search this blog

Thursday, May 5, 2016

On the modern genetic affinities of Ice Age Europeans


Apparently most of them didn't contribute much to our genomes. But for what it's worth, I tested their present-day affinities, and those of several other ancients, with f3 outgroup stats of the form f3(Test,X;Mbuti). For the archaeological contexts and other details about these ancient samples see here.


Not long ago Basques were thought to be by far the best proxies for Upper Paleolithic Europeans; almost a genetic relict from Ice Age Iberia. Clearly, this is not what they are, but they do share relatively high drift with the El Miron lady, which perhaps suggests some genetic continuity in Basque Country for almost 20,000 years. This would still be a remarkable story if confirmed via multiple lines of evidence.

Below are a couple of examples of what can be done with these stats. It's pretty clear that the Vestonice cave men show a preference for Northeast Europeans, in particular Balts, Estonians and Poles. On the other hand, Kostenki14 shares the highest drift with the Dutch, mostly north of the Rhine Dutch. Might that be meaningful in some way?

Are there any other interesting patterns in this data? Please note, however, that it's not possible to compare the stats directly, because they're based on different numbers of markers.

74 comments:

FrankN said...

Nice, Dave! Any possibility to include Afanasievo Gora, as a potentially better ENA/EHG proxy than MA1? Alos, for what its worth, I would like to see Motala in the stats/graphs as well.

FrankN said...

Edit: I have realised that AG3 is already in the Excel sheet. A plot comparing AG3 to MA1 would nevertheless be nice.

Davidski said...

@Frank

I just added the AG3 vs MA1 graph above.

And here are the stats for Motala_HG and also Karelia_HG.

https://drive.google.com/file/d/0B9o3EYTdM8lQX2plSFBoWHYydVU/view?usp=sharing

Ryukendo K said...
This comment has been removed by the author.
Ryukendo K said...
This comment has been removed by the author.
Shaikorth said...

Probably an indicator that the "ANE" cluster they got in some ADMIXTURE tests was actually something else.

Anonymous said...

@Davidski

Th Dutch also pop up rather high in the El Miron list, and Goyet116 list.

The Dutch north of the Rhine, as I mentioned before, are next to Lower Saxony probably the least affected by the Germanic Migrations. I also have a hunch that we find a tad more Funnel Beaker continuity than e.g. Poland, where IIRC there is archeological evidence for violence. I am currently reading an article on the FB to CWC transition in the Netherlands. It states that the transition was done less than a few centuries, probably in less than a few generations. However, FB cultures less associated with megaliths show more signs of continuity. As the Netherlands shows some clear evidence for WHG continuity next to LBK, in this case the local Ertebolla variants Swifterband culture and Vlaardingen culture (of the latter we found a number of fishing nets that are not unlike current day eel fishers use!).

I combine this with the surprising WHG uptick at the end of the neolithic that Haak et al mentions and have a hunch that maybe the North Sea coast, which used to have huge bogs, fenns and estuaries completely unsuitable for cereal growth but which were populated, might have been a slight WHG refuge.

Might be nationalistic bias though, as I am Dutch ;)

Open Genomes said...

An interactive 3-D version of the Ice Age Eurasians World PC Plot

Some interesting findings:

● Oase1 is very close to Ust'-Ishim who are both Y K2a-M2308. They both seem to be very "Austroasiatic" / ASI.
● The earliest Europeans, the Kostenki Aurignacians and the Vestonice cluster, are also somewhat "South Asian", but more like an ancestral ANI population. Kostenki K14 is South Asian mtDNA U2, and Y C1b-F1370 which is found from India to Oceania. Others in this cluster are in an extinct sub-haplogroup of mtDNA M as well as U2. On the Y, this cluster not only has C1a and C1b, but also IJ* and I*. It's as if I & J are also associated with ANI. The post-LGM El Miron cluster is very similar to the earlier Vestonice cluster.
● Mal'ta and Upper Paleolithic Afontova Gora-2 are in a very separate ANE cluster, which leads to Native Americans (not shown). Afontova Gora-2 is especially "Native American-like". While Y P1 is found among the Aeta of the Philippines, MS is found in New Guinea, and R2 and R1b* are found in India, it seems that P (QR) made it's way to northern Siberia very early, in a way that NO did not. We know that the Kets, who speak a language related to Na-Dene, have high percentages of Y Q1a. This Y K2b (MP) "Ket-like" cluster is completely isolated from the "Austroasiatic-like" Y K2a cluster. These would seem to have separated right at the start of settlement of the interior of Eurasia.
● The "Basal Eurasians" (Early Neolithic Farmers) appear to have been totally isolated from the other two early clusters. We know that the Starcevo-Koros skulls have metrics that are nearly identical to the Natufians. They seem to have bypassed any CHG or WHG admixture in Anatolia, unlike some of the northwest Anatolian Early Neolithic Farmers. It would seem that the Starcevo sample could serve as a proxy for the Levantine Hunter-Gatherers.
● The CHGs appear to be a mix of the "Basal Eurasians" and the "Austroasiatic-like" ANI-ASI group.
● The Karelian Hunter-Gatherers are very "ANE-like". The Swedish Hunter-Gatherers are also very "ANE-like", but are more mixed in the direction of the "Basal Eurasians".
● The Villabruna cluster is more similar to the Swedish Hunter-Gatherers than they are to the CHGs. The Villabruna pre-R1b1a-P297 is no different from any of the other Villabruna cluster members. Given that we find early R1a* and R1b1* in Karelia and Samara (including pre-R1b1a1-M478 and possibly a pre-R1b1c-V88), it seems that the earliest European pre-R1b1a-P297 originated somewhere close to the steppes. The El Miron / WHG cluster seems to be a mix of EHG and CHG.
● One Villabruna cluster member, La Brana-1 from Iberia, appears to be admixed with the El Miron people. La Brana-1 was C1a2-V20. It may be that the El Miron people survived longer in the "Franco-Cantabrian refugium" than elsewhere.
● Yamnaya, Afransievo, Poltavka, and some early Corded Ware samples don't appear to have much (if any) CHG admixture. They appear to be a mix of "Basal Eurasians" / Early Neolithic Farmers and EHGs.
● The Remedello culture of the Po Valley in Northern Italy appears to be a very late survival of the "Basal Eurasian" Early Neolitic Farmers of Europe, similar to the Sardinians today.

