This question have been following me since the last previous blogpost where I found this geographical distribution of chunkcounts PCA dimension 4 between Euroasian populations.
Everybody would probably agree about the northern distribution shown in blue and green apparently showing a genetic connection between North-East European populations like Saamis and Finns and Northern Siberian populations all the way from Fennoscandia to Beringia in Eastern Siberia as shown many times in this blog, from university research and from other bloggers. However if we look closer at the color distribution for the map:
We should of course not take this color distribution to literally as PCA plots distirbutions can be affected by many things but as we can see the red grade is close to the brown. This probably means the whole area from western continental Europe to East- and South-East Asia appear to show haplotype similarity. So it appears not only to have been a northern East-West influence but also a southern East-West influence as well.
To investigate this further I did a seperate run Chromopainter using 23k linked SNP at Chromosome 1 with a selection of individuals that had been phased together with over 2000 individuals using a high number of iterations in BEAGLE minimizing the error rate to a minimum. I used Chromopainters admixture functionality to design a admixture test using the relevant East-Asian, African and Siberian populations. As far I know from previous screening using ADMIXTURE all these individuals appear unmixed without any European admixture.
As we can see North-East Europeans appears to score low on East-Asian influence while continental Europeans from western and central Europe appear to score rather high vs the East-Asians. As expected we see the largest African like influence among the more southern populations and the North-Siberian influence in the more North-East European populations. Please note as this result is based on only 1 chromosome it doesn't always correlate 100% with the result of a whole genome analysis. There is a strong negative correlation between the East-Asian and Siberian component at -0.56 and there is even a stronger negative correlation between the Siberian and African component. The correlation between the East-Asian and Siberian component is weak at -0.17.
We can further see this using a area plot for the same data. The continental European populations from western and central Europe appear quite consistent to be closer to the East-Asian populations.
In Finestructure (from the previous run with the whole genome using superindividuals) the clustering appear to confirm what was observed above. The East-Asian influence appear rather consistent for all European populations included. In the North-East European populations this East-Asian influence appear to be lacking or less but the North-Siberian influence appear as expected from earlier analysis.
Individual results for project participants for PCA dimension 4 shown in the first image above. The first column show the actual PCA plot value and the second the ranking seen from top as in East-Asian and bottom as in Siberian.
This is a seperate PCA run using Europeans, East Asians and Siberians as reference. The X axis (horizontal) shows dimension 1 and the Y-axis (vetical) shows dimension 2. The X axis shows the European vs East-Asian influence, the Y-axis shows the European vs Siberian influence.
This is a seperate PCA run using Europeans, East Asians and Siberians as reference. The X axis (horizontal) shows dimension 1 and the Y-axis (vetical) shows dimension 2. The X axis shows the European vs East-Asian influence, the Y-axis shows the European vs Siberian influence.
Europe vs East Asia and Siberians overview
Europe vs East Asia and Siberians zoomed at Europeans
Europe vs East Asia and Siberians zoomed at Europeans detailed
Gradient map of PCA dimension 1 - blue most East-Asian, brown least East-Asian like.
Gradient map of PCA dimension 2 - blue least Siberian like, brown most Siberian like
What do NO, TU, HUT and AL stand for?
SvarSlettNO = Norwegian, TU = Turkish, HUT = Hutsul, AL = Albanian.
SlettThanks. Can Turkish participants join the Fennoscandia Project?
SlettI do have a Turisk participant TU1.
SlettDoes that mean yes?
SlettI am restrictive, there already exists public Turkish reference population available so strictly speaking I dont really need them but sure a few Turkish may of course be good to have with especially if the Turkish samples is restricted to certain areas of Turkey.
SlettI guess you mean Balkan Turks.
SlettThose references Turks is Balkan Turks?
SlettNo, the reference set of Turkish samples are either entirely or mostly composed of Anatolian Turkish samples. None of the reference set of Turkish samples I have seen so far is entirely or mostly composed of Balkan Turks. My question was related to your Fennoscandia Project Turkish sample(s) (e.g., TU1), not reference Turkish samples.
SlettThe individual TU1 cluster with Turks so I have no reason to believe this individual is not Turkish.
SlettI did not say he/she is not Turkish. I only asked whether your Fennoscandia Project Turkish sample(s) (e.g., TU1), and not your reference Turkish samples, are exclusively Balkan Turks. I asked that question because you said "a few Turkish may of course be good to have with especially if the Turkish samples is restricted to certain areas of Turkey" in response to my question "can Turkish participants join the Fennoscandia Project?" and also because that Fennoscandia is a Euro-focused project. My question was not based on TU1's genetic results.
SlettPerhaps it shows influence from the Mongol Empire in Continental Europe.
SvarSlettIs there any (except perhaps in some Noghai and Tatar groups)?
SlettLooking at the locations where it is high, Hun or Turkic influence is more likely.
SlettI am not sure but my gut feeling is that its quite old and may have come into Europe through the first farmers as it doesnt seem to have reached Saami and Finns..
