Here's a Principal Component Analysis (PCA) and an accompanying biplot based on output from an improved version of my ANE K7 ancestry test. Let's call it the West Eurasia K8. This one gives more accurate estimates of Western European Hunter-Gatherer (WHG) and Near Eastern admixture proportions, thanks to the use of new ancient samples.
Clearly, ANE is the main agent causing the west to east differentiation in dimension 2. Note that even a small rise in ANE, say, 4-5%, creates significant distance between samples on the PCA plot.
East and South Eurasian admixture has a similar effect, but must be more considerable to make an impact on a West Eurasian-specific PCA like this (and it does with the obvious Volga-Ural outliers, who come from Chuvashia and Tatarstan).
On the other hand, Near Eastern admixture without ANE creates almost the opposite effect. Note, for instance, that Neolithic genomes Stuttgart and NE1 show much higher levels of Near Eastern ancestry than most Europeans, and yet they're amongst the most western samples on the plot.
This suggests that the Near East, and in particular the Caucasus, experienced a significant rush of ANE admixture after early Neolithic farmers left the region for Europe. Alternatively, Caucasus populations may have carried even higher levels of ANE than they do today, before newcomers from the Near East mixed with them. But either way, a lot of ANE arrived in the Near East at some point.
It also suggests that, overall, the populations that moved west across northern Europe after the Neolithic, and shifted northern European genetic structure to the east, did not carry high ratios of Near Eastern ancestry. Instead, they harbored high ratios of ANE and WHG. What these ratios were exactly I haven't a clue, but ancient DNA should tell us that soon.
Below are the ancestry proportions for the five ancient genomes in this analysis, in chronological order. It's interesting to note (yet again) the rising and falling Near Eastern admixture, from the Mesolithic to Neolithic and then from the Neolithic to Bronze Age, respectively, as well as the steady rise of ANE from the Bronze Age to the Iron Age.
Loschbour (Mesolithic)
ANE 0
South_Eurasian 0
Near_Eastern 0
East_Eurasian 0
WHG 99.5
Oceanian 0.5
Pygmy 0
Sub-Saharan 0
Stuttgart (Neolithic)
ANE 0
South_Eurasian 0
Near_Eastern 72.19
East_Eurasian 0
WHG 27.8
Oceanian 0
Pygmy 0
Sub-Saharan 0
NE1 (Neolithic)
ANE 0
South_Eurasian 0
Near_Eastern 69.82
East_Eurasian 0
WHG 30.17
Oceanian 0
Pygmy 0
Sub-Saharan 0
BR2 (Bronze Age)
ANE 9.62
South_Eurasian 0.08
Near_Eastern 43.96
East_Eurasian 0
WHG 45.44
Oceanian 0.48
Pygmy 0.23
Sub-Saharan 0.19
Hinxton4 (Iron Age)
ANE 15.08
South_Eurasian 0.06
Near_Eastern 35.44
East_Eurasian 0.46
WHG 48.5
Oceanian 0
Pygmy 0
Sub-Saharan 0.46
See also...
The fateful triangle
Bell Beaker, Corded Ware, EHG and Yamnaya genomes in the fateful triangle
I'll try and offer this test one way or another after Christmas. It looks pretty good I reckon.
ReplyDeleteI'm basically off on holidays now.
Quite Noble David, at least this time you know what you are doing+Saying.
ReplyDeleteEnjoy your holidays!:)....
What's surprising from the spreadsheet, is how much middle eastern ancestry Ezrya people have and how little ANE ashkenazi jews have. Seems like significant results.
ReplyDeleteMaybe but Ashkenazi Jews have well over 60% Near Eastern ancestry here. Erzya have over 20%, but it's one of the lowest ratios in the analysis.
ReplyDeleteThese results are reflected in their PCA positions, which are the same as in in my PCA based on genotypes.
David, what do you propose was the timing of the increase or arrival of ANE in the near East and Caucasus relative to that in Europe ?
ReplyDeleteSpagetiMeatball, the Erzya Middle Easterness is expected considering it shows up in many different PCA structures and by presence of West Asian in the Volga-Ural region in many ADMIXTURE runs. It's possible to produce varying PCA's of West Eurasia. As quoted from Lazaridis:
ReplyDelete-Nelis et al. 2009 analyzed primarily north European populations in a global context to observe a southern European-Baltic cline stretching from southern Italy to Latvia [add. from which diverged a cline from S. Italy to Kuusamo isolate].
-Behar et al. 2010 co-analyzed European and Near Eastern populations to show a West Eurasian PCA that strongly resembles our Figure 2 in revealing parallel north-south clines in Europe and the Near East with qualitatively similar orderings of populations along the clines.
-Yunusbayev et al. 2011 oversampled Caucasus populations and documented a divide between Europe and the North Caucasus that corresponds geographically to the Black Sea.
-Skoglund et al. 2014 analyzed a quite different set of genotyping data and presented a PCA of diverse European populations that did not recapitulate a map of Europe. Instead, they find a complex picture dominated by European/Near Eastern variation along PC1 and Sardinian-Saami variation along PC2.
I'd add to those notes that it's possible to produce a PCA with Finnish-Sardinian dimension 1 and Basque-Maltese dimension 2 (Lazaridis figure S10.5). By adding Saami and Chuvash these would become a Sardinian-Saami dimension and a Basque-Chuvash dimension but otherwise the PCA would maintain this structure quite well (like this PCA I found, although labeling makes it a bit hard to read http://oi62.tinypic.com/21b6pz4.jpg). Mordovians and Northern Russians are shifted towards Maltese end of the dimension compared to Finns and Balts in figure S10.5, and that would not change with Chuvash and Saami added. The ideal situation in any case would be to have high quality ancient samples to define these dimensions.
ANE may have arrived in Iran and the Caucasus fairly early with pressure blade technology, but probably after the early Neolithic farmers had already left for Europe.
ReplyDeleteOr maybe it came down from the Caucasus and the steppe in a big rush after the Neolithic, with the Indo-European, Kura-Araxes and other expansions. It may have then spread around the Near East with the Muslim conquests, which would make sense actually, because it seems that the religious isolates in the Near East have the lowest ANE ratios.
Hard to say. We'll have to wait for some aDNA results from the Near East and Caucasus, which are surely on the way. But yes, something big happened there after the Neolithic transition, as big as whatever happened in Europe.
Most PCA of West Eurasia with decent sampling strategies and enough markers actually look very similar.
ReplyDeleteThe only time people screw them up is if they just use the HGDP samples and not much else, or they oversample Finns, especially the Kuusamo Finns.
The PCA I've been running here, including the one above, are very much like those that appeared in Lazaridis et al. and the Kostenki14 paper, and that's what we'll mostly be seeing from now on.
Kuusamo Finns aren't really dimension-definers in any of those PCA's except Nelis 2009 and irrelevant to the Mordovian "middle eastern" shift. Saamis are though - even a single one defines a dimension in one of these so oversampling is not an issue - and they, as well as Karelians and Vepsians, should be in PCA's more often than anything. The best situation would be to have non-projected high quality ancient samples define dimensions of a genotyping-based PCA and see if structure remains consistent, which likely will require a pure ancient Middle Easterner and EHG samples.
ReplyDeleteThanks David. Then I guess, its possible that ANE also may have entered Europe also from the Near East/ Caucasus area, from the late Neolithic ?
ReplyDeleteSome of the ANE in Europe is of post-Neolithic West Asian origin, but only in parts of southern and southeastern Europe, because the ratios of Near Eastern ancestry in other parts of Europe are too low.
ReplyDeleteImagine how much total admixture from the Near East/Caucasus would've had to come in to raise the level of ANE from 0% among the descendents of Middle Neolithic farmers to the present-day levels of around 15% or even more.
Europeans today would be mainly mixtures of Mediterranean-like Middle Neolithic farmers and Anatolians/Caucasians. But this isn't true, and there's actually a massive gap between most of Europe and the Near East/Caucasus on West Eurasian PCA like the one above.
Shaikorth,
ReplyDeleteNo, I can't see this PCA shape of West Eurasia changing much with new ancient samples.
Your certainly right about the latter , that's already been established by the autosomal studies of extant populations . But how much of that isn't due to more recent migrations of "Europeans", especially Easter Europeans and 'proto-Slavic" peoples migrating *east* ?
ReplyDeleteBut The relative contributions of the Sth Causasus / upper NE area and steppe region too seen Europeans wouldn't be too discrepant, naturally the former predominant in S/SE Europe the latter in N.
This then helps little in deciding the geographic origins of IE; for even if one region contributed even more pronounced proportion , this still does not demand that the spread of *a language* didn't spread from another direction. Language expansion is far more complex than imagined migration paths .
I don't understand How to use this test
ReplyDeleteHow does it work ? I must really be stupid !!!
That would depend on what's meant with "much". On a PCA of West Eurasia where dimensions are defined by WHG, EHG (or ANE) and AME samples of >20x read depth I don't expect to see Lezgins closest to WHG, but I don't expect it's going to be identical to Lazaridis figure 2 or to a PCA with just one ancient sample. I do think there's going to be a Chuvash, Mordovian and North Russian deviation towards Middle Eastern dimension which we already see in Lazaridis S10.5.
ReplyDeleteMike,
ReplyDeleteMost of Mesolithic Europe is where Loschbour is. Old Europe, created when Near Eastern migrants sort of similar to some present-day Bedouin mixed with Mesolithic Europeans, is where the two Neolithic samples are.
The stream of samples running more or less from Old Europe across northwestern Europe and deep into Russia is the trail left by whoever moved into Old Europe, mostly during the late Neolithic and Bronze Age.
I can't really see the data being interpreted any other way, although I think it is reasonable to assume some movements from Anatolia to parts of southern Europe after the Neolithic, and well into historic times.
So now we have to figure out who left that trail to Old Europe on this PCA. They probably weren't speakers of Uralic though.
Helgenes50,
The test isn't available yet. It will be after Christmas.
David,
ReplyDeleteI wonder if you could send PCA's out to the people that are interested and see where they plot like me in a mail?(like the ones you showed in this article)
Or is that not possible at the moment?
I understand better
ReplyDeleteYes, that's doable. I'll set something up after Christmas.
ReplyDeleteI'm going away tomorrow, so I won't have access to this computer till then.
Looks pretty damn good!
ReplyDeleteI would bet that ANE reached the Levant and South Caucasus around the 6th-5th millennia BCE.
ReplyDeleteMotala12 has too much ANE to be used as a WHG proxy, no?
ReplyDeleteMotala12 has ANE and lower quality than Loschbour. KO1 is lower quality than Loschbour too. In fact, MA-1 is also a low quality sample and the only reason it's used is that it's the only ANE sample.
ReplyDeleteStuttgart's results do suggest Middle Eastern inflation issues though. What exact samples were used in the initial run of this calculator?
David,
ReplyDeleteThis looks great!
After the holidays, could you also post the Fst distances between these components? Thanks in advance.
Looks nice for West Eurasia. The problem from the last test that seemed to emerge from the WHG turning into a ultra-North Atlantic component that was far southwest shifted compared to Loschbour looks solved.
ReplyDeleteThe levels look right and are fairly consistent with Laziridis. The Laziridis paper gave a low bound estimate for Near East of 66% Near East, 34% WHG in Stuttgart, which this more or less finds. And then that other esimates more or less match plugging these levels into Laziridis EEF, ANE and WHG estimates.
Further things I'd be interested in, when you get the chance, or if they are possible.
- FST matrix between components, especially since they all look more meaningful now.
- How well does this test work for populations outside West Eurasia? On this point, if you masked South Asian populations for components other ANE, WHG and ENA, I'd expect that they should pop out at around Component 1 -0.6 and Component 2 at 3.5 (as that's where they sit based on other plots using West Eurasian components only).
- Do the Scandinavian farmers show any enrichment of WHG compared to Stuttgart in this? Probably pretty slight.
- What does MA-1 look like run through this? I think that's important as Loschbour is placing correctly, so does MA-1 look 100% ANE?
- Is it possible to create masked WHG only samples with ( I guess a variation on this may be how you made this work?), and where are they on a European genotype / SNP PCA plot if so?
It looks like a big improvement on the old test, (which already was quite convincing for but with irregularities), in terms of where it places European samples.
The ANE is better in this one, just a hair off from the PCA, as far as BR2. I think He plotted around 11-12%, on the West side of the French. The last test had him around 5 or 6, I believe. So, this is a good improvement. Almost perfect! How does it do for BR1? Is it closer to her 12-13%, instead of 7, on the last one?