Open Genomes said...

Perhaps the most surprising thing is that Oase1, who was 1/8 Neanderthal, is so similar to Ust'-Ishim and lies between Ust'-Ishim and the Denisovan-admixed Australians. Perhaps the "Austroasiatic drift" in this direction is the result of additional Neanderthal admixture, after the initial admixture event in the Near East just after "Out of Africa".

We know that East Asians have an additional small percentage of Neanderthal admixture. We also know that Denisovan admixture exists in Eastern India and Tibet. If this additional Neanderthal and Denisovan admixture can account for this "Austroasiatic-Austronesian-East Asian drift", what can explain the strong drift in the direction of ANE? Was this the result of an encounter with an early, isolated AMH population?

Rob said...

Hey OG
Off topic , but do u have the bAM for Bichon by chance ?

a said...

O.G.


Fantastic work.
Thank you, for taking time and effort to provide us with a 3d plot generated from Eurogenes- Ice Age calculations.

Ryukendo K said...
This comment has been removed by the author.
Davidski said...

rk,

Here's the updated sheet with the Papuans. I don't have Australians in this dataset.

https://drive.google.com/file/d/0B9o3EYTdM8lQdHo2ZmlESHJ1QTA/view?usp=sharing

I can't put the ancient samples in the rows because most are missing too many markers relative to the modern samples. So plotting them with the modern samples won't produce coherent results. Even the Anatolians are short of about 25K SNPs, which would make a difference for f3 stats.

Which D-stats were you wanting?

Rami said...

can u run K8 on the samples , curious to see how they look from that perspective. I am not a fan of graphs.

Unknown said...

Blogger Open Genomes said...

Perhaps the most surprising thing is that Oase1, who was 1/8 Neanderthal, is so similar to Ust'-Ishim and lies between Ust'-Ishim and the Denisovan-admixed Australians. Perhaps the "Austroasiatic drift" in this direction is the result of additional Neanderthal admixture, after the initial admixture event in the Near East just after "Out of Africa".

We know that East Asians have an additional small percentage of Neanderthal admixture. We also know that Denisovan admixture exists in Eastern India and Tibet. If this additional Neanderthal and Denisovan admixture can account for this "Austroasiatic-Austronesian-East Asian drift", what can explain the strong drift in the direction of ANE? Was this the result of an encounter with an early, isolated AMH population?


OG,

We have to keep in mind that any drift calculation based on Eurasian ascertained panels, whether it be calculated via formal stats or inferred through ADMIXTURE is bound to be inaccurate if the sample is archaic admixed. This is for the simple reason that Eurasians, have a very different drift history than archaics, or SSA groups for that matter, and any panel ascertained using them is bound to contain monomorphic or ancestral archaic and African bases at Eurasian polymorphic sites. For example an African and a Neanderthal sample will appear to share more drift with each other than say Neanderthal and Eurasian, simply because of agreement at many loci for which Neanderthal and African are monomorphic, rather because of shared drift. If whole genome data were to be used, which captures all archaic and African polymorphic sites, the results would drastically change.

For Satsurbila admixed pops, such as W and S Asians, there would be a longer drift path with ANE, than populations that are purely W Eurasian derived, since the non-Basal portion of Sats shares a longer drift path with ANE than other W Eurasians, although total shared drift in Sats vs W Eurasians would be diminished due to Basal Eurasian input into Sats. Although, as you point out, I do suspect that an encounter with an early AMH pop is very likely for S and E Asians.

Ryukendo K said...
This comment has been removed by the author.
Krefter said...

@Davidski,

Can you add ancient genomes(CHG, LBK, WHG, Karelia_HG, Yamnaya) to your spreadsheet? By add I mean put them I mean test their drift with the Ice Age Euros, not to test moderns drift with them. I want to try to model modern Europeans F3-stats as a mixture of those ancient genomes.

Ryukendo K said...
This comment has been removed by the author.
Ryukendo K said...
This comment has been removed by the author.
Chad said...

Okay. I am about to blow this up with some stats and ideas. First, I wouldn't use Chimp, due to that ancient attraction. There is also the more than double Neandertal ancestry in UP samples to worry about.

First off, in regards to CHG, it can't just be a mix of BE and ANE.
Ust_Ishim Satsurblia Villabruna GoyetQ116-1 -0.0448 -5.618 469034
Ust_Ishim Satsurblia Villabruna Kostenki14 -0.0317 -4.258 587354
Ust_Ishim Satsurblia Villabruna MA1 -0.0097 -1.192 439142
Ust_Ishim Satsurblia Villabruna AfontovaGora3 0.0143 1.470 170274

We also see the Epi-Gravettian CHG is much closer to Euro Epi-Gravettian than UP Gravettian. However, we see the reduction in comparison to Kostenki, who may have some input into CHG as well. Also, CHG is not even close to being very significantly ANE. Underneath the BE is a pretty even mix of Villabruna and AG3.