SlettThe values don't reveal anything about Mongol or Hunnic influence. They reflect two things: the movement of Siberians into NE Europe, and the fact that the Siberian element likely contains a measure of ANE, which it shares in common with Europeans in varying degrees. The higher 'East Asian' values are the result of the imbalance within Europe regarding relative Siberian and East Asian affinities caused by the two mentioned factors.
SlettThe PCA is still interesting, it doesn't use relative African/EA/Siberian comparison and still shows peculiarities. There is an odd difference between Finns and Mordovians/Vologda Russians in affinity to Siberians and East Asians alike (a difference is visible in the heatmap too), while Saami (North especially) do show high Siberian affinity but not as high as you'd expect considering the heatmap.
SlettThe East Balkans also correlate in PCA and heatmap regarding East Asian affinity.
"There is an odd difference between Finns and Mordovians/Vologda Russians in affinity to Siberians and East Asians alike (a difference is visible in the heatmap too), while Saami (North especially) do show high Siberian affinity but not as high as you'd expect considering the heatmap."
SlettI'm also somewhat surprised. Besides, it seems to me that as Fi14 I'm personally slightly drifting into some strange direction. I wonder if there's an important reference group missing, maybe related to some eastern baltic relic population?
All this shows is that Eurasian populations that lack direct Siberian admixture are more related to East Asians than to Siberians when they're forced to 'choose' between the two. Obviously Finns have more East Asian admixture than Spaniards, for example, but because both the 'Siberian' and 'East Asian' (which are both East Asian) components are posited opposite each other in this analysis, then, because Finns have direct 'Siberian' admixture, it follows that they should have an additional percentage in the Siberian column vs. the Spanish, and thus a lower East Asian affinity than the Spanish, but only because both values (ignoring the African column for theoretical simplicity) must add up to 100%.
SvarSlettI should change my first statement to this (to better fit the values presented): "Eurasian populations that lack direct Siberian and East Asian admixture are relatively more related to East Asians than populations that have direct Siberian admixture but lack (at least relatively) East Asian admixture."
SvarSlettThis, of course, is not necessarily evidence of a positive admixture event in the populations that score relatively higher 'East Asian' affinities.
It doesn't seem to be relative for all populations, in the genomewide Finestructure run Tuscans for example do not show elevated affinity to East Asians but East Balkanites and Hungarians do, even more than Russian/Ukrainian cluster. That is an area where recent admixture is possible.
SlettThe Spanish have a different recent admixture event (tested by Lazaridis et al to involve either a small "true" Sub-Saharan element or a large SSA-influenced Mozabite-like element) which likely would not increase East Asian affinities - and in the heatmap it indeed does not.
About placement on the PCA map, is the position determined strictly out from the affination to the reference populations (SIB, AFR.EA) so that internal sharing betweed participants don't affect one's placement.
SvarSlettI have always think that it is odd that people in fennoscandia would have more East asian genes than than for example east europeens if one think about the geographical situation and the enviroment that separate these populations.
If internal sharing pull was the main reason for placement, I think Vologda Russians and Mordovians should have been pulled further away from Siberians by Lithuanians (to which they are very similar as seen in the European-wide PCA).
SlettThose European plot with the "V" shape with the South Europeans at the base and with Fennoscandians on one wing and East Europeans on the other I believe shows an internal variation in Europe. The plots seen here shows external variations or at least in the first dimensions like here. They are independent. It is fully possible that even Lithuanians and neighbouring populations in their internal variation is very similar but when compared to with external influences can be different. I so on the individual PCA using Europeans vs Sib and EA that most Lithuanians didnt have much of this EA-SIB pull but at least one Lithuanians did and appeared similar to Ukrainians and Belorussians.
SlettThe admixture test above was based on Chr1 and using just a few random picked individuals. It is possible that there is some margin of error here. On the other hand the admixture test above catches everything not only the variation from PCA dimension 4. NO1 do also pull towards the East-Asians on the other individual PCA plot and more than most Scandinavians.
Slett"I have always think that it is odd that people in fennoscandia would have more East asian genes than than for example east europeens if one think about the geographical situation and the enviroment that separate these populations."
"
the PCA seems to include just SIB and EA, i wrote wrongly SIB,AFR,EA.
SvarSlettFrom this PCA, in genes which contrast Cambodians to Yakuts, Finns and to a lesser extent Swedes (etc) fall a little more on the Yakut side, while continental Europeans seem to fall neutral, without any more affinity to either.
SvarSlettThis would be expected I'd think, for a population which lacked influence from and did not influence either Cambodian or Yakut ancestors.
To actually check whether the above is true, it would help if you (Anders) showed a graphic fitting the color scale to numbers. If continental Europeans do fall neutral, then the number corresponding to their color should be 0. We don't really have any idea where 0 is just from the color scale, and it isn't necessarily in the exact centre.
Its difficult to get that color scale to numbers but the scale is made on the basis of the PCA number range so the middle scale colour should approx show the 0 number.
SlettI have updated the post with a image showing individual results. Please note that the gradient map is based on population averages in Europe not on selected individuals.
The dimension 4 difference between Mordovian/Vologdan and Finnish samples is quite considerable. I think this might have something to do with the idea presented in Lazaridis et al, about the "easternness" of various Northeast Europeans having different sources.
Slett