ReplyDeleteI found a couple old F4's. It does look like Oetzi and Gok4 are more Eastern shifted than Sardinians. Of course, Sardinians do have some African ancestry, which may kind of negate much or all of the addition Eastern affinities. Although, Oetzi does shift closer to the Balkans than Sardinians. Again, considering the small African ancestry of Sardinians, it could be possible for Oetzi to also be just 4%ANE, but appear more due to less or no African affinities.
ReplyDeletef4(Sardinian,Oetzi;Karitiana,San) = -0.00221783 (Z=-3.06251)
f4(Sardinian,Gok4;Karitiana,San) = -0.00167365 (Z=-1.23616)
Here's the Fst table. Keep in mind though, the East and South Eurasian clusters are probably mixed.
ReplyDeletehttps://drive.google.com/file/d/0B9o3EYTdM8lQVHp4bUYtZkFpMXc/view?usp=sharing
I'll try running some of the low coverage ancient genomes with this test soon, but apart from their low quality, they're also missing lots of markers, so I have no idea what will happen. I doubt MA-1 will get 100% ANE, since one of its sequences that I have is missing half of the markers needed for this, and the other is clearly contaminated.
Motala12 might work reasonably well though. I'm guessing it'll be around 15-20% ANE.
By the way, here's that table from Laz et al. in which they show estimates of Near Eastern ancestry for Stuttgart (page 92 of the supp info). They come up with a range of 55-100%, but they admit that both the low bound and high bound aren't right.
https://drive.google.com/file/d/0B9o3EYTdM8lQZFJaenB2VXdXbVE/view?usp=sharing
It's hard to say which of these estimates is correct, but those achieved with the Han and 5.1% assumed Sub-Saharan admix in the Bedouin are almost exactly what I got: 68.9% vs. 68.61%. These seem to match Stuttgart's PCA position on various plots.
Davidski based on the Reich PCAs including EHG samples you've seen, how much ANE/WHG ancestry do you suspect EHG had?
ReplyDeleteAlso, do you really think MA1 represents an ancient population? I suspect that he represents a real type of ancestry in Eurasia(and America) that is very likely admixed in some way.
Mesolithic and early Neolithic samples are obviously much better proxies for the ancestors of modern people, and so once the EHG genomes are published are you thinking about replacing MA-1 with EHG as an ancestor of Euros? I know that won't work for all Euros, but it's got to be a decent ancestor-proxy for most.
Looks great David. :)
ReplyDeleteKrefter,
ReplyDeleteMA-1 might be mixed in some way, but I didn't use MA-1 to get this ANE cluster. Like last time, I used ANE-like allele frequencies from modern samples, and the impression I have is that it's a real and very robust ancient component.
I don't know what the EHG are exactly, but I suspect they're just a more easterly version of Swedish hunter-gatherers. In other words, they probably have a lot more ANE, and also possibly some ENA.
So I think that essentially the WHG/ANE/Near East model basically explains present-day European genetic structure, and other models will just be variations on the same theme.
David,
ReplyDeleteWhich populations did you take those alleles from?
Those Near Eastern ancestry estimates are what they got when trying to fit Stuttgart as HG + Near Eastern mix using Loschbour to get HG drift. That's a bit different from "West Eurasian" ("pre-Loschbour") + Near Eastern mix used in their successful models which had less contribution from Basal/Near Eastern. Of course if the WHG component in this calculator is something other than Loschbour, such as the WHG-like parts of Lithuanians or Ukrainians etc. the result will reflect that mostly.
ReplyDeleteWhen looking at those Lazaridis NE-estimates in their own context, I'd rely more on the ones based on contrast with She than Han, since She are more distant from both Europeans and MA-1 than Han or Dai and thus likely more "pure" East Eurasian.
I started the process with non-East Asian segments from South American Indians, but the end product is based on allele frequencies from all of these samples...
ReplyDeletehttps://docs.google.com/spreadsheets/d/1v4zYizoWtsoW1MNBN7SUrLf8R62NHPbMRySUJ2J48_Q/edit?usp=sharing
The ANE test methodology is described in this post and the comments section.
http://bga101.blogspot.com.au/2014/09/eurogenes-ane-k7.html
Shaikorth,
ReplyDeleteIf Stuttgart does have more than 70% Near Eastern ancestry, then that means my test underestimates Near Eastern ancestry across the board. But if I somehow managed to crank it up a few per cent, then Loschbour would probably sail off the plot.
"If Stuttgart does have more than 70% Near Eastern ancestry, then that means my test underestimates Near Eastern ancestry across the board. But if I somehow managed to crank it up a few per cent, then Loschbour would probably sail off the plot"
ReplyDeleteWe have a general idea what Basal Eurasian and WHG mix Near eastern is, and it seems no or very few modern Middle easterns have more basal Eurasian than your Near eastern component. I'm very surprised by how strictly pure Middle eastern SW Asians are.
WHG obviously must have varied in ancient Middle easterns, and so Europeans probably have more recent(Neolithic-present) Middle eastern ancestry than what they score in this test.
Based on the wealth of Y DNA and mtDNA from Early farmers in central Europe we know there's no way Stuttgart was ~31% WHG. I would guess something over 10%, but I would be surprised if it was much more.
There are even more variables that could affect these Lazaridis estimates, for example is the African part in Bedouins like Dinka, Ju'hoan or Yoruba (which they could not use in that model) or even a mix? The reduction estimates are based on assumptions of either Dinka or Ju'hoan only.
ReplyDeleteAnyway, Loschbour's movement on the plot one way or another is no biggie, he's outside modern variation anyway and will drag the PCA edge with him. The European-Loschbour dimension is already stretched compared to the genotype-based PCAs like Laz figure 2, and to your West Eurasia PCA with Loschbour though that one had WHG in a different direction.
Thanks David!
ReplyDeleteIf possible, could you post results for HGDP00444 (a Burusho sample, likely to be the most ANE-admixed Eurasian sample we have), and HGDP00244 (a Pashtun sample, very similar to myself, but obviously it shares more markers with your test)?
As always, this would be greatly appreciated. Thanks again.
Shaikorth,
ReplyDeleteThe European-Loschbour dimension isn't stretched.
This PCA is scaled to fit the values of the principal components. My PCA based on genotypes are rotated and stretched to fit geography.
That's all it is. Apart from that, they're identical, and Loschbour's position is too.
Lazaridis' PCA is taller probably because of aesthetic reasons. If they also scaled it to fit the values of the principal components, they'd end up with the same outcome.
Krefter,
ReplyDeleteYou can't accurately estimate WHG with uniparental markers. If we could, we'd have to say that there's no way Baltic Europeans were mostly WHG. But they clearly are.
Also, Neolithic farmers at basically all sites show hunter-gatherer Y-haploroups like I2 and C6, so we know there was considerable admixture into their gene pool as soon as they crossed into Europe. This admixture could have been maintained due to rapid population growth right after the admixture event, even if the hunter-gatherer uniparental markers disappeared over time because of drift and maybe selection.
Just over 10% simply doesn't work, because it doesn't fit with any of the PCA. Stuttgart is obviously more or less 50/50 Loschbour and Bedouin-like. But a decent proportion of the Loschbour-like ancestry comes from the Near East. Hence, this is why Stuttgart can't be described as 50/50 European/Near Eastern.
Farmers had been in central Europe for only 1,000-2,000 years by the time Stuttgart, NE1, etc. were alive. If she was 20% Euro WHG, I would expect more WHG uniparental haplogroups found in her people, because they had less time for drift.
ReplyDeleteMy point overall is that we can only guess how much true Loschbour WHG ancestry Euros have, because we don't know how much UHG Stuttgart's near eastern ancestors had. True Loschbour-WHG in Europe is probably somewhere between your and Laz's WHG scores. Anyways it's surprising how well it survived.
That's great! Very wide selection. Is there a chance you could throw in the Balochi and Brahui?
ReplyDeleteDavidski,
ReplyDeleteI assume that the SNPs your test samples are based on what 23andme reports on.
Is it correct to assume that if a different DNA test company is used that does not test the same SNPs tested by 23andme, then their Eurogenes admixture results may not be as accurate for that individual.
In other words, if their testing company does not report on the SNPs that your test looks for.
kd
ReplyDeleteCurrently, the AncestryDNA files have the best overlap with my tests. 23andMe nosedived after bringing out their latest 600K SNP file.
But the results won't be affected much in the K13 and K15 tests that look at modern ancestry and are based on over 180K SNPs, because even people missing 20% of the markers get accurate oracle outcomes.
Sein,
Burusho HGDP00444 37.17
Pathan HGDP00244 33.53
Krefter,
Arguments like "1,000-2,000 years isn't enough time" aren't very persuasive, because they're not based on any realistic benchmarks. All the results available show very obvious and direct WHG ancestry in Stuttgart. The estimate of around 30% produces a realistic outcome on PCA plots. Estimates much different from this in either direction would not produce realistic outcomes.
Yeah, I forgot about the stretching of the PCA, no issue with Loschbour's general direction then.
ReplyDeleteThe differences to genotype PCA's with a WHG sample aren't huge. In Lazaridis' PCA Bedouins extend considerably further along dimension 2 than Sardinians so it wouldn't be quite the same outcome even if they squeezed it. One could say this is because Loschbour is projected, but it also happens in their unlinked Finestructure PCA (Figure S19.2) which gets the same structure, and Loschbour isn't projected there yet does not define dimension 2 either. The position of Sardinians and Basques also overlaps on Lazaridis PCA's along dimension 2.
Anyway this PCA still gets close to the structure so I won't nitpick about that futher.
Chad,
ReplyDeleteThe Balochi are listed here...
https://docs.google.com/spreadsheet/ccc?key=0Ato3EYTdM8lQdC01RENtMGpRc3FHTGZpSUFKR3hHY1E#gid=0
The Brahui show lower levels of ANE than the Balochi, but they're hard to test because they're very drifted.
I'm not sure where to throw this, but here... "13-60% of the modern horse genome comes from extinct species..."
ReplyDeletehttp://www.sciencedaily.com/releases/2014/12/141215154627.htm
David,
ReplyDeleteWhat is the story behind UAE 1H, 3H, and 14H? That's exceptional levels of ANE, for the region.
@Davidski
ReplyDeleteI just looked at this analysis.
So, you have:
Loschbour (Mesolithic)
ANE 0
South_Eurasian 0
Near_Eastern 0
East_Eurasian 1.92
WHG 96.05
Oceanian 2.02
Stuttgart (Neolithic)
ANE 0
South_Eurasian 0
Near_Eastern 68.61
East_Eurasian 0
WHG 31.38
Oceanian 0
I agree that there is a contribution to Europeans (not present in Stuttgart or Loschbour) that pulls modern Europeans north and east (on the plot) compared to Stuttgart and Loschbour.
However, I don't think it is necessary to assume that this is a process that occured only since the Mesolithic/Neolithic.
I think it is likely that, during the Ice Age, Europeans/North Africans differentiated themselves according to ecozone/fauna. The Northern European component (largely absent from Loschbour and Stuttgart) could be a result of this specialization.
Just because it was absent in Loschbour and Stuttgart doesn't mean that "ANE" had to have arrived in Europe from the Eastern Steppe during the Neolithic.
I can think of a lot of other places where "ANE" might have been and what technocomplex it might have been associated with.
Also, there are some linguistic and genetic reasons why a single wave massive over-run of Europe by PIE during the Neolithic is unlikely. (Not to mention the continuity of the Mesolithic archaeological record across Northern Europe all the way from Scotland to the Volga.)
One other thing I would mention, and it's a little odd that people miss this, is that something could have been pulling La Brana (Spain) and Loschbour into their far "western" position on the PCA. Migration from the East is always discussed, but there's a good change that migration from Northwest Africa could be the source for some of that "south westerness" in La Brana and Loschbour.
So it could be these specific WHGs (represented by La Brana and Loschbour) that have been subjected to long standing genetic flow from the Southwest, rather than the rest of the European population being subjected to a dramatic event that shifted them East.
I think your estimates for WHG are off, there is no way those groups have that much WHG. What you are calling WHG is probably more of a Basal Eurasian and SNPs truly unique to WHGs probably make up at most 5-10%. WHGs were tiny populations that reproduced barely more than replacement and did not expand until being assimilated by later migrants, it is illogical to believe that their genetic impact was so significant compared to the other booming farmer/pastoralist populations.
ReplyDeleteMdgh, he used WHG genomes to create the WHG component. If not WHG what makes for example English and Kurdish different?
ReplyDeletemdgh,
ReplyDeleteIt's very difficult to mix up Basal Eurasian with WHG. It's easier to mix it up with Near Eastern. I'd say you need to do some more reading.