Below, AG3 is very ANE, and a better source for moderns.
Mbuti AfontovaGora3 Villabruna GoyetQ116-1 -0.0156 -1.953 188801
Mbuti AfontovaGora3 Villabruna MA1 0.1525 17.398 161042
Mbuti AfontovaGora3 Villabruna Kostenki14 -0.0278 -3.527 205374
Mbuti MA1 Karelia_HG AfontovaGora3 0.0920 9.683 164435
Mbuti Han Karelia_HG AfontovaGora3 -0.0101 -1.949 224163
Mbuti Karitiana Karelia_HG AfontovaGora3 0.0219 3.354 224163
Mbuti Karitiana MA1 AfontovaGora3 0.0241 3.553 170325
Mbuti Satsurblia MA1 AfontovaGora3 0.0149 1.603 117071
Mbuti Villabruna Karelia_HG AfontovaGora3 -0.0685 -8.694 213028
Mbuti Villabruna Karelia_HG MA1 -0.0646 -9.367 522097
Mbuti Villabruna MA1 AfontovaGora3 -0.0009 -0.095 161042

Chad said...

Also, the idea that these Epi-paleolithic people were from Europe is likely incorrect. Some of their ancestry may be from AfontovaGora, but a good chunk was native to West Asia and potentially the Balkans.

Mbuti Satsurblia Villabruna GoyetQ116-1 -0.0296 -4.635 407577
Mbuti Satsurblia Villabruna Kostenki14 -0.0243 -4.228 508450
Mbuti Satsurblia Villabruna MA1 -0.0024 -0.373 381469
Mbuti Satsurblia Villabruna AfontovaGora3 0.0095 1.139 150285
Mbuti BedouinB Villabruna GoyetQ116-1 -0.0301 -7.845 592963
Mbuti BedouinB Villabruna Kostenki14 -0.0333 -8.825 735615
Mbuti BedouinB Villabruna MA1 -0.0242 -6.511 551104
Ust_Ishim Satsurblia Villabruna GoyetQ116-1 -0.0448 -5.618 469034
Ust_Ishim Satsurblia Villabruna Kostenki14 -0.0317 -4.258 587354
Ust_Ishim Satsurblia Villabruna MA1 -0.0097 -1.192 439142
Ust_Ishim Satsurblia Villabruna AfontovaGora3 0.0143 1.470 170274
Ust_Ishim BedouinB Villabruna GoyetQ116-1 -0.0450 -7.133 591557
Ust_Ishim BedouinB Villabruna Kostenki14 -0.0373 -6.164 734000
Ust_Ishim BedouinB Villabruna MA1 -0.0301 -4.899 549915
Ust_Ishim BedouinB Villabruna AfontovaGora3 -0.0233 -3.500 218613

Unknown said...

Ryu,

I am not sure if you remember the 3D scatter plots I used to post last year, but I used Origin Pro for them, the PCAs, and hierarchical clusters. There is a bit of a learning curve, but their tech support is good.

I believe the license is around $500 /yr, but if you have a student ID or are an educator, you can get version for only $100 year. I have never seen free ware for this type of purpose. There is some less expensive software, but I think it is a waste of money, because they are very limited in what can be done.

Chad said...

There is also the idea that some of these Epi-paleolithic samples weren't mixed with UP Europeans. I think this is plainly false. They all seem admixed, and some by more direct local sources. Magdalenian mixed Mesolithic samples are closer to the Aurignacian. Balkan and Southern Euros are closer to Gravettians.
Mbuti Villabruna Vestonice16 Ostuni1 0.0133 1.619 237094
Mbuti Villabruna Vestonice16 GoyetQ116-1 -0.0257 -3.463 486767
Mbuti Villabruna GoyetQ116-1 Ostuni1 0.0386 4.704 231046

Even Villabruna looks significantly admixed with Magdalenian ancestry, yet Loschbour is even more significantly Magdalenian than Villabruna. Yet, Villabruna is much closer to Gravettians than GoyetQ. This tale is far from resolved.
Mbuti GoyetQ116-1 Loschbour Villabruna -0.0237 -3.523 586237
Mbuti GoyetQ116-1 Villabruna Motala_HG -0.0072 -1.328 580572
Mbuti GoyetQ116-1 AfontovaGora3 LaBrana1 0.0559 6.733 190547
Mbuti GoyetQ116-1 AfontovaGora3 Villabruna 0.0336 4.078 188801

I think the best place to assume the real admixing population comes from is where the line intersects running from Anatolians through EEF, and from UP samples through WHG. This more 'western' location should be where this population sits. I think this will show that all Epi-paleolithic and Mesolithic Europeans have a decent amount of UP ancestry and likely some minor MA1/AG3 as they're just as close to both the UP and ANE samples. Judging by the f-stats here, there is no way that our Epi-gravettian and Mesolithic samples are just 16% UP. It looks like a decent amount more than that. Just how far WHG shoots off the line from Anatolians to EEF was a hint, for quite some time.

