mode in my opinion tends to overestimate/make up smaller components which is why I don't like it much. It clusters segments without an historically informed departing point and tends to amalgamate more exotic variation with some of the dominant component. It is also more chunky and less reliable IMO than some of my blindly supervised results.
Green in Med populations I think corresponds to the Nile Core Element (ancient Northeast African, which doesn't exist "unadmixed" anymore) which is likely intermediate between the Mesopotamian Core and very anciently related also to West and East African maybe through the Green Sahara of thousands of years ago. In the Tigrinya sample the Yoruba modal component is likewise drawing on East not West African elements.
Red and light blue elements in more Eastern populations are likely representative of assimilated Coastal Migration elements. In Europeans they are most likely noise, even though at least in some Romanians they might indicate Roma admixture.
Yellow segments could be interpreted as Central Asian elements but their unevenness and presence in unlikely places (Spain) may mean its just noise.

My point in presenting this unsupervised run is to prove ADMIXTURE does find what can be interpreted as both Fertile Crescent replacement and Amerindian-like residual segments in Northern Europe and the Caucasus on its own at the adequate K.
One very interesting new indication of the new samples is the near complete absence of Siberian/Amerindian-like segments in Germany. Germany was an entry point for early Neolithic colonization of much of Northern Europe, the Danubian Wing of the 1st Wave up the Balkans, Danube and Rhein valleys from Western Anatolia -other sources were likely the Med-Atlantic Wave via Iberia and British Isles also from Western Anatolia and the Wave from Eastern Anatolia via the Caucasus and the Ukranian/South Russian river valleys.
This region has (in West/South Germany) excedingly fertile river valleys, congenial to wheat/barley agriculture. Residual forager elements are expected to reach a nadir here. And the results speak for themselves.

Tomorrow I'll present New World data, and also the results of a few South Asian samples I threw in the analysis without my South Asian populations to keep them company. Their results are a bit artificial I think, but still revealing.

No comments:

Post a Comment