search this blog

Tuesday, September 3, 2013

A multidimensional view of Europe + West Asia


I'm wondering whether something like this can say anything meaningful about the peopling of Europe? Any thoughts?

See also...

Principal component analysis (PCA) of West Eurasia

PCA of the world

9 comments:

Eduardo Pinto said...

Hello David,

Before any conjectures, I would like the following answered first...

Who are those West Asians branching in two different clusters on 1&3?

On 2&3 who are those West Asians overlapping with the European cluster, and who they meet with?

Same as above for 2&4.

Davidski said...

Have a look now.

Eduardo Pinto said...

2&4 looks very interesting! At a first sight it looks just like a West-East cline but it's more than that... On the bottom you have Sardinians representing South Levant farmers with Southern Europeans closeby, while at the top you have Armenians and other West Asians representing Danubian farmers cutting away through Central and Eastern Europe.

Maju said...

@Eduardo: that split is almost certainly (see previous similar post by David in which individual populations are labeled) Highland West Asia (peaking in Caucasus) vs. Lowland West Asia (peaking among Bedouins), a well known duality of West Asian population genetics, closely related to the J2 vs J1 Y-DNA distribution.

Maju said...

"Any thoughts?"

My thoughts are David that we get something similar to a K=4 or K=5 Admixture run. I do appreciate your intention to expand our understanding by doing various PC vectors instead of just the usual two but, unavoidably, it is still limited and in certain way conditioned by sampling choices.

You may say "but PC10", for example, "only weights 1%". True (I guess) for the overall pooled sample but it may well weight 30% or whatever in some specific population(s), whose specificity is therefore hidden by these big scale comparisons. One of the issues we know that exist is, for example, that Western European populations are quite diverse and specific, while at the same time also having relevant levels of admixture with other populations be them European or trans-Mediterranean, so they will show here their admixture factors but not their specificities.

Anyhow:

PC1 is the main European-specific vector and indicates NE-European affinity, peaking among Finns and nearby populations like Lithuanians and such. By default all Europeans tend more or less to that pole... but also highland West Asians, especially Caucasians. The less akin are Lowland West Asians and North Africans.

PC2 can also be considered a European vector but only marginally so: it points to the North Caucasus and Chuvash peoples, which are indeed European but in a very peculiar way genetically speaking. Eastern Europeans and Highland West Asians are generally akin to it. North Africans, Sardinians and Basques are not akin (what says very little in favor of linguistic speculations on Basque-Caucasian affinity, BTW).

PC3 is somewhat similar to the previous but more clearly Highland West Asian (the Chuvash are out) and certainly not European. The less akin are again North Africans (most of them).

PC4 is the Lowland West Asian component and curiously again the less akin are again (most) North Africans.

There's very little information about European settlement here. We can however draw a dichotomy of West Asia vs Europe at PC1 vs PC2. The extreme attributes of Europeanness are here high PC1 and low PC2, while the extreme attributes of West-Asianness are the opposite: high PC2 and low PC1.

Rather than Finns or Eastern Europeans, the extreme characteristic of Europeanness is at Basques. Draw a diagonal with those values and then perpendicular lines: right after Basques most other Europeans (incl. Finns, Lithuanians and Sardinians) perform similarly and then there is another tier which I believe corresponds to Balcanic peoples. West Asians instead (excluding Cypriots, I believe) show no or almost no gradation being all similarly non-European (although diverse). North Africans appear more "European" than West Asians in this axis, what makes good sense for what I know (lots of mtDNA H and V, surely since Paleolithic times) but may also be a product of their outlier nature in this duality, what makes them tend to neutral values (because they are neither apples nor oranges but, say, bananas and the number of samples and comparisons do not favor banana-tagging).

My two cents anyhow.

mikej2 said...

In general the Eurepean plot with dimensions 1 and 2 looks great. However I still call for the sample quality, and in this particular case see oversampling of some ethnic groups. Albeit this look reasonable and perhaps nothing changes after the recalibration of the sample set, it is a fact that sampling have a big effect on the result and using somehow standard and systematic arrangements makes comparisons more reliable.

Onur said...

West Asians instead (excluding Cypriots, I believe) show no or almost no gradation being all similarly non-European (although diverse).

No. On PC1 Anatolian Turks lean more towards Europeans than Cypriots are and Balkan Turks are in alignment with some of the other Balkan populations. To see the position of Turks in the West Eurasian context more clearly, see the PCA plot in this link, in which Turks and all other populations are explicitly labeled:

http://eurogenes.blogspot.com.au/2013/08/principal-component-analysis-pca-of.html

North Africans appear more "European" than West Asians in this axis, what makes good sense for what I know (lots of mtDNA H and V, surely since Paleolithic times) but may also be a product of their outlier nature in this duality, what makes them tend to neutral values (because they are neither apples nor oranges but, say, bananas and the number of samples and comparisons do not favor banana-tagging).

Again, no. On PC1 North Africans are more distant from Europeans than Northern West Asians are and are in alignment with many of Southern West Asians. Those "North African" samples who are in alignment with Sephardic Jews are mostly North African Jews, not indigenous North Africans. To see that clearly, again look at the link I provided.

Maju said...

"No. On PC1 Anatolian Turks lean more towards Europeans than Cypriots"....

No.

Or rather yes... but only on PC1 and what I was pointing to was to the combined effect of PC1 and PC2 as described above.

So NO because you did not get my point at all, Onur: in the diagonal axis I was proposing Turks are almost indifferent in "Europeanness" or "West Asianness" relative to other West Asians, certainly less than Cypriots.

"Again, no. On PC1 North Africans"...

Again NO to you because I was talking of the combined diagonal axis, not PC1 or PC2 separately. The diagonal is the only axis that contrasts Europe vs. West Asia, perpendicularly to the means of each continental region.

Onur said...

I see. Anatolian Turks lean more towards Europeans on that diagonal axis than do all the other northern West Asian populations except Cypriots, and Balkan Turks are apparently in alignment with southern Balkan populations. These are consistent with the results of other kinds of genetic analysis.