Central Eurasian genetic specifity detected

Found at Dienekes and GNXP:

Hui Li et al., Genetic Landscape of Eurasia and “Admixture” in Uyghurs. AJHG 2009. Now freely available at PubMed.

The interesting stuff is in figure 1:


So we have now a cluster centered at the Khanty (an East Uralic population) but strong in Central Asians that is distinct from both West and East Eurasians.

I think this is the coolest discovery of autosomal genetics in quite a long time.

Dienekes protests about the inconsistence of this with Y-DNA but I fail to see the connection, because Y-DNA or even mtDNA, surely scattered in the early Upper Paleolithic when Eurasians were still very much undifferentiated, while these components are surely shallower and represent regional homogeneization processes that happened surely only after the LGM. Also the impact of Y-DNA flows may be really weak: Buriats and Finns share Y-DNA lineage, as do Polish and many Indians or West Europeans and some Central Africans but it's obvious that these don't correlate too well with autosomal genetic clustering which has its own processes of regional homogeneization, while haploid genetics and specially Y-DNA is subject to high odds of fixation by mere drift and founder effects.

Razib says that the authors seem to be arguing for greater number of (strategically chosen?) populations, what makes total sense and has served, when done, to add many shades to the tricomy of the old HapMap continental hyper-simplification.

However it's very possible that a deeper cluster analysis would have revealed futher clusters, maybe an Indian one (very small sample though) or a distinction between the Khanty and Central Asians or...

Where does this Central Eurasian cluster stem from. I'd say, based on haploid genetics, that it was probably created by the admixture of a West Eurasian and East Eurasian migration converging in that peripheral area and then coalescing into a rather homogeneous cluster on its own right. That would explain why the population appears almost exactly intermediate between the two major continental groups and also why there is no haploid lineage that specifically belongs to that region (all are shared with either West Eurasia or East Asia).

East Asian variations

It is noticeable that three quite distinct populations emerge in East Asia: inland, coastal and south. These are probably essentially comparable with the blue, yellow and one (or all together) of the SE Asian components detected in the HUGO Consortium paper.

More on "something to ponder"

This post is a continuation of Something to ponder, posted a few days ago.

In PLOS-Genetics, March 2007, C. M. Nievergelt et al. published an article with the quite unattractive name of General Analysis of Molecular Variance. In the article they published this neighbour-joining tree:


(For the legend and better image access it directly: HERE).

When compared with the more recent graph (of Jun Z. Lin et al) commented before, the similitued is striking: West Eurasians are again closest to Sud-Saharan Africans, while East Asians are the most distant ones. Sadly there is no data in this graph re. South Asia but another thing is noticeable: unlike in Lin's result, here Amerindians (or rather the root of the American branch) are closer to the root than Oceanians (New Guineans only in this graph). But as the American branches are also much longer, this may partly explain that difference.

Leaving aside this divergence, the two graphs are surprisingly similar in placing East Asians as the most divergent group from the tree root (independently on wether you place it inside the African branch or at the African-Eurasian divergence point). The authors mention that the distance from Addis Ababa is the strongest predictor of genetic background similarity among the individuals and populations. But China is still much closer to Addis Ababa than North America unless you consider the transatlantic route (which we know was not the case).

Nievergelt et al. add: but the world regions explain variation in genetic background similarity over and above this measure, suggesting that diversity among individuals within populations situated within the same world region is not completely captured by their distance from Addis Ababa. Worthy note indeed but still unable to justify the extra genetic distance of East Asians.

I am still pondering the why of this anomaly.

In other notes, in this graph it's quite noticeable the apparent homogeneity of European samples and their belonging to the larger West Eurasian group. Inside this, Near Easterners show much larger differences among them than Europeans do. Another clear single branch (for those that still speculate with several prehistoric migrations into America) is that of Native Americans.