Mbuti Villabruna GoyetQ116-1 MA1 -0.0165 -2.292 431693
Mbuti Villabruna Ostuni1 MA1 -0.0536 -6.349 195544
Mbuti Villabruna MA1 AfontovaGora3 -0.0009 -0.095 161042
Ust_Ishim Villabruna GoyetQ116-1 AfontovaGora3 0.0067 0.713 214488
Ust_Ishim Villabruna Ostuni1 AfontovaGora3 -0.0473 -4.397 113346
Ust_Ishim Villabruna MA1 AfontovaGora3 0.0139 1.363 182363

Chad said...

Excuse the line above about being equal to MA1/AG3 and UP. The stats above show that is incorrect.

Chad said...

I think that here we see some minor ANE being given to post-UP Europeans.


Mbuti AfontovaGora3 GoyetQ116-1 Villabruna 0.0156 1.953 188801
Mbuti AfontovaGora3 Ostuni1 Villabruna 0.0178 1.807 101247
Ust_Ishim AfontovaGora3 GoyetQ116-1 Villabruna 0.0375 4.111 214488
Ust_Ishim AfontovaGora3 Ostuni1 Villabruna 0.0131 1.186 113346

SNP counts of 500k would definitely make these all significant.

Ryukendo K said...
This comment has been removed by the author.
Chad said...

A migration starting around 35-40kya would be possible and explain the appearance of I in the Gravettian. It certainly is a possibility.

On qpAdm I get 30.7% Villabruna 69.3% Kostenki with a chi-square of 3.194 and a tail of .6701

Chad said...

Still, I think the assumption that Villabruna is not significantly admixed with UP ancestry is wrong. If all Gravettian samples are mixed, then we wouldn't expect that to disappear in the Epi-gravettian. I wouldn't be surprised if all WHG are 20%+ UP.

Krefter said...

@Chad,

Thanks for the stats. So, is WHG definitely not very descended from any of our Upper Paleolithic genomes? WHG's ancestors were living somewhere else? Vestonice and Kostinki's people went extinct?

Ryukendo K said...
This comment has been removed by the author.
Anonymous said...


@Chad Rohlfson

"There is also the idea that some of these Epi-paleolithic samples weren't mixed with UP Europeans. I think this is plainly false. They all seem admixed, and some by more direct local sources. Magdalenian mixed Mesolithic samples are closer to the Aurignacian."

"I wouldn't be surprised if all WHG are 20%+ UP."

The paper states something similar. They claim Magdalenian is Aurignacian + something WHG. That something WHG could very well be shared drift. How else could the Magdalenians coming out of the Iberian refuge after the LGM have something shared with the Villabruna group emerging somewhere else? It also makes it clear that the 3rd population adding to WHG can't have been emerging too far east.

The point that confused me in the first place was that they keep using GoyetQ116-1 as marker for Magdalenian admixture in WHG. This is offcourse understandable, but a statement that Loschbour is 15% GoyetQ116-1, and Magdalenian is roughly two thirds GoyetQ116-1 (Fig 4) means *Magdalenian* admixture in Loschbour is roughly one quarter.

Anonymous said...

@Chad

"Mbuti AfontovaGora3 GoyetQ116-1 Villabruna 0.0156 1.953 188801
Mbuti AfontovaGora3 Ostuni1 Villabruna 0.0178 1.807 101247
Ust_Ishim AfontovaGora3 GoyetQ116-1 Villabruna 0.0375 4.111 214488
Ust_Ishim AfontovaGora3 Ostuni1 Villabruna 0.0131 1.186 113346"

You should rerun these with samples that show a clear Asian signal, such as Bichon, as well as with sample that clearly no Asian signal, such as Villabruna (See fig 4).

Anonymous said...

@RK

"To explore this question, why don't we try:
MButi Villabruna1 Villabruna2 X
Where Villabruna1 and 2 are two samples from Villabruna cluster"

Possibly KO1 (high Asian, low Aurignacian) and Villabruna (Low asian, higher Aurignacian)?

Aram said...

OG

Can You expand more on this please.

Yamnaya, Afransievo, Poltavka, and some early Corded Ware samples don't appear to have much (if any) CHG admixture. They appear to be a mix of "Basal Eurasians" / Early Neolithic Farmers and EHGs.

No or little CHG in Yamna/ Poltavka / Afanasievo? This makes the whole thing upside down.

Karl_K said...

@OG

Your 3D graph is very nice, but some of your conclusions are pretty wild. You are making claims quite contrary to what formal stats show, and your ideas are at odds with many of the interpretations of the authors of the current paper under discussion.

I know that these people have put years into the analysis of the data. Why should anyone trust your own interpretation over theirs?

Karl_K said...

@Kurd dgk

"If whole genome data were to be used, which captures all archaic and African polymorphic sites, the results would drastically change."

I think this should be something that everyone must keep in mind for all of this new data. The enriched and sequenced sites were those known to have SNPs in the few ancient fully sequenced genomes, but mostly from sites known to contain SNPs in living people.

The major limitation is that this emphasizes how these ancient people are related to living people. But it misses most of the story of how they are related to each other. Especially if several groups failed to pass on their genes to the current day.

Ryukendo K said...
This comment has been removed by the author.
Shaikorth said...

RK, this might help:

http://eurogenes.blogspot.com/2013/09/twelve-different-views-of-europe-and.html

edit: this maybe as well
http://eurogenes.blogspot.com/2013/12/cluster-analysis-of-west-eurasia-13.html

Chad said...

There is a big problem with making Vestonice as Kostenki with minor Villabruna ancestry. Vestonice is closer to Villabruna than Kostenki, by a near significant margin.

Karl_K said...

@Chad

"by a near significant margin"

Which is what most people say as, "by an insignificant margin"... because it is not significant.