Also, the WHG here is not inflated, because it's the total WHG present in Europe, and not just the excess WHG on top of what Stuttgart had, which is what Lazaridis et al. were estimating.
You seem to be confused by the lack of Stuttgart's EEF component in this test. But that's not a real component; it's a composite of WHG and Near Eastern admixture.
Your estimate of 5-10% WHG for Europe doesn't make any sense. It would not produce coherent results on a PCA.
You have to realize that this is all a very complex puzzle, and all the pieces have to fit. You can't just pluck figures out of the air independently of everything else.
Davidski, I gues your argument holds if we assume (and ultimately prove) that highland West Asia was no ANE-rich to begin with, and had nto become less so due to genetic Semiticization due to Neolithics, 2nd Mill civilizxations, later Arabization, not to mention Turkicization.
ReplyDeleteThat EEF were not ANE rich does not disprove it , either; as they were probably more 'southern Levantine' in nature, as their Bedouin-like picutre suggests.
Thus, there is reason to beleive that there was an "ANE-triangle" between Europe, NW Asia, and northern Eurasia (?) making any deductions as to ultimate soyurce of the apparent-ANE shift in post-Neolithci Europe, still an open question.
Either way, as I have prevously maintained, even if there was an ANE-shift in Europe during the LN/ EB periods, then this does not *a priori* mean that this introduced IE. A smaller genetic shift from elsewhere could have (paradoxially) introduced it, as langauge expansion is far more complex than mere (apparent) genetic patterning.
Anyway, looking at [http://www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.0105090], the 'northeastern' component does not dominate over the 'southwestern' component when looks at iontermediate populations like Nth Italy, French, or Spain.
Pierre,
ReplyDeleteI have several ancient genomes in this analysis, from the Mesolithic, Neolithic, Bronze Age and Iron Age, so you can actually see that this shift to the northeast happened from the Neolithic to the Iron Age.
Mike,
ReplyDeleteYes, these Northeast European-like alleles that seemingly spread from the east after the Neolithic show the lowest impact in Sardinia and nearby parts of southwestern Europe.
But this is of course one of the areas of Europe most distant from the steppe, the generally recognized PIE homeland, and it's where non-Indo-European languages, like Nuragic, survived into historic times. So how does this back up your argument? Doesn't it work against it?
@Davidski,
ReplyDeleteI can see that BR2 (Bronze age) [ANE 9.7%],
Hungarian,
and
Hinxton4 (Iron Age) [15.11%], English, [http://sra.dnanexus.com/studies/ERP003900]
So . . .
9.7% "ANE" in the Bronze age, in Hungary.
Based on this, I'm not convinced that "ANE" couldn't have had one or many reservoirs in some place like Poland or the Baltic, or River basins like the Dnieper or Volga, during the Paleolithic.
Still not convinced of a single wave Steppe expansion into Europe only since the Neolithic.
Depends on your presumptions. The converse would argue against your ideas, that the greater impact of SW Asian in SEE, naturally diluting as one goes north, but still perceptible even in Atlantic Europe. So this genetic diffusion arguement does not support either standpoint.
ReplyDeleteI think the difference in your and my asumptions is that you are ready to automatically accept that East-central Europe was already Indo-European speaking by / during Bronze Age. I beg to differ, or at least entertain the possibility of a much later Indo-Europeanization process. Certianly, this is supported by linguists who date the northern IE hydronyms (from Rhine to Volga) to 1000 BC, not to mention their uniformity ,which futher supports a later date. If appreciable areas of SW and S Europe were still not IE as late as the pre-Roman Era, why not northern Europe also ?
So whatever one makes of genetic shifts and however one explains the advent of (apparently) Kurganoid features appearing in Corded Ware culture, etc, this might have little to do with IE. And this is not to mention the disregard of the fact that central European (eg cremation burials) or Aegean (certain metal ornaments) elements *also* appear in the steppe at exactly the same time, but noone (or at least only a few people) claims an expansion from central Europe or Greece to the steppe (!).
So there was a whole lot of interaction going on, and individual and even groups of individuals moving around in CE Europe and the Pontic region at this time. But to claim precedence of one direction (E to W), is to elevate one set of data to evidence of a migration, and reduce the rest as 'trade' or 'exchange'. The special pleading with such argumentation requitres no elaboration.
No doubt IE began expanding during thie priod (M3), but was still in early phase, and although already expansive, it was not intensive, and not a foregone conclusion. It could have very easily been an abortive epiphenomenon if not for historical, sociolinguistic *chance*
Pierre,
ReplyDeleteMy assumption for now is that ANE spread across much of Europe after the Neolithic from river basins like the Dnieper and Volga, and not from east of the Urals.
ANE might well have been a European component since the Paleolithic, or even originally so. It was present at fairly high ratios in Mesolithic Scandinavians like Motala12 and StoraForvar11.
But the early Neolithic farmers definitely came from the Fertile Crescent in the Near East. It's just that those populations don't exist anymore, because of post-Neolithic migrations into the Near East from Africa, Central Asia and even South Asia.
Mike,
ReplyDeleteUniparental markers also fit with the linguistic, archeological and genomic model of PIE expansions from the steppe.
For instance, R1a-M417 looks like it expanded rapidly during the early Indo-European period, because its STR markers form a star-like cluster, which is also known as the Genghis-Khan effect.
In fact, the downstream R1a-Z645 is dated to the late Neolithic and ancestral to the European and South Asian "twins" R1a-Z282 and R1a-Z93.
So, who brought these markers to Europe and India just as the genomics and archeological cultures of Europe were changing so drastically? What language group do you have in mind, considering that it made such a huge impact, and thus was unlikely to have gone extinct very quickly?
Any ideas?
David:
ReplyDelete* your reply to Pierre. "ANE might well have been a European component since the Paleolithic,"
Something I had pointed out much earlier, also.
" It's just that those populations don't exist anymore, because of post-Neolithic migrations into the Near East ". Also agree, although some a DNA form NE would be nice !
To me;
'For instance, R1a-M417 looks like it expanded rapidly during the early Indo-European period, because its STR markers form a star-like cluster"
Yes, but given the uncertainity in dating, and gaps in knowledge from a DNA, the Late neolithic expansion, and source, remains to be demonstrated, but certainly possible. Moreover, the apparent star-like expansion of Z645 is more likeley a product of expansion of its downstream clades.
"So, who brought these markers to Europe and India"
Possible, even probable, but not definitive. We have not established with certainty the presence/ absence of R1a in central - south Asia before 2000 BC.
'What language group do you have in mind, considering that it made such a huge impact, and thus was unlikely to have gone extinct very quickly"
Who said the expansion of R1a, even if it wen as you predict, had to have been a monolingual phenomenon ?
David, I look forward to having more discussion with you as new data becomes available, as Im sure it will. If not before the NY, have great holidays.
Davidski,
ReplyDelete"My assumption for now is that ANE spread across much of Europe after the Neolithic from river basins like the Dnieper and Volga, and not from east of the Urals . . ."
Surprise! We agree.
So, my beef is essentially not with your model, but with Reich's statement in an October 30th New York Times article:
http://www.nytimes.com/2014/10/30/science/from-ancient-dna-a-clearer-picture-of-europeans-today.html?_r=0
"But most living Europeans also carry genes from a third population, which appears to have arrived more recently. Dr. Reich and his colleagues found the closest match in DNA taken from a 24,000-year-old individual in Siberia, suggesting that the third wave of immigrants hailed from north Eurasia. The ancient Europeans that the scientists studied did not share this North Eurasian DNA. They concluded that this third wave must have moved into Europe after 7,000 years ago."
Also, I think it is meaningless to refer to this Northern European component as being "ancient".
Compared to what?
The "third wave of immigrants" statements makes it sound like the North Eurasians arrived in Western Europe on the Queen Mary in 1950. Goofy and confusing.
The NYT fudged that article. Reich actually did a presentation recently on new results where he said that the source of ANE in much of Europe was probably late Neolithic/Bronze Age European Russia.
ReplyDeletehttp://eurogenes.blogspot.com.au/2014/06/coming-soon-genome-wide-data-from-more.html
It may be interesting to add the the ancestry proportions of one of the Gokhem genomes. Do they have a trace of Ane? How much WHG and Middle Eastern do they have, as compared to other Neolthics?
ReplyDelete@ Pierre - "One other thing I would mention, and it's a little odd that people miss this, is that something could have been pulling La Brana (Spain) and Loschbour into their far "western" position on the PCA. Migration from the East is always discussed, but there's a good change that migration from Northwest Africa could be the source for some of that "south westerness" in La Brana and Loschbour."
ReplyDeletePeople don't necessarily miss it, it's just that most choose to ignore it or deny it.
Well, we have to ignore it, because it's not there.
ReplyDeleteLoschbour and La Brana-1 aren't "southwestern" in the slightest. They're both ridiculously northern.
Davidski,
ReplyDelete"Loschbour and La Brana-1 aren't "southwestern" in the slightest. They're both ridiculously northern."
That's a semantic argument.
There's plenty of archaeological evidence to indicate that hunter gatherers had wide ranges and could have followed migrating animals.
Loschbour and La Brana do tilt "Southwest" on the PCA.
On which PCA do they tilt "Southwest"? Can you post this PCA here?
ReplyDeleteDavidski,
ReplyDelete"The NYT fudged that article. Reich actually did a presentation recently on new results where he said that the source of ANE in much of Europe was probably late Neolithic/Bronze Age European Russia."
Zimmer and Reich know each other. Zimmer has published announcements of Reich's work many times. I highly doubt that his statements were "fudged". Reich is perfectly happy to play along with the storymaking, especially when so much of his work is popularized in the New York Times.
It would be one thing if this storymaking were reasonably accurate, but it is not.
You admit yourself that this Northern "Eurasian" component is probably contiguous with Paleolithic Europe.
The Paleolithic record in places your are talking about, Poland for instance, is very clear.
Reich is a liar, and I'm going to call him one.
Those of you who are playing along are liars as well, not scientists.
"Reich is a liar, and I'm going to call him one.
ReplyDeleteThose of you who are playing along are liars as well, not scientists."
Clearly you have no idea how science works. Was Newton a liar because he published his law of gravity without including a single mention of general relativity? Was Darwin a liar because he didn't mention DNA?
Scientists formulate hypotheses. When further evidence arises, those hypotheses are replaced wirh better ones.
@Michael
ReplyDeleteI don't need a lesson from you on the scientific method.
Reich has allowed himself to be quoted in the New York Times as saying that a major portion of the Northern European population "arrived" in Europe in the last 7000 year.
He's been quoted in other publications saying the same thing.
At a minimum, even Davidski, who's analysis above does appear to hold water, admits that this "Northern European" component is associated with the Dnieper and Volga, and not from east of the Urals.
And he also states that the "ANE" component was present in Mesolithic Scandinavians.
Mesolithic.
David Reich says in the New York Times that the first "arrival" of the ANE in Europe is 7000 years ago.
Against any application of the scientific method, that doesn't hold water.
Also, a close look at Admixture data, as well as several recent papers, will tell you that the three European components split at some time since K14 and MA1. So this Northern Eurasian component is part of a continuum of people that have been "European" since K14. Calling them "immigrants" is a joke.
David Reich is a liar.
"the Caucasus, experienced a significant rush of ANE admixture after early Neolithic farmers left the region for Europe"
ReplyDeletelooks like, Eastern-Asian R group brought ANE and then proceed to western Europe "hinxton"
Davidski do you have ANE K8 scores for Germans, Austrians, Swiss, and SW French?
ReplyDelete@Pierre
ReplyDeleteDid Reich spit in your cereal and rape your puppy, man?
The NYT article says that the new population "arrived in Europe" rather than "arrived in Central and Western Europe", yes. It also says that they came from "Northern Eurasia", which obviously includes, you know, Europe. It's just a newspaper article. Suck it up.
"I don't need a lesson from you on the scientific method."
ReplyDeleteObviously. I looked back at what you've been saying here and I apologize. The three European components clearly split at some time since MA1, and no population "arrived" in Europe in the last 7000 years.
Thank you for your clarity.
Pierre states:
ReplyDelete"he also states that the "ANE" component was present in Mesolithic Scandinavians.
Mesolithic.
David Reich says in the New York Times that the first "arrival" of the ANE in Europe is 7000 years ago."
That is why it would be so interesting to see how the Gokhems would do in list of the ancestry proportions. We are pretty sure - see Skoglund for that - that TBR people picked up more HG than LBK or Oetzi. We know the HG in the neighbourhood carried ANE. If they don't have a trace of it they might have picked up the extra HG somewhere else, if they do they picjed it up in Sweden. (That is, if Gotland HG's actually were the same as mainland Sweden HG's, as everybody seems to be assuming.)