Something to ponder (genetics)

Popular blogger on genetics, Razib (Gene Expression), has recently included a graph from Jun Z. Li et al, 'Worldwide Human Relationships Inferred from Genome-Wide Patterns of Variation' (Science, 2008). The graph is:

Soon I noticed that the two axis aren't proportional: first the distance marks of the PC2 (horizontal) axis are 250% larger than those of the PC1 (vertical) axis. Also the fact that the PC1 represents 52% of the genome and the PC2 only 28% introduces another large distortion. The horizontal axis should be a lot shorter (or the vertical axis a lot larger) for the representation to be realistic. So I corrected it:

I know it's not very pretty but it actually represents more accurately the genetic distances - and you can check for the details in the original above. As you can see I also added concentric rings, centered in the two African dots (representing real people) that are closest to the Eurasian samples.

After doing that, it seems apparent that East Asians and Native Americans are more distant genetically from the common African origins than West Eurasians, South Asians and the natives of Oceania. Do not try doing that in the original graph because the distortion of axis makes East Asians actually appear closer to Africans and Europeans the more distant - but it's just an effect of that convenient (but misleading) distortion.

The greater separation of East Asians and Native Americans from the common African origin may be a function of distance but that alone should place Native Americans more distant and that's not the case. It is really something to think about and may be related to the unique circumstances of East Asian paleodemographics, that seem to involve greater bottlenecks than the rest and a large Neolithic colonization possibly. In any case something to ponder.

Biased European genetics (again)

I am pretty sure that it's not really intentional but each time they study autosomal European genetic variation, they forget to study the central strip (French, Austrians, Hungarians, etc.). They seem to like going to the extremes and somehow "demonstrate" a fallacious discontinuity between Northern and Southern Europe.

This is the case of the new research by a US team led by Chao Tian. The overall results are most visible in figure 2C (excluding Askhenazis) that show both a distribution along the East-West geographical axis and along the North-South axis (maybe more apparent for lack of intermediate populations). Smaller components (figure 6) also seem to emphasize the E-W dominant cline, though they are dismissed by the authors.

This last is surely wrong. Not much older studies of the same kind evidenced (Bauchet et al, 2007) that when taken only two components the results are actually much distorted. Often smaller components in the overall, are very important and even dominant in one specific population. These locally dominant components become invisible when only the two or three more extended are considered, making large geographically defined populations to be classified by a minor component of their genetic pool.



The above image is a five-pole diagram I drew some time ago based in the K=5 graph below from that other study (the study reached to K=6 but the 6th component was too diffuse to matter, maybe it is a Balcanic or Eastern European element, as these areas were not studied).

Considered only two components (K=2, not shown but corresponding to the red and blue ones), Spanish samples, for instance, fell almost completely in the red "Near Eastern" zone, while Basques resulted extremely ambiguous (due to near lack of either of these two components).

Instead, seen as a plot of five components, Spanish and Basques clearly cluster primarily with themselves and no one else. Some Spanish are somewhat intermediate with Eastern Mediterraneans while others are intermediate with Basques but mostly they cluster on their own.

Another find of the K=5 plot is a Central-Northern European cluster (green) distinct of the "Finnic" blue marker. Also it's noticeable that many Northern Europeans show tendencies towards not just Finns but also Basques or even Southern Europeans in some cases. Again the lack of representation of the intermediate strip (France is only represented by one sample, while the Danubian basin, the Balcans and Eastern Europe are totally absent) creates some distortion, enhancing N/S differences.

By the way, how do I read these clusters? In my opinion the Iberian (cyan), Basque (orange) and Central-North (green) clusters must represent late Paleolithic Magdalenian and/or Epipaleolithic populations: those of the Iberian, Franco-Cantabrian and Rhin-Danub regions respectively. The two principal components instead would represent two later arrivals: Neolithic for the red (Eastern Mediterranean) one and Uralic (Fino-Ugric) for the blue one - though this last one poses some difficulties of interpretation actually (it is very possible that this "Uralic" element has been distributed by Indo-European migrations as well, specially those linked to Scandinavia and the Baltic region, like Germanic peoples).