In my critique of Moorjani et al. (2011), I noticed how the authors projected West Eurasian samples on an African/East Eurasian axis and assumed that African-shift along that axis was due to the presence of African admixture.
I showed, that while some populations are shifted towards Africans, others are shifted towards Asians, so a projection along an African-East Asian axis is in reality a palimpsest of the two phenomena.
I argued that assessment of African admixture using a simple 2-population model that does not account for the "East Asian factor" leads to erroneous results. I then compared five methods of admixture estimation, showing that the Sub-Saharan admixture results of Moorjani et al. (2011) were higher than all the other methods, consistent with my hypothesis.
In the present, I show how many different European/West Asian populations are shifted towards Africans/East Asians.
Principal Components Analysis
First, a PCA plot of just the West Eurasian populations; labels are mapped onto each population's average position:
Second, a PCA plot together with 25 Chinese and 25 Yoruba; Eurasians are separated from Africans along eigenvector 1, and East Eurasians from Eurafricans along eigenvector 2.
Third, a blowup of the West Eurasian portion of the above plot:
Third, a further blowup of the plot, excluding Chuvash to make it even clearer:
Finally, here is a spreadsheet with the average PC co-ordinates of the studied populations. The first two eigenvalues are 25.92 and 11.91.
With respect to the Asian- and African- shift of West Eurasian populations, I note that northern Europeans (and Basques) are less African-shifted than southern Europeans, and, at the same time they are more Asian-shifted: the 16 least Asian-shifted populations have a coastline in the Mediterranean (excluding the Portuguese), while the 16 least African-shifted populations do not (excluding the French).
This analysis suggests the importance of choosing appropriate populations to represent Caucasoids in the global context. One often sees CEU used for that purpose. I see no major problem with that in general, as it is good for different studies to have a similar reference point, and CEU have been used for years for that purpose.
However, when dealing with the problem of admixture, this becomes an issue. CEU emerges as one of several populations with minimal African-shift, but are intermediate in terms of their Asian-shift. Sardinians, on the other hand, have mininmal Asian-shift (by far), and are intermediate in terms of their African-shift.
If one were to choose a single population to serve as a Caucasoid pole according to a criterion of maximal differentiation, then Basques are the obvious candidate, as they are tied for 1st place in having least-African shift, and 2nd in terms of Asian-shift. Indeed, a K=3 ADMIXTURE analysis of this dataset demonstrates that they are in fact the population showing the maximal contribution of the Caucasoid-specific component.
The analysis presented here also demonstrates the relative value of different population isolates in ancestry analysis. The Chuvashs, for example, are clearly not part of the genetic continuum of Europe, and neither are French Basques and Sardinians: all of these populations form very distinct clusters within the West Eurasian-specific context (first plot of this post). Nonetheless, their analysis within a global context demonstrates that they are distinct in different ways: Chuvash because of their substantial Asian-shift, Sardinians because of the substantial lack thereof.
In conclusion, It is a good idea not to employ a simple 2-way population mixture model to assess either African or Asian admixture in West Eurasians. Such a model may lead to erroneous results if it employs West Eurasians' shift on the African-East Asian axis.