In any case, you are correct in saying that the original statement was a problem.

Chad said...

The Z is over 2. More SNPs would make it significant. I think all post UP samples have some Aurignacian. Gravettian probably has some of this ghost pop that makes up maybe 80% of the Epi-Paleolithic and WHG samples.

Chad said...

I don't know about the Gravettian having any Kostenki either. Villabruna is significantly closer to Kostenki than AG3 too. I'll work on it more tonight.

Karl_K said...

@Chad

"The Z is over 2. More SNPs would make it significant."

This is in your opinion. Without actually having the additional SNPs, or better yet a whole genome, you can't actually be sure.

If you believe that you actually CAN be sure with less data, then you should write a paper. If you can make accurate predictions (as verified by available whole genomes) from only a tiny fraction of the nucleotide positions... that would be awesome.

It would be a nice paper. I believe it is true. But if noone has published it, then it seems like there are too many ifs involved.

Plus... you would probably have to know which SNPs to sample. And that implies that you already knew the result before doing the test.

Ryukendo K said...
This comment has been removed by the author.
Davidski said...

Krefter,

Not sure if this will work, but by all means, have a crack.

https://drive.google.com/file/d/0B9o3EYTdM8lQc1NOZklBajFyTnc/view?usp=sharing

rk,

Like this?

https://drive.google.com/file/d/0B9o3EYTdM8lQYzVkZV9rM21HSVU/view?usp=sharing

Ryukendo K said...
This comment has been removed by the author.
Ryukendo K said...
This comment has been removed by the author.
Ryukendo K said...
This comment has been removed by the author.
Davidski said...

rk,

This is my main dataset with around 1 million SNPs. I think Chad uses the same IDs for all the samples.

https://drive.google.com/file/d/0B9o3EYTdM8lQUkphZHB2cGN5NHc/view?usp=sharing

I also have two other datasets; one with many more modern European groups, and another with many more Asian groups, both based on around 500K SNPs.

I like to run most things with the 1 million SNP dataset, and use the other datasets if I'm missing a modern population.

So if anyone wants any stats, either myself or Chad can probably run them. Just post a text list of what you're wanting with the correct IDs.

Davidski said...

I do have these sorted into clusters in one of my datasets. Except they're not actually labeled as "clusters".

Mbuti Anatolia_Neolithic Villabruna ElMiron -0.0205 -6.98 589228
Mbuti Anatolia_Neolithic Villabruna GoyetQ116-1 -0.0392 -11.459 658130
Mbuti Anatolia_Neolithic Villabruna Kostenki14 -0.0449 -13.697 905480
Mbuti Anatolia_Neolithic Villabruna Vestonice -0.0302 -12.596 736405
Mbuti Anatolia_Neolithic ElMiron GoyetQ116-1 -0.0203 -4.769 479322
Mbuti Anatolia_Neolithic GoyetQ116-1 Kostenki14 -0.0084 -1.949 631187
Mbuti Anatolia_Neolithic GoyetQ116-1 Vestonice 0.0088 2.342 574875
Mbuti Anatolia_Neolithic ElMiron Kostenki14 -0.0278 -6.878 563460
Mbuti Anatolia_Neolithic ElMiron Vestonice -0.0132 -3.989 534060
Mbuti Anatolia_Neolithic Kostenki14 Vestonice 0.0179 4.971 708579
Mbuti Karelia_HG Villabruna ElMiron -0.0434 -9.616 568639
Mbuti Karelia_HG Villabruna GoyetQ116-1 -0.0654 -12.425 625624
Mbuti Karelia_HG Villabruna Kostenki14 -0.0733 -14.719 827456
Mbuti Karelia_HG Villabruna Vestonice -0.0619 -16.221 701000
Mbuti Karelia_HG ElMiron GoyetQ116-1 -0.0296 -4.762 466724
Mbuti Karelia_HG ElMiron Kostenki14 -0.0362 -5.817 543495
Mbuti Karelia_HG ElMiron Vestonice -0.0231 -4.644 520449
Mbuti Karelia_HG GoyetQ116-1 Kostenki14 -0.0099 -1.525 600100
Mbuti Karelia_HG GoyetQ116-1 Vestonice 0.0059 1.067 555217
Mbuti Karelia_HG Kostenki14 Vestonice 0.0153 2.848 674697
Mbuti Karitiana Villabruna ElMiron -0.0089 -2.279 589338
Mbuti Karitiana Villabruna GoyetQ116-1 -0.0127 -2.887 658456
Mbuti Karitiana Villabruna Kostenki14 -0.0132 -3.16 907963
Mbuti Karitiana Villabruna Vestonice -0.0157 -4.594 736700
Mbuti Karitiana ElMiron GoyetQ116-1 -0.0047 -0.912 479333
Mbuti Karitiana ElMiron Kostenki14 -0.0059 -1.181 563521
Mbuti Karitiana ElMiron Vestonice -0.0077 -1.743 534074
Mbuti Karitiana GoyetQ116-1 Kostenki14 -0.0004 -0.081 631360
Mbuti Karitiana GoyetQ116-1 Vestonice -0.0018 -0.369 574904
Mbuti Karitiana Kostenki14 Vestonice -0.0015 -0.35 708737
AfontovaGora3 MA1 Villabruna ElMiron 0.0092 1.116 156191
AfontovaGora3 MA1 Villabruna GoyetQ116-1 0.0225 2.443 164195
AfontovaGora3 MA1 Villabruna Kostenki14 0.02 2.218 182581
AfontovaGora3 MA1 Villabruna Vestonice 0.0138 2.002 178192
AfontovaGora3 MA1 ElMiron GoyetQ116-1 0.0113 0.975 138333
AfontovaGora3 MA1 ElMiron Kostenki14 0.0128 1.152 145183
AfontovaGora3 MA1 ElMiron Vestonice 0.0099 1.035 149426
AfontovaGora3 MA1 GoyetQ116-1 Kostenki14 -0.0075 -0.625 153385
AfontovaGora3 MA1 GoyetQ116-1 Vestonice -0.0082 -0.822 155129
AfontovaGora3 MA1 Kostenki14 Vestonice -0.0079 -0.801 167134

Alberto said...