Gokhem2 doesn't show any ANE in this test, while a merged genome of Motala12 and Ajvide58 shows 16%.
ReplyDeleteI actually ran the test again with more hunters and farmers, after merging some files, and managed to push up the level of Near Eastern admix in Stuttgart to 72%.
This is as far as it'll go, but it looks good, because Loschbour now gets 100% WHG, and synthetic samples made from the WHG component cluster very close to Loschbour and La Brana-1 on a PCA based on genotypes.
Hopefully I'll be able to post these results here later today.
To try and get some more understanding about how the levels of the components contribute to position on this graph, I added some simulated modified populations to the datasheet, based on Caucasian populations plus extra ANE and/or WHG and with less Near East (seemed simpler to use them as a basis rather than Volga populations with more of the East Eurasian).
ReplyDeletehttp://i.imgur.com/J83wNJs.jpg
Shaikorth In Lazaridis' PCA Bedouins extend considerably further along dimension 2 than Sardinians so it wouldn't be quite the same outcome even if they squeezed it.
Yeah, Gamba 2014's PCA is also the same way, with the Bedouin extending towards the southwest - http://www.nature.com/ncomms/2014/141021/ncomms6257/images/ncomms6257-f2.jpg. However, the relative position of KO1 to the Sardinians and Bedouin is similar with the Sardinians still fitting a KO1-Bedouin cline point.
Similar plot from Dienekes back in 2011 - http://blogs.discovermagazine.com/gnxp/files/2011/01/MDS1600.png.
Some variation is still riding on those Bedouins and how much they carry African admixture and other admixture. adna from the Near East and Southwest Asia still has some potential to change the form slightly.
And there is some dependence on sampling strategy - I feel like the ideal sampling strategy for Europe at least would probably be a balance between Popres of the sort used by Novembre et al (which is lacking in Northeast Europeans) and Human Origins (which is lacking in Central Europeans). 23andme seems to have a really good European sample via its users (3000 customers is a lot more than in any published PCA in a study) but may be used disproportionately -
http://blog.23andme.com/wp-content/uploads/2011/09/Genes-mirror-geography-for-Europeans.jpg (fairly even with geography, as it should be, apart from too many British).
The FSTs are interesting. There are big differences between the drift from Africa in the clusters and just overall drift.
ReplyDeleteMaybe ANE contributes more strongly to loss of genetic diversity in Eurasian populations than East Eurasian or WHG. That may make some sense in terms of how many of the lowest genetic diversity populations seem to have ANE? Like if Native Americans were a composite of a low diversity ANE plus ENA then went through another bottleneck, that might explain their low genetic diversity better than back when we thought they were just Siberian East Eurasians who'd gone through an almighty bottleneck. Unless it is an artifact of ascertaining ANE here.
To get an f3 statistic like effect, I tried to subtract or add to distances of each column and row to equalize distance from Pygmy.
http://i.imgur.com/QjKVSdW.png
When you do adjust for overall drift from Pygmy though Near Eastern is still quite a bit further from ANE than WHG, and East Eurasion than ANE, so it's still not quite purely Basal Eurasian.
Based on Laziridis, if you combine the estimates of 66% Near Eastern into Stuttgart, remaining 33% WHG and 44% Basal Eurasian into Stuttgart, the rest HG, then the Near Eastern here should be 66% Basal Eurasian (44/66) and 33% WHG. And the relative distances of the drift corrected Near Eastern, WHG and East Eurasian components from one another seem to more or less fit with that.
Capra and the rest of you,
ReplyDeleteYou are professors, some even directors, at major research institutes.
Many of you have been secretly and not so secretly collaborating and often screening out researchers (or potential researchers) that challenge your work.
In some cases, some of you have taken to attempting to seduce (unsuccessfully) informed online members of the public who are curious about prehistory. Clearly, this is meant to silence members of the public from questioning very aspects of your work.
Some of you have smeared (online) members of the public.
It may very well be that there is a population from the "Steppe" that arrived in Europe in the last 7000 years.
However, it is highly unlikely (given your unscrupulous pattern of silencing the competition) that this is the full story.
Often, your victims are women students and scientists.
I know who many of you are and I will be keeping an eye on your online behavior as well as your statements in the press.
Women do talk to each other and it isn't very difficult to show what you have been up to.
Pierre,
ReplyDeleteAre you done now?
No.
ReplyDeleteJust beginning.
Matt,
ReplyDeleteI removed a few of the Bedouin from my Human Origins dataset that showed inflated pairwise IBS sharing and SSA admix, and this prevented the Beoduin cluster from extending as far to the west as in the Laz et al. PCA.
These are the sorts of issues that will skew the results of PCA based on genotypes, but won't affect the shape of a PCA based on ancient ancestry proportions.
I'm not sure where to throw this, but here... "13-60% of the modern horse genome comes from extinct species..."
ReplyDeletehttp://www.sciencedaily.com/releases/2014/12/141215154627.htm
A more detailed version of the above:
Summary:
Two ancient wild horses recovered from russian permafrost (tamyr peninsula) are more diverged from domestic horses than even przewalski. Its not a good candidate for the ancestor or initial domestication. some of its gene pool was absorbed after domestication described as restocking. see below..
https://www.genomeweb.com/genetic-research/ancient-horse-samples-offer-hints-domestication
Matt, whether the original ANE part of Native Americans was low diversity or not, the drift in them since populating of Americans has certainly removed variation and shaped it further. It's really quite similar to how low read depth has made MA-1 more homozygous than Charles II of Spain. We don't know whether this has relevant effects until someone finds a better ANE genome than MA-1.
ReplyDeleteAs for removing bedouin (or other) samples from the set, it doesn't really matter whether they are there or not because the overall PCA shape doesn't change when they define dimension 2 instead of Loschbour. Real less informative dimension of "Bedouin inbreeding" would make the plot look very different, a good example is the Lazaridis' linked mode Finestructure PCA. In the end PCA's and ADMIXTURE results are approximations of genomewide variation and PCA of admixture results is an approximation of an approximation, so it's good that this one looks close to the genotype one.
Looking at the datasheet and the sims, one thing I do think with this is that, although it is basically likely to be correct in its percentages and even the clusters when mapped onto genotype space, which are by far the most important elements and are a big achievement, I do think this PCA may be overestimating the degree to which ANE causes shift south.
ReplyDeleteIf we think about how PCA analysis works, it takes a set of dimensional variables and tries to reduce them to a more parsimonious lower set of dimensions.
This means that if you have three points (1,0,0), (0,1,0) and (0,0,1), then they are turned into the points of a equilateral triangle.
See the following example - http://i.imgur.com/dmnaFhz.png - image from PAST's PCA function.
This is true whether or not we call our dimensions ANE, WHG and Near Eastern or X, Y, Z, or A, B, C.
See a version of the dataset with a 100% ANE and 100% Near Eastern added - http://imgur.com/SA1C39f. I think (geometry is not my strongest point) it's not quite equilateral, but that this is because there other dimensions for the other 5 components, even though they are small compared.
But this thing is that if we translated this into genetic distances, it would imply ANE, Near Eastern and WHG were equidistant just as the points of an equilateral triangle are. And they're not.
What's actually true is that ANE and Near Eastern are both quite a bit closer to WHG than either are to one another. We know this from Laziridis as Near Eastern contains WHG genetics, making WHG and Near Eastern alike directly, while ANE and WHG are both alike more indirectly by lacking Basal Eurasian and descending from the same clade.
To try and imagine how this would play out, imagine shifting an equilateral triangle to one where one point is closer to the other two than they are to one another. The easiest way to imagine this is to imagine the ANE point moving on the diagram, so it keeps the same distance to WHG as it has, and WHG and Near Eastern keep the same distance, while the distance between ANE and Near Eastern increases.
Now as this happens, the more ANE a point has, the more it should follow the shifting position of ANE.
Effectively, translated to the original PCA posted up all the ANE points would move northeast.
This isn't really a problem for the samples we know about as, the levels of ANE are not high enough for this to be a major issue, but might cause strange estimates for the positon of samples we don't.
David, if you're not sure about this, and its possible, try merging up your synthetic genotype samples of ANE and WHG (e.g. 50-50). I think you'll find that they won't shift south that much without a lot of synthetic Near Eastern, even with quite high levels of ANE and that they'll be relatively closer to WHG and more distant from Near Eastern than you would expect for a similar composition on this PCA.
Matt,
ReplyDeleteHave you seen these PCA?
http://imageshack.com/a/img538/2013/ehUOcC.png
http://imageshack.com/a/img540/4412/blmKW4.png
The second is based on a freehand sketch, but makes the same point, which is that ANE looks kind of southeastern, relative to northeast Europeans at least.
Projections are also known to give MA-1 a "southern" position while WHG's stay further north even projected, it's visible in Lazaridis PCA but perhaps better here.
ReplyDeletehttp://3.bp.blogspot.com/-hoBLZ0dZXto/U1luxCudqwI/AAAAAAAAJk4/J3X428YdPzo/s1600/PCA.png
This Human Origins PCA isn't affected by projection bias, but it shows the same thing, probably even more clearly.
ReplyDeletehttps://drive.google.com/file/d/0B9o3EYTdM8lQSGFuWHNpOFlOVEU/edit?usp=sharing
I'm not sure where Matt saw a different result? Maybe on one of my early PCA of MA-1 which was based on a 140K sequence with more homozygous calls than the Human Origins sequence?
Anyway, MA-1 clearly falls along a trajectory between Eastern Europe and the Caucasus, like the ANE line on my biplot above. So ANE admixture is indeed likely to pull Europeans southeast rather than northeast.
Here's a PCA with synthetic ANE, ENF and WHG samples made from the allele frequencies of the latest version of this test, in which Stuttgart scores 72% Near Eastern admixture.
ReplyDeletehttps://drive.google.com/file/d/0B9o3EYTdM8lQc3k1dGdoR01aNk0/view?usp=sharing
This is the best I can do with current sampling of ancient genomes. But I reckon it's pretty damn close. If it wasn't close, then the WHG synthetics wouldn't cluster with Loschbour.
I'll update the post above with the new results later today, if I manage to get online.
What did you make the synthetic WHG, ENF, and ANE out of?
ReplyDeleteLike I just said, the allele frequencies from the ANE, WHG and ENF clusters.
ReplyDeleteBasque have as much WHG as far northern Euros, and that WHG doesn't vary much in Europe as a whole. Also, Gok2 scored 66% WHG in ANE K7. ANE is what makes North Euros significantly more related to WHG than south Euros The main variation is in Near eastern and ANE ancestry.
ReplyDeleteBR1 at a quick glance looks like he could be Basque, with a few more percentage points of ANE.
Non-EEFs coming into northern Europe after the Neolithic didn't have to be something crazy like 70% WHG to explain modern levels, if people like Basque, BR1, and Gok2 were there.
Considering north Euros ANE levals, there's no doubt Yamna made a very big impact, like well over 30%. S Euros(including Basque) also obviously have a big chunk of Yamna ancestry, like well over 20%(assuming all or most ANE is Yamna).
Because of significant Yamna ancestry in all of Europe, it is very important to know Yamna's WHG levels.
I'm starting to question the accuracy of Laz's statement that Samara Yamna can be fit as Armenian/EHG, because their WHG would have to of been around a south Italian level at most, which is way to low.
On the ancient genomes PCA Davidski(link below) has posted, Yamna looks like it's in between where NE Euros and IRI would fit, and so they should have a North Euro-like WHG level.
http://imagizer.imageshack.us/a/img540/4412/blmKW4.png
LNE/EBA looks like NW Euros and CWC looks like NE Euros, suggesting they had alot of WHG. MNE is to southern to be the pop Yamna mixed to to create CWC if Yamna was really 50% Armenian and 50% EHG.
I think the problem is that Laz's models for admixture can't show complex ancestry and forces him to assume CWC was a mix of only Samara Yamna and MNE. There are other elements not shown in Laz's models which must have had lots of WHG.
I don't want to complain too much about the work those guys have done, because where would we be without it now? But unfortunately, there are some serious limitations to their methodology, and they also rely heavily on MA-1, which is an awesome find, but low quality and contaminated.
ReplyDeleteI think lots of high coverage genomes will ultimately solve all the problems. But the test I've got here is producing results beyond my expectations, and I can't wait to run more ancient Europeans with it, like the eastern hunter-gatherers and Yamnaya nomads.