Thanks to everyone for the test and stats, and for the comments. Still catching up with all and with the paper, but something I've not seen mentioned (might have missed it) is if Hungary_HG (KO1) represents the "pure" WHG, while the rest have some degree of ElMiron cluster admixture that pulls them further away from the Near East?

I'm looking at the stats in the paper (Table S8.2):

D(Mbuti, GoyetQ1161; Villabruna, Kostenki14) Z=-0.5
D(Mbuti, GoyetQ1161; Rochedane, Kostenki14) Z=-1.4
D(Mbuti, GoyetQ1161; Ranchot88, Kostenki14) Z=-2.4
D(Mbuti, GoyetQ1161; Loschbour, Kostenki14) Z=-3.1
D(Mbuti, GoyetQ1161; LaBrana1, Kostenki14) Z=-3.7
D(Mbuti, GoyetQ1161; Hungarian.KO1, Kostenki14) Z=1

And:

D(Mbuti, Villabruna; Stuttgart, Kostenki14) Z=-0.7
D(Mbuti, Loschbour; Stuttgart, Kostenki14) Z=-1.5
D(Mbuti, LaBrana1; Stuttgart, Kostenki14) Z=-0.2
D(Mbuti, Hungarian.KO1; Stuttgart, Kostenki14) Z=-4

One possible caveat of the second set of stats would be if Stuttgart had recent admixture from a group directly related to KO1, which could make that D-stat significant. It could be tested by using Anatolia_Neolithic instead of Stuttgart. But otherwise, since it looks hardly possible that KO1 has some EEF admixture, and it's the one without any sign of ElMiron/Magdalenian admixture, it seems to me that he might be the best representative of the WHGs that entered Europe (from/through Anatolia?) and that moving into Western Europe got some ElMiron cluster (and deeper in time related to GoyetQ116-1) admixture.

Could someone test the following 2 stats to see if they confirm the above ones?

Mbuti Loschbour Anatolia_Neolithic Kostenki14
Mbuti Hungary_HG Anatolia_Neolithic Kostenki14

Chad said...

KO1 is significantly further from Ust Ishim, possibly due to BE/EEF admixture.

Rokus said...

Kostenki14 shares the highest drift with the Dutch, mostly north of the Rhine Dutch. Might that be meaningful in some way?
Sure. It means that Brace was just another one feeling comfortable with being ignorant about the Dutch as a separate ethnic group. Brace (1996) in Cro-Magnon and Qafzeh - Vive La Difference:
The craniofacial form of Cro-Magnon allies it with the living populations of northwestern Europe specifically with the fringes in Scandinavia and England, but not with the European continent.

Everything from the detail of mastoid process form and nuchal muscle attachments to fully "bun-shaped" occiputs demonstrates a continuity from Neanderthal morphology to that visible in the inhabitants of the fringes of western Europe today in Norway, the Faeroe Islands, and England

Apparently the Dutch represent this conservative fringe of western Europe even better, and hence don't represent the European continent such as normally assumed.

Fu et al. (2016): GoyetQ116-1 from presentday Belgium [...] is similar to the ~37,000-year-old Kostenki14
GoyetQ116-1 is chronologically associated with the Aurignacian cultural complex.
From the supplement:
ŸEl Miron fits as a mixture of (mostly) GoyetQ116-1-related ancestry, and a lesser amount of Villabruna-related ancestry.
Loschbour fits as a mixture of (mostly) Villabruna-related ancestry, an a lesser amount of GoyetQ116-1-related ancestry.
Some populations that lived ~35,000 years ago in northwest Europe (i.e. GoyetQ116-1) have an unambiguous link to populations of the El Miron Cluster that lived around 19,000-14,000 BP.Of course, the link might not involve direct descent from a population that lived near Goyet Cave. Instead, it may be mediated by populations related to GoyetQ116-1 that lived elsewhere.

Hence, GoyetQ116-1 must have been intermediate in time and space between a more conservative Kostenki14-element in the north and east, and the Iberian El Miron cluster, while in Loschbour it also survived as a component in WHG.
Apparently GoyetQ116-1 as much as Loschbour south of the Rhine were not fully representative of the Kostenki14-derivates that roamed the lands north of the Rhine. Given Figure S7.2 this may have been even an unadmixed type of Villabruna.

Shaikorth said...

Do we know that Kostenki-types rather than Goyet116-types ever were in the Low Countries though? Kostenki14 is a sample from Russia, Dutch proximity may well be the result of later migrations.

Re: skull morphology, Kostenki didn't really represent that "Cro-Magnon" type (rather he was associated with some tropical populations like Melanesians) so no surprise any modern northern Europeans were not even considered as its possible closest modern relatives.

Alberto said...

@Chad

The possibility of KO1 having EEF admixture looks very unlikely, especially when he was found among farmers. An EEF parent would already make him 50% EEF, while if he had been there for even more generations it would only decrease his WHG %. Besides, there is no sign of him having EEF admixture of any kind. He always appeared as 100% WHG. He is further from Ust-Ishim for some reason than Loschbour et all, but not from Oase1 or Han. Nor further away from MA1 either.

The fact that he's the WHG without (or with less) ElMiron/GoyetQ116-1 admixture seems to explain better his higher affinity to EEFs, I think.

Matt said...

To be honest, I don't really think there's much information to be had in the differences to the Upper Paleolithic groups, really, for recent people.

But... in case anyone was interested:

Clustering with all the f3 stats including the Upper Paleolithics - http://i.imgur.com/M5BSAJQ.png

Clustering with just the f3 stats for Anatolia_Neolithic, AfontovaGora3, Satsurblia_cluster, Oase1, Ust Ishim, Villabruna cluster (e.g. all the groups that are supposed to be in play by the Mesolithic) - http://i.imgur.com/1BsYIFc.png

Very marginal differences (the UP groups just don't contribute much extra information), however some slight differences where the Nordic countries tend to cluster closer to the Balts, when there are all the Upper Paleolithic groups in (though this could just be because there's an increased weight on HG ancestry or something?). Some shuffling elsewhere in the tree as well. Not sure if either is finding a more accurate measure of relatedness.

Ryukendo K said...
This comment has been removed by the author.
bellbeakerblogger said...

David, will we be seeing an updated deep ancestry calculator like on gedmatch?

Grey said...

"Apparently the Dutch represent this conservative fringe of western Europe even better, and hence don't represent the European continent such as normally assumed."

If swamps and mountains make the best defensive terrain and making an assumption about what the Low Countries were like before they were drained that makes sense imo.

Davidski said...

bbb,

I'll try and put something together once I figure out what these new samples represent.

Krefter said...

@Everyone,

Don't take the highest ranking moderns in shared drift with Ice age genomes. A few decimals doesn't mean anything. The vast majority of European's ancestry comes from the Villabruna, Mal'ta, Anatolia Neolithic, and Kotais clusters. There's no significant Goyet Q116 or Kostinki ancestry.

@David, Chad, Tobus.

It's interesting WHG's closest Upper Paleolithic relative is Ostuni1 who lived in the Southern tip of Italy 27,000 years ago. There's another South Italian genome named Paglicci71 who's about 19,000 years old.

She had mtDNA U5b2b which is typical of WHGs. I know she's of bad coverage, but maybe she is our oldest WHG(or the primary component in WHG) and this can be confirmed with a just a few thousand SNPs.

German Dziebel said...

@Davidski

"Not long ago Basques were thought to be by far the best proxies for Upper Paleolithic Europeans; almost a genetic relict from Ice Age Iberia. Clearly, this is not what they are, but they do share relatively high drift with the El Miron lady, which perhaps suggests some genetic continuity in Basque Country for almost 20,000 years."

This makes them the best proxy for UP Europeans.

On a separate note, your first PCA shows that Oase (and less so Ust-Ishim) are closest to ANE. This is another good reason to believe in out-of-America. Thanks, soldier! Good work.

Krefter said...

@German,

"On a separate note, your first PCA shows that Oase (and less so Ust-Ishim) are closest to ANE. This is another good reason to believe in out-of-America. Thanks, soldier! Good work."

How do you explain the diversity of mtDNA and Y DNA in Africa? How do you explain all non-Africans descending from a single population that lived far less than 100,000 years ago?

IMO, the most basic human functions are adapted to whatever habitat we evolved in in Africa. Our bodies for example can't survive in freezing temperatures without clothing and buildings. If we evolved as far North as America, we wouldn't have the bodies we do.

Grey said...

Krefter

"There's no significant Goyet Q116 or Kostinki ancestry. "

Yeah but how much is significant - especially at the time?

The Tibetan high altitude gene is a tiny part of the whole genome but the difference between survival and not.

Grey said...

Krefter

"Our bodies for example can't survive in freezing temperatures without clothing and buildings."

Or maybe some could once but the ability came with a cost so we lost the ability?

https://en.wikipedia.org/wiki/Yaghan_people#Adaptations_to_climate

"Over time, they had evolved significantly higher metabolisms than average humans, allowing them to generate more internal body heat"

Alberto said...

@RK

Alberto, you might remember that that was the conclusion reached by the paper where Anatolia Neolithic was presented. KO1 was the closest to WHG in EEFs.

What??? KO1 was defined as an EEF with extra WHG admixture? I certainly don't remember reading such thing ever. Could you point me to it?

And we have new data now, from this paper. Just look at Table S8.2. The stats in the form D(Mbuti, Ust-Ishim; X, Kostenki14) where X is a sample from the Villabruna cluster:

Villabruna: Z= 0.6
Rochedane: Z= 1.5
Ranchot88: Z= 2.3
Loschbour: Z= 0.2
LaBrana1: Z= 0.8
Hungarian.KO1: Z= 2.3

So if KO1 has the same score as Ranchot88, who is impossible that he has EEF admixture for obvious reasons. This alone makes the claim based on Ust-Ishim stats very dubious. But then just look at the ones with other samples: With Oase1: no signs of BEA. With Dai: No signs of BEA (quite the contrary). With Han, same as Dai. With Karitiana: no signs of BEA. With Malta1: No signs of BEA.

Plus we have seen literally dozens of Admixture runs with WHG and EEF, and not once KO1 has shown even noise levels of EEF admixture (and Admixture is actually good at this).