@Krefter
ReplyDelete"I think the problem is that Laz's models for admixture can't show complex ancestry and forces him to assume CWC was a mix of only Samara Yamna and MNE. There are other elements not shown in Laz's models which must have had lots of WHG."
@Davidski
"I don't want to complain too much about the work those guys have done, because where would we be without it now?"
Where would we be now without the Lazarides model?
Probably further along, since the Lazarides models is fundamentally flawed and makes inferences beyond the capability of any known statistical method.
"You don't want to complain too much . . ."
Of course not, especially when they could be reviewing your next paper.
I'm planning to send off a manuscript to PLoS One next year. I guess we'll see how that goes.
ReplyDeletePierre, you shouldn't hate Laz because you dis agree. And I don't understand why you dis agree. To say Europeans could only have 10% WHG ancestry is ignoring the facts.
ReplyDelete@Krefter
ReplyDeleteI didn't say Europeans only have 10% WHG ancestry.
Dick Shrinker didn't say that either. This is what Dick said:
"90% of the population of Western Europe is formed from HGs that were in Europe before the Neolithic"
I mean, what exactly were those "ANEs" doing in Poland, Sweden or the Ukraine, or whatever, during the Paleolithic and Mesolithic? Eating tofu kielbasa?
The "ANEs" and "WHGs" split from the SAME population after K14 and MA1. If "WHG" was a hunter gatherer, then so was "ANE", right up to the Neolithic and in some areas, well into the Neolithic.
Farming didn't start in Poland, or Sweden or Western Russia, as far as I know.
So if you take the "WHGs" and the "ANEs" and put them together, in Western Europe, it comes out to about 90%.
That's what Dick said.
Anyway, I think ANE and EEF are ridiculous terms.
ANE should be Northern European Hunter Gatherers. NEHG.
EEF should be Southern European Hunter Gatherers. SEHG.
And if you don't like that, we can just cut the assumptions and go with SE, WE, and SE.
I admit that that wouldn't play well in the New York Times.
Reich agreed that Yamnaya is not a good fit, but it is all we have. Obviously IR1 is Yamnaya derived, but before the Iron Age, it is likely much more WHG than Yamnaya.
ReplyDeletePierre, La Brana-1 was pretty much identical to Loschbour and he lived in Spain. We also have K01 from Hungary, which is pretty far south.
ReplyDeleteNear eastern ancestry and mtDNA is first found in Europe with samples from cultures who archaeologist say came from the near east, not southern Europe. How in the world does ancient DNA suggest any form of near eastern(EEF whatever you want to call it) ancestry is native to south Europe?
Why should ANE be called North European hunter gatherer, if Swedish hunter gatherers were mostly the same thing as Loschbour?
"So if you take the "WHGs" and the "ANEs" and put them together, in Western Europe, it comes out to about 90%."
What tests are you basing this on? All Euros have a big chunk of near eastern ancestry, which did not exist in Mesolithic Euros.
When I say near eastern I mean the exact same type of ancestry which prevails in the near east today. There's no way any Europeans out there are 90% what was in Europe before the Neolithic.
Lithuanians have the most at just over 70%.
Discussion on this blog is like a broken record. We need to move on.
ReplyDeleteBTW, I'm willing to trade research Files, if anyone wants to.
BR2 looks to be hanging between 12-13%ANE. Maybe, closer to 13.
ReplyDeleteBR2 is 10% ANE. Basques are a bit lower, French a bit higher.
ReplyDeletePierre,
ReplyDeleteEnough with the horseshit.
La Brana-1 is a southern European hunter-gatherer.
EEF are not.
David,
ReplyDeleteSounds exciting! I hope everything goes well. If possible, would you be willing to share with us (in a general, broad-brush sort of way) the ideas/findings that are behind your paper? If not, that's absolutely okay. We'll wait. ;-)
Just a random side note: MA1 is very "southern" in a pan-Eurasian context. When South Asians are included on PCA plots with MA1, and Native Americans+Africans are excluded, MA1 clusters in South Asia. Depending on the plot, he can cluster anywhere from South Central Asia (near a Pashtun sample, or among the Burusho, etc) to South Asia proper (among scheduled caste North Indians, or South Indians). At least that's how the MA1 sample we had a few months ago behaved, not really sure about the "Human Origins" sequence. But I think even a high quality MA1 sequence would behave like this, since David has demonstrated with these tests that whatever ANE happens to be, it's Eurasian peak occurs in South Asia (contrary to what people thought, this isn't due to some confounding admixture in MA1's genetic history, since David's ANE cluster is based on modern Native Americans, and is still heavily skewed towards South Asians, in terms of distribution).
I know you guys are talking "southern" in an intra-European context, not a broadly Eurasian context. But I just wanted to make a note of how MA1 actually clusters in an intra-Eurasian context.
Basques should be about 9.7% average correct? French were around 14%, I believe. What are you getting for the Basque and French, with the new ANE?
ReplyDeleteCO1 and BR1 would be great to see, as well!
ReplyDeleteI think 9.7% for Basques is too high. They're more like 6-7%, which is what this test shows, and that's probably why they can actually be modeled as 0% ANE.
ReplyDeleteThe French in the north are around 14%, while in the south they're 9% or even lower.
Yeah, I suppose that is a possibility. Their main haplotype is more of a founder effect. They probably don't have much Bronze Age ancestry.
ReplyDeleteWhat do you get for CO1 and BR1, on ANE?
Basques kind of look like the Ojibwe of Europe. 80% R1b without changing the genetic and cultural make-up a great deal. Receiving end of traders maybe?
ReplyDeleteThe French average certainly is closer to 12%, it is lower than that of the English. In the new test their average dropped, from the spreadsheet, SE English about 14 % ?
ReplyDeleteOK, I've updated the blog post with the new results. Now Stuttgart is just over 72% Near Eastern, which means everyone else is more Near Eastern too. I think this works, because the synthetic WHG samples basically look like real WHG genomes on the PCA. They're actually closer to La Brana-1 than Loschbour, but whatever.
ReplyDeleteAlso, here are the results for Gokhem2 and a composite of Motala12 and Ajvide58 (I had to merge them to get enough markers).
Gokhem2
ANE 0
South_Eurasian 0
Near_Eastern 49.14
East_Eurasian 0
WHG 49.83
Oceanian 1.03
Pygmy 0
Sub-Saharan 0
SHG_merge
ANE 15.76
South_Eurasian 0
Near_Eastern 0
East_Eurasian 0
WHG 81.94
Oceanian 1.64
Pygmy 0
Sub-Saharan 0.65
I'll post the results for some of the other genomes and other stuff after Christmas.
Seinundseit-carlos-Krefter-DDeden-capra ...-Simon_W-Fanty,
ReplyDeleteDavidski,
I notice several new blog posts in the last hour or so, put up by people who know a lot more about estimating admixture than you do.
You might want to have a look for that.
I doubt that I'm the first person that has noticed a problem with the Lazarides model.
If these people are as clever as you are in interpreting ADMIXTURE then I'll pass.
ReplyDelete@Davidski,
ReplyDeleteNo problem, but they will probably be reviewing your papers at some point and are organizing a major conference in population genetics next year.
But don't trouble yourself about that.
Hi guys, back after a while.
ReplyDelete@ Davidski
Where are the positions of the finnish, balts and poles on the pcas? Could you mark them out?
Are the finnish just below the -1.8 number on the x-axis?
Thanks.
I'm unable to mark them now, but the PC scores are available here.
ReplyDeletehttps://drive.google.com/file/d/0B9o3EYTdM8lQeTNPNEdyTzJkS2c/view?usp=sharing
The Fst table for the revised analysis is here.
https://drive.google.com/file/d/0B9o3EYTdM8lQekNXdWczVTVDdDg/view?usp=sharing
Davidski,
ReplyDeleteAny chance you can let the user define the number of returns on population sharing in Oracle, or increase it to 50. At the present it is set to 20.
Presently, with 20 returns, individuals with genotypes very close to the reference populations are not able to get any idea of what their population sharing looks like at a distance of for example 15 or 20.
Thank you.
Ryukendo, if you want to make a quick comparison, here's how those samples you asked about place in a genotype-based Eurogenes West Eurasian PCA. I marked the location of samples you asked about used in the ADMIXTURE-based PCA and North Russians + Mordvins. For Poles just PL1 (IIRC that's David?) who's in non-southern Polish range to avoid errors, since the Polish ID's aren't exactly the same in PC-score list and the genotype PCA.
ReplyDeletehttp://s23.postimg.org/69244nt3d/PCAloc.jpg
"Loschbour and La Brana do tilt "Southwest" on the PCA."
ReplyDeleteI may be misremembering but wasn't it the other way round i.e. they don't but modern Iberians do? Suggesting a change in the interim (Atlantic megalith imo).
.
"Near eastern ancestry and mtDNA is first found in Europe with samples from cultures who archaeologist say came from the near east, not southern Europe. How in the world does ancient DNA suggest any form of near eastern(EEF whatever you want to call it) ancestry is native to south Europe?"
Isn't EEF made up of two components including one that is WHG-like but not exactly?
(I don't want to knock it as the three way split seems a good first stab at the problem but I think that component being a composite is a source of confusion.)
.
"Receiving end of traders maybe?"
Or miners.
Davidski, do you have an opinon on which version of ANE K8 is better?
ReplyDeleteIn the first one the near eastern scores were in nearly perfect correlation with the EEF scores of Laz. English for example scored around 34% and Stuttgart scored ~68%, making English about 50% EEF.
Now the near eastern scores are higher and the WHG scores are lower according to Laz standards. OF course that one test Laz ran doesn't have to be perfect.
How do you explain the strangely close relation between Near eastern and WHG?
"Stuttgart and NE1 show much higher levels of Near Eastern ancestry than most Europeans, and yet they're amongst the most western samples on the plot."
ReplyDeleteIf "Basal" had a coastal distribution in Europe before farming then farmer "near eastern" might be similar to pre-farming Atlantic coastal "near eastern".
Stuttgart is now 72% Near Eastern, and if so, then it's just over 1/4 WHG. From memory, the English are around 40% Near Eastern now. So if they really are 50% Stuttgart-like EEF, then they should indeed be around 40% Near Eastern.
ReplyDeleteI don't have a calculator handy, but 40 divided by three is 13.33, and times four it's 53.33. That's pretty close to what we want. Or not?
Keep in mind also, that I tested a lot of English samples, while Laz et al. only ran six or eight.
Anyway, the synthetic samples made from the WHG allele frequencies from the last test clustered a bit too far south for my liking, just below La Brana-1. So I think they carried some Near Eastern admixture.
The synthetic WHG samples from this test cluster between Loschbour and La Brana-1, so they look pretty clean, which suggests that Near Eastern ancestry is not being underestimated because the WHG component is pretty clean too.
Maybe it's now being overestimated, but if so, then not by much. It's a fuzzy boundary anyway, because as you know, the Near Eastern component is made up of the so called Basal Eurasian and something WHG-like that is probably ancestral to WHG.
David,
ReplyDeleteThank you very much for your new work. It looks worth all the effort.
When will the new calculator be available at GEDmatch? Also could you post the results for KO1?
KO1 is 100% WHG in this test, but it's a low coverage genome and missing around 50% of the markers, so I don't think that's a reliable result.
ReplyDeleteI won't be able to work on these files until after Christmas, so I don't know when the test will be available at GEDmatch at this stage.
Assuming the result of KO1 is reliable, this should indicate that ANE admixture was non-existent in what is now Hungary during the Mesolithic times. What this suggests is that the Mesolithic- and also Neolithic-era western limits of ANE admixture were somewhere further east, at least that south in Europe.
ReplyDeleteI am supposed to be 52 % EEF, Stuttgart-like.
ReplyDeleteIf Stuttgart is 72 % Near Eastern, in my case I get 37 %
37 % is exactly what I got yesterday by using my K13 average
ie 77,1 % SE English and 22,9 Cantabre
Your last Near Eastern results seem overestimated, that's my point of view.
Seinundseit-carlos-Krefter-DDeden-capra ...-Simon_W-Fanty,
ReplyDelete"In the first one the near eastern scores were in nearly perfect correlation with the EEF scores of Laz. English for example scored around 34% and Stuttgart scored ~68%, making English about 50% EEF."
"Now the near eastern scores are higher and the WHG scores are lower according to Laz standards. OF course that one test Laz ran doesn't have to be perfect."
As Davidski moves the centroid of his statistical weights away from "ANEs" [Northern Europeans], he will necessarily more heavily weight toward "EEFs" [Southern Europeans].
Usually, when I see this kind of sensitivity in a data set, a little alarm bell goes off in my head.