So it looks more than bogus to me to suggest hat KO1 has EEF/BEA.

Then looking at the stats I posted above, or from page 75 in the supp.info of the paper:

We observed that Hungarian.KO1 does not show the same affinity to ElMiron and GoyetQ116-1 as do the other Villabruna Cluster (Table S13.2). We were concerned that these results might be an artifact of the fact that the libraries that were used to build Hungarian.KO1 were not UDG-treated. However, this does not appear to explain the observations, as the pattern replicates when restricting to transversion SNPs that are not affected by ancient DNA damage (Table S13.3). These results suggest the possibility that Hungarian.KO1 may have less ElMiron and GoyetQ116-1 affinity than does Villabruna.

If Villabruna types brought increased relatedness to Near East over all previous Europeans (including GoyetQ116-1 and ElMiron cluster), and all Villabruna cluster samples have admixture from them above KO1, then it's expected that KO1 will have increased Near East affinity, as confirmed by the stats. This looks quite straight forward.

Matt said...

Grey Yeah but how much is significant - especially at the time?

The Tibetan high altitude gene is a tiny part of the whole genome but the difference between survival and not.


This is one of those things that's quite interesting for me here, because if the fraction of each previous population is high in the succeeding population, introgression of a selected variant could be quite high all the way back to the Aurignacian.

Like say El Miron are 60% Aurignancian era Goyet, then Mesolithic Europeans are 20% El Miron, then modern Europeans are 10% WHG.

So that would give you something like 1.2% Goyet in modern Europeans. Very little overall ancestry (and less if the WHG ancestry were mainly through KO1 with lower El Miron contribution).

But some gene that was at 90% frequency in the Goyet population, then would enter El Miron at 54%, then rise in El Miron again to 90% frequency, then enter WHG as 18% frequency, get selected to a higher frequency again, and so on to present day people.

There are probably quite few variants that are actually like this though, if any. But there could actually be some adaptive evolution in present day Europeans that goes all the way back to the Upper Paleolithic, even if the contribution from the people living in Europe during the Upper Paleolithic, outside genes is pretty low.

Shaikorth said...

@German

"This makes them the best proxy for UP Europeans."

Only a better proxy than their neighbours, not better than Northeast Europeans or Scandinavians who are even closer or at least equally close to El Miron.

Grey said...

Matt

"But there could actually be some adaptive evolution in present day Europeans that goes all the way back to the Upper Paleolithic, even if the contribution from the people living in Europe during the Upper Paleolithic, outside genes is pretty low."

Yes, that's what I was thinking - tiny as a percentage but maybe not in effect, especially when you think of some of the traits that might have useful to an UP person for example time orientation.

Lukasz said...

Hi. can you make also dendrograms from those data? There sometimes give new insight into dataset.

Lee said...

Looking at the El miron population. It looks likely that they were the original population that expanded with the Magdelenian culture. This was then replaced by the WHG-Villubrana genetic culture ~14K years ago around the time of the Younger dryas.
So we get a expansion from the Franco-Catabrian Region that was later replaced by people from where? Italy/Baltics/Greece? Where did they come from-the WHG?



With El-miron having Near Eastern related ancestry that also fed into WHG- Was this a step wise population expansion? A large population out of the near east/North Africa that stopped off in Italy/Greece and another small subset of the population continuing into Iberia?


Another question-El miron by PCA analysis is fairly different than the WHG that came later. They do look like the Basque, and other of the Atlantic fringe types. Could this genetic group be associated with the later Bell Beakers and/or Megalithic cultures?

Davidski said...

El Miron looks nothing like Basques. You're misinterpreting something you've seen.

Lee said...

Davidski--I am confused by your statement.

Your graph posted in this blog shows a high amount of shared drift with French Basque: "Not long ago Basques were thought to be by far the best proxies for Upper Paleolithic Europeans; almost a genetic relict from Ice Age Iberia. Clearly, this is not what they are, but they do share relatively high drift with the El Miron lady"

Gunther and Jakobsson published a review with a PCA plot including El miron overlaid ontop of modern populations (http://www.sciencedirect.com/science/article/pii/S0959437X16301150)

It is apparent that the Basque have more Southern European ancestry than El-miron--probably reflecting an influx of Neolithic farmers into the mix later on (say ~5-8K BCE). Interestingly the El miron lady is tracking between WHG and modern Orcadians. As the other members of the cluster have less WHG ancestry than El miron lady--they should "pull" further east on the plot closer to other northern europeans.

So with that cluster being paleolithic and closely tracking with modern north Europeans-how much could neolithic populations have contributed to the structure of these populations?

in other words- what proportion of el miron cluster is found in northern europeans? was it replace in whole or part by WHG? only for the mix to be restored by WHG and Neolithic & EHG mixture?

Maternal DNA does suggest a high replacement -on the maternal side at around 14K years ago which matches with the Villabruna influx.

Davidski said...

Your graph posted in this blog shows a high amount of shared drift with French Basque.

Not high, only relatively high. It's actually pretty low.

Gunther and Jakobsson published a review with a PCA plot including El miron overlaid ontop of modern populations.

Their PCA suffers from projection bias, which means that the ancient samples are pulled towards the middle of the plot, skewing their results.

But it's not possible to accurately project ancient samples older than the late UP onto modern variation even after accounting for projection bias. I made this point a while back when running a PCA with the samples from Fu et al.

http://eurogenes.blogspot.com.au/2016/05/for-typical-west-eurasian-pca.html