But not for you, or your colleagues, apparently.
Just wondering, if there isn't a clear boundary between "EEFs" and "WHGs", then how is it possible to place a hard boundary, in time/place, between "WHGs" and "ANEs"?
Patterson is going around making statements to the effect that he can pinpoint the exact source of PIE to the Maikop culture and can rule out the Yamnaya culture "because their DNA showed the group had hunter-gatherer ancestry".
http://news.harvard.edu/gazette/story/2014/12/the-surprising-origins-of-europeans/
If a clear boundary between EEF and WHG can't be resolved, then I would really like to know how a clear boundary between the Yamnaya, Maikop, or other groups from the "Steppe" can be resolved.
Any thoughts?
Looking at FST distances something looks strange:
ReplyDeleteWHG - NE = 0.062
WHG - ANE = 0.127
Isn't Basal Eurasian population supposed to have separated at an earlier stage from the branch from where later ANE and WHG diverged? If so, how is it possible that WHG are twice closer to NE than to ANE?
Do you think it's possible that WHG evolved from Basal Eurasian?
"If so, how is it possible that WHG are twice closer to NE than to ANE?"
ReplyDeleteBecause NE contains a component that is very closely related or identical to WHG.
"Because NE contains a component that is very closely related or identical to WHG."
ReplyDeleteAnd how is that possible if Basal Eurasian split earlier than ANE and WHG?
Or, again, do you think it was the other way around? ANE and Basal Eurasian split first, and much later WHG split from Basal Eurasian?
"Because NE contains a component that is very closely related or identical to WHG."
ReplyDeleteTable S14.1 from Laz has results showing Loschbour and La Brana-1 are only slightly closer to Stuttgart than to MA-1. Stuttgart has far more WHG than synthetic Near eastern, so we should expect synthetic WHG to be closer to synthetic ANE.
Table S14.1
https://docs.google.com/spreadsheets/d/1KV2zOTqmhMPFNzJiUYYGSmljLAv_hhGtXEdTqGolTt0/edit#gid=0
What Table S14.1 reveals.
https://docs.google.com/document/d/189cNQ_sX7POFsgOvPtb_liI7Ca8Bga9fHXMZ0nNOlAY/edit?usp=sharing
Table S13.1 from Laz shows a correlation with ANE ancestry and shared drift with Loschbour for near easterns. I guess that could be simply because ANE has no basal Eurasian...
Also, here you can see that using D-Statistics in Table S7 from the recent K-14 paper, closeness to K-14 has the same ANE correlation as closeness to Loschbour for near easterns.
https://docs.google.com/spreadsheets/d/1sGychnMdEMg9UvW5oPTYtWqi7ZOG7p_ZU1uzzXoCm4g/edit?usp=sharing
Karitiana are about as close to K-14 as Adygei. If you go by Laz's model Karitiana are about 56% East Asian and 44% ANE, and Lezgin are about 45-50% Basal Eurasian, 25-30% ANE, and 20-30% WHG, and so that makes sense.
All I just posted suggests MA1, K-14, and Mesolithic Europeans are apart of the same clade, and are closer to each other than to Near eastern.
Seinundseit-carlos-Krefter-DDeden-capra ...-Simon_W-Fanty,
ReplyDelete"All I just posted suggests MA1, K-14, and Mesolithic Europeans are apart of the same clade, and are closer to each other than to Near eastern."
Drift?
Thomas Mailund's post on estimating admixture proportions, including a discussion about the possible effects of drift:
ReplyDeletehttp://www.mailund.dk/index.php/2014/12/17/estimating-admixture-proportions/
Alberto,
ReplyDeleteNear Eastern is a mixed component of the so called Basal Eurasian and something WHG-like.
The fact that these two components mixed at some point to create NE doesn't tell us when they diverged.
And the Fst distances are probably also skewed by recent drift, because the components are sourced from modern samples.
I want to be clear here, about middle eastern populations and what is being called the "EEF" component in the Lazarides model.
ReplyDeleteThere are other components, in addition to the EEF component, that usually appear in Admixture runs of Middle Eastern populations.
However, when referring specifically to the EEF component (Southern European including the Levant, but also to North Africa), that fact that an EEF like component appears in K14 is significant.
Furthermore, the fact that I and J ydna appear in both Europe and the Levant suggests that the EEF component is downstream from K14.
I say suggest, because I would like to say more ancient DNA samples from Eurasia for it to be possible to draw a conclusion with reasonable confidence.
However, I do think that the part of the Lazarides model that proposes a separate early "Basal Eurasian" branch from other paleolithic Eurasians is flawed.
sorry, I meant
ReplyDelete. . . I would like to SEE more ancient DNA ...
Davidski,
ReplyDeleteI thought you had separated the WHG part from the EEF to create this NE component (which should represent the Basal Eurasian). My mistake then.
Obviously the drift is difficult to be accounted accurately, but it would be strange to explain such discrepancies in those FST values just based on drift.
Wouldn't it be possible to get a synthetic Basal Eurasian component also, just like with ANE? Wouldn't that give a better idea than using a NE component which already contains part of WHG?
In any case, I still think that something is not matching.
ReplyDeleteANE - WHG = 0.127
ANE - NE = 0.119
If ANE was closer to WHG than to Basal Eurasian, then it should be closer to WHG than to NE (which is part WHG, part Basal Eurasian). But again this show that ANE is closer to Basal Eurasian than to WHG, which should be the case.
... which SHOULDN'T be the case.
ReplyDeleteHelgenes50,
ReplyDeleteThe figures obtained in the run where Stuttgart is 68% Near Eastern can probably be thought of as the low bound estimates. The figures from the run in which Stuttgart is 72% Near Eastern might be high bound, but my opinion is that the latter are more accurate. In other words, the truth might be somewhere in the middle, but I'd say it's closer to the 72% test.
Also, I suspect that in the near future, when we get ancient genomes from the Near East, the estimates of Near Eastern admixture for Europeans will vary depending on which ancient Near Easterner they're based on. I think ancient samples from closer to Europe will result in higher Near Eastern estimates in Europeans, and those from further away in lower estimates, because the ratio of Basal Eurasian vs. WHG-like ancestry will increase with distance from Europe.
Alberto,
I would need fairly accurate estimates of Basal Eurasian ancestry for a wide variety of ethnic groups, and probably also a Basal Eurasian genome to be able to create a Basal Eurasian component.
Pierre,
Mailund's post is about the finer points of formal mixture testing and really has no direct relevance to what we're discussing here. It certainly doesn't in any way support your impossible claim that early Neolithic farmers in Europe were by and large unmixed descendants of southern European hunter-gatherers, which is contradicted by the ancient genomes we already have available.
Alberto,
ReplyDeleteTwo groups of villagers from Sardinia are sometimes more distant from each other in terms of Fst than Iberians and Russians, or even than some Europeans and Near Easterners.
In other words, Fst is easily skewed by factors like drift, inbreeding and even the number of samples used. So to work out the topology of these ancient components you'd need to build trees with formal statistics.
@Davidski
ReplyDeleteThe matter of drift and the path and timescale over which it occurs is highly relevant to this discussion about admixture. Mailand's post is highly relevant.
"It certainly doesn't in any way support your impossible claim that early Neolithic farmers in Europe were by and large unmixed descendants of southern European hunter-gatherers"
I didn't say that. I said that the Western Europeans are mostly descended from Mesolithic European hunter gatherers. I'm willing to concede that perhaps it is not as high as 90% ("WHG"/Northern European Mesolithic/Southern European Mesolithic).
As I have stated above, the balance between "ANE", "WHG" and "EEF" depends on how you choose the centroid of your statistical weights.
As you move this around, you can get numbers where "ANE" is low and "WHG" and "EEF" are high.
Or you can go the other way and get results where "ANE" and "WHG" are high and "EEF" is low. If you look at weighted Admixture runs, as, for instance in the Rasmussen paper (doi:10.1038/nature13025), Extended Data, Figure 3, at K10 and K11, the contribution to Orcadians, French, Basques, Hungarians, Estonians, Lithuanians, from the Middle East is about 20%.
In Southern Europeans, it is a bit higher, about 30%.
It just depends where you set your weights, or where they are set by the populations you run in Admixture.
The fact that this can be shifted so easily should raise a red flag, but, apparently, for you, it doesn't.
Pierre,
ReplyDeleteIt's not as complicated as you would like it to be.
Loschbour is a pure Western European hunter-gatherer, and it should be very obvious to anyone by now, including you, that modern Western Europeans are not even close to 90% Loschbour-like.
We can argue about the precise figure, which obviously is much lower than 90%, but that's not what you're doing here. You're throwing up red herrings and claiming that they support your ridiculous views.
And Rasmussen's ADMIXTURE run isn't informative about the level of actual Near Eastern admixture in Europeans, because the European-specific clusters in his analysis are partly Near Eastern.
That's probably because he wasn't attempting to distill a pure indigenous European component in the first place.
David,
ReplyDeleteWould you run CO1 if you get a chance, please? Thank you!
David,
ReplyDeleteCould you please define South Eurasian, and the region this component peaks.
Thanks
kd,
ReplyDeleteSouth Eurasian is a very widespread component that peaks in Malaysia at 60%, followed by Cambodia at 50%.
It's hard to say what it represents, because this test is mainly designed to distill ANE and WHG, so as to give Europeans an idea how much post-Mesolithic ancestry from outside of Europe they might have.
David,
ReplyDeleteOn a somewhat unrelated note, are you able to test something for me? I am wondering if the West Asian component can pop in and out of populations. Perhaps, using k13 or 15, cross an Adyghei or any Caucasus population with Stuttgart, and Loschbour. Just to see if it drops by half, or can disappear all together. Also, see if you can create it by mixing your Bedouins with Lithuanians, Estonians, or Finns. This should tell us if it is representative of actual descent or a way of expressing some likelihood.
David,
ReplyDeleteSo with this new test then the Near Eastern component contains some alleles that are either closely related to or are ancestral to the WHG component? If that is so then does that mean that the WHG component of this test is pretty much a pure Mesolithic Western European hunter-gatherer component?
@ Shaikorth
ReplyDeleteThanks!
I'm gonna make a prediction: Yamnaya would place on pca below the arch of East Europeans + Chuvash + Eryza and above caucasus groups, and will be closer to NE Euros than to Caucasus groups.
NE Euros will appear to be IE-ized Finns+Balts, French will appear to be IE-ized Basques; aka you could draw a straight line between yamnaya, the autocthonous relic, and the modern IE-ized pop.
I am holding on to my prediction that the increase in WHG represented by BR1 cannot be a result of IE, and that the ascription of cultures assoc with BR1 to kurganization is dubious. I think it must have been produced by something northwestern assoc with North Sea and Atlantic components in Eurogenes' ADMIXTURE.
Let's see how this turns out.
@ All
Tree-like modeling implies that basal eurasian splits off before East Asian and WHG-ANE do from each other, but by genetic distances Europeans are closer to Middle Easterns today than they are to Dai if we model without any other information except just modern genomes.
The reason why Basal was proposed was not because Europeans were closer to Dai than Middle Easterns, but because Europeans were closer to Dai than Midddle Easterns were. Europeans and Middle Easterns are in fact closer to each other than to Dai.
This discordance between trees and genetic distances seems to me to mean that the degree of Crown Eurasian admixture into the Middle East from the (WHG-ANE) clade must have been very substantial, larger than people often assume a least.
Ryu,
ReplyDelete"I'm gonna make a prediction: Yamnaya would place on pca below the arch of East Europeans + Chuvash + Eryza and above caucasus groups, and will be closer to NE Euros than to Caucasus groups."
Far north Euros(NE have more) have as much WHG as Basque, but far more ANE. The Yamna described by laz probably had 20% WHG(In laz terms) at the most. Assuming Samara Yamna is the only source for ANE in north Euros all are 50%> Yamna.
This leads me to believe there's an element to the copper age genetic shift that was very WHG-like(much more than Basque) Laz didn't describe/reveal, or Samara Yamna isn't the only ANE source for north Euros.
There had to of been people running around 5000-6000YBP who were something over 60% WHG(In laz Terms), if what Laz said is literally true, to explain current levels.
IF you don't already know, Davidski posted this PCA awhile ago. The info originally came from Riech's crew.
http://imagizer.imageshack.us/a/img540/4412/blmKW4.png
Yamnaya is almost in the middle of the triangle formed by Eastern Hunter-Gatherers, Middle Neolithic Europeans and Mal'ta Boy on the plot David recreated from hand drawn notes from ASHG 2014.
ReplyDeleteCloser to the line formed between EHG and MNE than to Ma-1, but still closer to MA-1 than any of the other groups shown.
Chad,
ReplyDeleteCO1 is missing almost 60% of the markers, but comes out 55.7% Near Eastern and 44.29% WHG.
I'll check BR1 and IR1 at some point today or tomorrow if I can.
Jackson,
Yes, I believe the WHG component in this latest version of the test is purely WHG. In fact, as I discussed above, it might even be a very conservative variant of WHG, because the synthetic samples made from it are even more northern than La Brana-1. But I think that's OK, because I'm able to slide things to increase it or decrease it across the board.
rk,
This PCA is based on a freehand sketch from the ASHG 2014, so it probably shouldn't be scrutinized too much, but Yamnaya is shown very close to Corded Ware (CWC), and that's likely to be more or less correct. As we know, CWC clusters in North-Central Europe, so Yamnaya can't be too far east of there.
http://imageshack.com/a/img540/4412/blmKW4.png
Here's a similar PCA for reference from the Kostenki14 paper.
http://imageshack.com/a/img673/9131/528oyd.png
By the way, I don't pay much attention to the Fst distances in these sorts of analyses, because from experience they seem to test more variables than just phylogeny, such as the genetic diversity within each cluster.
However, I also suspect that the Near Eastern cluster is more complex than just being a straight Basal Eurasian/WHG-like mix. It might well be part ANE too, at least in this test, or perhaps part ANE-like, and even hiding somewhat different forms of Basal Eurasian.
@David
ReplyDeletehttp://imageshack.com/a/img673/9131/528oyd.png
Did you notice on this PCA that
the distance between Motala and Ma1 is almost like that of the French average
The amount of ANE is however not the same
Helgenes50,
ReplyDeleteDon't forget PC2. That also has an impact.
But there might be other issues too, because even though this plot seems accurate, MA-1 is way too close to everyone, so it looks like the ancient samples were projected, which means that they don't affect the shape of the PCA. If you're wanting to estimate ancestry proportions from PCA, it's better if all of the samples affect the outcome.
Amazing stuff David thanks.
ReplyDeleteI can't run BR1 and IR1, because they're missing 75% of the markers. But somewhat surprisingly, CO1's result makes perfect sense on the PCA, despite its lack of markers. It's just right (south) of Gokhem2.
ReplyDeleteaqua green = CO1, aqua blue = Gokhem2, light purple = SHG merge (Motala12 + Ajvide58), red = me.
https://drive.google.com/file/d/0B9o3EYTdM8lQV2FrakV5VEJUSDg/view?usp=sharing
@ David
ReplyDeleteDamn thats a pity. BR1's position would be interesting on this one.
@ Krefter
Thanks for the image. I did not see it before this.
I have long suspected the same thing, that something(s) happened to drive up WHG levels independently of the intrusion from the east.
We do know that the Eastern European gene pool was also 'westernized' to some extent during this period of increased migration. My suspicion is that BB drove up WHG and ANE in Western Europe to some degree independently of IE. Also, something that has always bugged me is why Yamnaya ancestry is announced to peak in 'North Central Euros' instead of Eastern Euros, as would be expected; I think one way to reconcile this is to propose that WHG was also driven up in East Europe in recent times by the expansion of the agricultural slavs from the forest-steppe border, which created a large ethnic and historical disruption in the historically nomad-dominated PIE area.
I don't think anyone said that Yamnaya ancestry peaked in North-Central Europe, as opposed to Eastern Europe. What was said was that the Corded Ware genomes landed in North-Central Europe. The Yamnaya genomes are just east of them.
ReplyDelete@ Davidski
ReplyDeleteRazib tweeted that Yamnaya ancestry peaked in North Europe, in his 20 Oct tweets. Which is ambiguous of course, but I'm taking him at face value.
The thing I'm trying to figure out is the increase in WHG phenomenon. East Europeans are distinguished by their tendency towards finnish, and balts especially, and I highly doubt they would increase the affinity to Yamnaya.
To illustrate what I mean, if you look at the
http://imagizer.imageshack.us/a/img540/4412/blmKW4.png
freehand sketch, I think Eastern Europeans will differ from corded ware in the direction of WHG/EHG, because they obviously have less ANE than Yamnaya and I'm quite confident Yamnaya will have more ANE than any Euro pop today, making Eastern Europeans unable to occupy that position. Balts will tip the spear, quite far in the WHG direction I expect. This would make East Euros in general lie away from Yamnaya.
"What was said was that the Corded Ware genomes landed in North-Central Europe. The Yamnaya genomes are just east of them."
ReplyDeleteThen how could Yamna be 50% Armenian and 50% EHG? BR1 and BR2 are probably similar to Bell Beaker and Unetice, just less ANE, and he had alot of WHG. IF Yamna, CWC, Bell Beaker, and Unetice are very close on a PCA, there has to be alot of WHG in all of them.
Either Laz's statement about Yamna is wrong or your drawing of that PCA is wrong. They both can't be right.
Looking back at what Patterson said about Maykop being PIE, because Yamna had hunter gatherer ancestry: This must mean Yamna had significant WHG-type ancestry, because Indians have a ton of ANE-type ancestry.
ReplyDeleteIndians have alot of ANE and near eastern-type ancestry, and if they were 1/3 Yamna and Yamna was 20% WHG, it would be a surprise if they even showed single digit WHG numbers like Greeks do.
"Razib tweeted that Yamnaya ancestry peaked in North Europe, in his 20 Oct tweets. Which is ambiguous of course, but I'm taking him at face value."
ReplyDeleteRazib! Now there's an authority on population genetics. Isn't he a grad student working on the cat genome.
I'm glad you guys have got it all figured out.
@Krefter
ReplyDelete"This leads me to believe there's an element to the copper age genetic shift that was very WHG-like(much more than Basque) Laz didn't describe/reveal, or Samara Yamna isn't the only ANE source for north Euros.
There had to of been people running around 5000-6000YBP who were something over 60% WHG(In laz Terms), if what Laz said is literally true, to explain current levels. "
Dienekes had a link on Swedish Neolithic (paper by Sjoglund, 2012) where alles sharing maps are shown. It shows that Swedes had very few shared allels with TBR Gojhem, but far more with Fotland Pitted Ware. More, for instance, than Russia did. Only Poland shared more. That indicates that there must have been a source of local WHG ancestry after TBR, if I understand it all correctly.
http://dienekes.blogspot.nl/2012/04/ancient-dna-from-neolithic-sweden.html
http://2.bp.blogspot.com/-JXffzwQQek0/T5WDLwnr2fI/AAAAAAAAEyY/TuDjzH5UJI0/s1600/allele_sharing.png
Mind you, it also shows at least one of the Gokhem finds (Gok4) to be relatively far more related to WHG's than the others.
http://3.bp.blogspot.com/-hoBLZ0dZXto/U1luxCudqwI/AAAAAAAAJk4/J3X428YdPzo/s1600/PCA.png
@ryukendo kendow
ReplyDelete"East Europeans are distinguished by their tendency towards finnish, and balts especially, and I highly doubt they would increase the affinity to Yamnaya."
You of all people, should be able to discern the variation between Finns and Balts. Obviously, from a passing glace at a PCA plot, Finns are have more of a Steppe influence than Balts. But you can't differentiate the process by which this occurred, even using current ancient DNA samples.
So, maybe all you guys are picking at straws, at least for now.
This endless conversation, trying to tease out "hunter-gatherers" from "farmers" from "steppe people" looks very cult like to me.
Just saying.
I agree epoch1970, Yamna+Basque probably can't explain current WHG levals. EHG may have had less WHG than some northern Europeans today, and something western could be the source of high WHG today. I doubt Samara Yamna had even 30% WHG(In laz terms), but maybe another type of Yamna had much more and very little near eastern ancestry, and still over 30% ANE. But maybe what Laz said at the conference is flawed in some way.
ReplyDeleteI don't know how the coverage of DNA works but it looks like Gok4 could have as much or more WHG than Lithuanians. Gok2 had around as much as Basque and NW Europeans.
"This endless conversation, trying to tease out "hunter-gatherers" from "farmers" from "steppe people" looks very cult like to me."
ReplyDeleteYou're not arguing, you're picking and twisting what people are posting to discredit what were saying.
We have HG and farmer DNA, we know they were two very different people in stone age Europe.
You're arguing for something that has literally been proven incorrect by Ancient DNA. You ignore the truth.
"This endless conversation, trying to tease out "hunter-gatherers" from "farmers" from "steppe people" looks very cult like to me."
ReplyDeleteYou're not arguing, you're picking and twisting what people are posting to discredit what were saying.
We have HG and farmer DNA, we know they were two very different people in stone age Europe.
You're arguing for something that has literally been proven incorrect by Ancient DNA. You ignore the truth.
@All
ReplyDeleteSuggestion:
Maybe we could move this conversation away the almost religious significance people are attaching to these population components in Laz's model, the WHG, EHG, ANE, EEF and whatever else has been added in the last few months.
Why doesn't someone do a sensitivity analysis, showing how these proportions of these components vary as the centroid of the weighting is moved around. I'd be really interested in that.
@Krefter
ReplyDelete"You're not arguing, you're picking and twisting what people are posting to discredit what were saying."
I feel sorry for you. There's not "farmer" or "hunter-gatherer" DNA. You can go ahead and waste the next ten years on this, even invalidate your own research and that of your colleagues, but others, more qualified, are working on this, and sooner or later, your very simplistic "farmer-hunter/gatherer" is going to go the way of the doo-doo.
But don't expect me to quietly play along.
@Krefter
ReplyDeleteBy the way, do you really think it's appropriate to be impersonating a person of African ancestry when your not of African ancestry?
Yes, something is going on. CO1 should have some ANE. Even just 3-5% would be something.
ReplyDelete@ Pierre
ReplyDeleteHe was a grad student. He was also actually there at the conference.
@ Krefter
"but maybe another type of Yamna"
I'm somewhat dubious of this. The PIE were once a single dialect area, and it seems quite unlikely that continent-sized differences in genetic composition could have sustained themselves in that pop for any length of time.
@ Davidski
Went through some of the older posts. The central Asian paper once again places Papuans outside of a clade of West+East Eurasian, with a contribution from East Eurasian to Papuan.
It seems like this is the default position papuan is placed in whenever Denisovan is not present in the test. This is weird to me as Denisovan ancestry should not bias Papuans away or towards groups without Denisovan ancestry, which is basically anyone else. The Bedouin don't change position when the Basal arrow from the African stem is removed, for example.
The previous time we tried doing stuff using treemix, but now after the whole K14 Ust-Ishim shenanigans I think D-stats might be much more productive in this regard. At least, the authors of the paper on K14 used it to propose that Papuans have something from outside the Crown Eurasian clade, if that's anything to go by.
I might suggest some stuff in a bit. I hope you have the time?
Thanks!
@ryukendo kendow
ReplyDelete"I think D-stats might be much more productive in this regard"
You can use whatever statistics your want, but in the end, you're looking at a lot of noise, and a long diffusion process, with multiple trajectories in time and space.
I very much doubt that any statistics you choose will allow you to tease out "farmers" from "hunter-gatherers".
Davidski: However, I also suspect that the Near Eastern cluster is more complex than just being a straight Basal Eurasian/WHG-like mix. It might well be part ANE too, at least in this test, or perhaps part ANE-like, and even hiding somewhat different forms of Basal Eurasian.
ReplyDeleteThis does seem like it might be possible.
In the absence of adna, which would still be needed for a definite check, using West Eurasian segments from East African genomes, as per the African Genomic Variation Project and Pickerell 2014, might provide us with a truer ancient Near Eastern.
It's perhaps crude but when looking at ADMIXTURE, East Africans often have more of a component which seems tight with other West Eurasian components but further away from Caucasus peoples in particular than Arabic groups should be based on the same ADMIXTURE.
Likewise, on ADMIXTUREs once the African elements are dispensed with, North Africans (Moroccan, Algerian) seem to often look like a mix between Sardinian and a population that is less Caucasus like than the Bedouin are.
rk,
ReplyDeleteSince Yamnaya is fitted as Armenian + Karelian (EHG) it could be that West Asian can be used to mark Yamnaya style "steppe" influence in Europeans. It would fit with the IE-explanation too because the West Asian follows IE (and also is present in Volga-Ural region) and the populations where it is absent tend to be Baltic Finns and Basques. This would obviously not mean that more West Asian in ADMIXTURE means more Yamnaya (or proto-IE) influence because Yamnaya could have less than half the West Asian of Armenians in those calculators if Karelian is divergent enough, and for instance South Slavs, (never mind South Italians), may well be less Yamnaya-like than Poles or Ukrainains.
WHG in Europeans exceeding Basque + Yamnaya combo is in any case probably better explained with WHG survival, especially since WHG increases towards north, than with a WHG-rich version of Yamnaya unmentioned by Reich et al. Corded Ware is northwest of Yamnaya, so it having picked up WHG should not be surprising.
rk,
ReplyDeleteEast Europe is often described as part of North Europe.
For instance, in Skoglund et al. 2013 where the hunter-gatherer genomes shared most alleles with Poles, the conclusion was that they were most similar to North Europeans.
https://drive.google.com/file/d/0B9o3EYTdM8lQVU9GY1FkNnVRaWc/view?usp=sharing
So again, no one ever said that Yamnaya ancestry peaked in North Europe as opposed to East Europe.
Krefter,
Patterson just said that he was collecting DNA to test a theory.
Shaikorth,
Armenians don't show exceedingly high West Asian, like Georgians for instance. They actually have a lot of East Med-type ancestry.
But Yamnaya can probably be modeled as part Armenian, because it has more ANE than the eastern hunter-gatherers can provide alone in a mixture of them and something like Stuttgart.
Or less WHG, more Near Eastern etc. If it was only an ANE question in comparison to Stuttgart, just increasing EHG should make Yamnaya fit with it. Additionally it may be that EHG has something that can't be explained with just WHG's and MA-1. We should know soon enough.
ReplyDeleteBig thanks for the holiday present Davidski. Still can't wrap my head around many of the implications of this data, but here are a few things on my mind…
ReplyDeleteSomething’s fishy with the Bedouin and Saudi results? Half of each group appears to have 0% ANE, while the other half exhibits a wide range, from a few percent all the way up to ~10%. What’s going on?
I’m generally surprised by the high levels of ANE across much of the Middle East. To me it looks like there had to be a singular, consistent, and widespread introduction of ANE ancestry. The Yemeni Jews have 0% ANE, while all the Levantine populations have about ~5-10%. Thus the ANE intrusion must have happened sometime after the Yemenis left — not *that* long ago. Persians then? Still, even if this intrusive population had very high levels of ANE, to get to some of these levels there had to be lots of mixing, which I’m not sure there is strong evidence for. Admittedly I know next to nothing about the history of this region so perhaps someone more educated that I am could take a stab.
Here's another PCA made from the K8 run, but this one is rotated and stretched to fit geography, just like my PCA of West Eurasia based on genotype data.
ReplyDeletehttps://drive.google.com/file/d/0B9o3EYTdM8lQNEJOdWp5RlRXYWc/view?usp=sharing
So like I said, the two PCA are basically identical, despite the very different methodologies used to create them, which means their shapes are largely dictated by ANE, WHG and Near Eastern ancestry proportions.
There are indeed two different Bedouin groups, clearly separated on both PCA. One of these groups has no ANE, while the other has levels comparable to those seen among most Middle Eastern groups.
The Bedouins with no ANE are labeled BedouinB in the Human Origins dataset, and were used in place of Stuttgart in one of the formal tests in Laz et al., presumably because they lack ANE.
The Saudis can also be divided into two such groups, one with ANE and the other without.
Shaikorth,
ReplyDeleteNo, you can't just increase the contribution of one reference group to get the right result if this reference group doesn't have the correct ratios of the components in question.
What you're saying is something similar to the claim that all we have to do is to increase the contribution from SHG to get the right level of ANE in modern Scandinavians. No, we can't, because their WHG/ANE ratio isn't right, which means that if we increase the ANE contribution then the WHG has to go up as well, and then it's too high.
David,
ReplyDeleteCould you include NE7, just as a pre-CO1 reference? Thanks!
"I feel sorry for you. There's not "farmer" or "hunter-gatherer" DNA."
ReplyDeleteIf there wasn't I'd believe it. I simply say what the evidence says, and I'm open to HG and farmer not being significant in Euro prehistory.
I'm not here to be your enemy.
"By the way, do you really think it's appropriate to be impersonating a person of African ancestry when your not of African ancestry?"
I'm not African and never claimed to be.
Davidski, can you post all the ancient individual's scores?
ReplyDeleteNow it's pretty obvious to me that BR1 and BR2 are basically Gok2 or Gok4, plus something from the east. Assuming Samara Yamna was over 30% ANE, that could make both of them close to 30% Samara Yamna. If Yamna was only 20% something WHG(In laz terms), and BR1+2 are Yamna+Gok2, then their WHG would be way lower than it actually is.
ReplyDeleteSamara Yamna+Gok2 simply doesn't work for any Europeans. Considering CWC's position on that PCA I think Yamna mixed with people who had more WHG than Gok2 in northern Europe.
Lets say "EHG in comparison to Stuttgart" then, although it is very likely that by adding EHG to Stuttgart you can get Yamnaya's ANE level, and other stuff in good enough proportions for the end product to approximate Corded Ware which is close to Yamnaya, while SHG + Stuttgart can't really get modern Scandinavian ANE level without looking full SHG.
ReplyDeleteIn any case difference Armenians make as southern half of Yamnaya isn't just about ANE. Their WHG decreases more than ANE increases compared to Stuttgart or Sardinians, and on top of that Near Eastern also increases. Possible effects of the South Eurasian stuff are something to be accounted for as well.
I can't post results for most of the other ancient genomes because they don't have enough markers. NE7, for instance, is missing 80% of the SNPs. Ust'-Ishim is doable, but not very interesting in this context, because it's too old.
ReplyDeleteBut it's pretty easy to work out what they should all be getting by looking at their PCA and other results.
Krefter,
ReplyDeleteTake a look at this...
https://drive.google.com/file/d/0B9o3EYTdM8lQYlU4VXFJLVFXdXM/view?usp=sharing
Compare it to this...
http://imageshack.com/a/img540/4412/blmKW4.png
Note also that on my plot Yamnaya can be modeled as 50% Armenian and 50% something from just east of SHG.
I think it is quite impossible to expect that Yamnaya would score west of any present-day Europeans, if it had such high ANE. Including Kargopol Russians.
ReplyDeleteAlso, it is by no means clear that a line from Gok to Basque should represent IE. Shouldn't a line from Basque to French, or a line from Finnish to Polish--this one especially, be more accurate?
Shouldn't we include finnish in this plot?
Thanks for showing that Davidski, I never considered the possibility. If you're right, then CWC should be near Poland, Ukriane, and Russia, and Unetice and Bell Beaker should be by NW Euros and central Euros.
ReplyDeleteIf Yamna was like 35% ANE, we should expect them to be far more eastern. Maybe EHG was like 70% WHG. Yamna could have had only slightly more ANE than Balts. I can't believe I didn't consider the possibility EHG weren't much different than SHG.
I'm not saying that's the definitive model. For instance, if we increase the level of Gokhem2 related ancestry in BR2, that would push Yamnaya further east. As a result, it would also push the position of the source of the northern ancestry in Yamnaya further east.
ReplyDeleteThe main points I'm making are:
a) BR1 and BR2 can easily be modeled as part Yamnaya, which fits with archeological data, because the emergence of Mako culture follows a major Yamnaya migration into the Carpathian Basin.
b) The fact that Yamnaya can be modeled as 50% Armenian doesn't preclude it from clustering within the range of modern Eastern Europeans, or even Northeastern Europeans. This shouldn't be surprising though, since Corded Ware genomes can be modeled as 73% Yamnaya, and they cluster in North-Central Europe.
RK
ReplyDeletePerhaps. In the sketch from Reich lab presentation, EHG is a lot more eastern compared to SHG than SHG is compared to WHG.
Also Mordvins, or Eastern Ukrainians if it's more western than Kargopol Russians, are a better proxy than Finns for the Yamnaya-alike in that PCA, and probably generally in ADMIXTURE, given its position compared to Corded Ware which should cluster in Polish or Baltic range.
rk,
ReplyDeleteFinns are included on the plot, and the red line runs from Gokhem2 to LBA_Hungary_BR2, in other words, from the very late Neolithic of Northern Europe to the late Bronze Age of the Carpathian Basin.
By the way, have you seen this map? Check out figure 1.
https://www.academia.edu/5717048/Transition_to_the_Bronze_Age_Issues_of_Continuity_and_Discontinuity_in_the_First_Half_of_the_Third_Millennium_BC_in_the_Carpathian_Basin
CWC could be fit as 36% EHG. I don't think they were ever tested directly with Yamna. If you multiple 36, you get 72, and so I think somehow that could be why people started saying CWC was 73% Yamna.
ReplyDeleteMaybe NE Euros and Scandinavians have extra SHG-EHG type ancestry.
Lazaridis actually said that CWC are better fitted as 73% Yamnaya than 36% EHG and Middle Neolithic from Germany.
ReplyDeleteAlso, there might have been somewhat different Yamnaya-derived groups moving west, some with much more WHG, SHG or EHG type of ancestry. I'm fairly sure that not all of the people from across the vast Yamnaya horizon were clones of the six Samara Yamnaya genomes sequenced for this study.
Exactly. Samara is a long ways from the West Pontic. There are 9 archaeological groups in Yamnaya. There's a good chance that those groups were as much or more WHG than anything else, considering the lack of ANE in Hungarian and Danubian groups. Those western groups would more likely be mixed with types ranging from CO1 and Gok2. Europe could be a mimic for how the steppes were, as far as ydna ranges.
ReplyDelete@Davidski
ReplyDeleteI looked at your little plot of ancient DNA (http://imagizer.imageshack.us/a/img540/4412/blmKW4.png.)
"Lazaridis actually said that CWC are better fitted as 73% Yamnaya than 36% EHG and Middle Neolithic from Germany."
If this is it, I don't see how you can differentiate CWC from Yamnaya. They're too close together.
"I'm fairly sure that not all of the people from across the vast Yamnaya horizon were clones of the six Samara Yamnaya genomes sequenced for this study."
No kidding.
From wiki:
http://en.wikipedia.org/wiki/Yamna_culture
"Pavel Dolukhanov argues that the emergence of the Pit-Grave culture represents a social development of various local Bronze Age cultures, representing "an expression of social stratification and the emergence of chiefdom-type nomadic social structures", which in turn intensified inter-group contacts between essentially heterogeneous social groups."
@ Davidski
ReplyDeleteOk I spotted the finns. Looks like a line from finns to polish would land us in Western Europe lol, so that was wrong.
@ Davidski @ Chad
The point that I constantly make is, there must be some core of autosomal similarity across all early IE genomes; I think its highly unlikely that some groups were composed of A+B, and then another group was composed of C+D, and both were early IE. Far more likely that one group was ABCD, and another CDEF, which gives us CD as the evidence for the ramification of the IE protopop across space, no matter what groups they admixed with locally.
I'm glad you used BR2, because if you used BR1 instead, and if you accept BR1 as IE-ized--which, for the reasons I gave above, I don't--then you would get a line landing squarely in WHG territory.
As it is the pca fits reasonably well within my expectations, below E.Euros and above Caucasus, just like I said. I would bet that the results when they come out will show that Yamnaya is somewhat further to the south and east, which would leave a bit more space for WHG survival in places like Estonia, Russia or Finland, and fit in ruler-wise with an approx 30-40% contribution to north Europeans in general, if we take Gokhem as the other end of the line.
By the way, I'm gonna say that the fact that Corded Ware genomes center in Central-North Europe, while the Corded Ware horizon is itself centered on Poland--meaning that populations with a more southern and western profile than those today existed in that horizon prior to later developments--probably means that the slavic expansion did SHG-ize Eastern Europe in recent times, pulling east europeans north of the line formed by Gokhem, Corded ware, and Yamnaya.
ReplyDeleteIt'll be interesting to see how Hinxton 4(Iron age Britian) compares to Bell Beaker and Unetice genomes. It looks like Bell Beaker and Unetice is going to cluster around modern British and central Europeans, and so south of Hinxton 4.
ReplyDeleteToday most Europeans are generally in the same cluster with some outliers in France, Baltic, and Balkans. Copper age central Europe is probably where and when this cluster was born.
French have extra EEF ancestry, Balts have extra HG-ancestry, and Balkans have extra near eastern-ancestry.
Why are there so few women on this blog?
ReplyDeleteMuch of the data that you are sharing here will ultimately be published. Yet, there is clearly a pattern of excluding women on this information sharing forum. Thus, you very clearly are deliberately excluding women from your research discussions.
So, I'd like to know why you are sharing data and doing research, which you will likely publish, in a place that is not friendly to women?
Lots of examples of that, by the way. Just a few days ago, Davidski referred to women as "commodities".