search this blog

Monday, October 3, 2016

On Gamkrelidze-Ivanov's dubious map skills and then some


I've put together a new D-stats sheet that might be useful in the Indo-European homeland debate (see here). It features new samples from the EGDP dataset, with most of the stats based on over 750K SNPs. The stats are of the form D(Chimp,Ancient)(Mbuti,X).

The idea that Indo-Iranian languages arrived in Central and South Asia from the Armenian Plateau and/or eastern Anatolia during the Bronze Age, as per Gamkrelidze and Ivanov (check out their article here and, if you don't have access, crazy map here), is still popular with a lot of people. But it's most certainly a dud.

There's too much Bronze Age steppe ancestry in this part of the world, particularly among the more isolated Indo-Iranian populations like Pamir Tajiks and the Kalasha, as well as upper caste Indians, to ignore. At the same time, there is no hard data linking any of these groups to Bronze Age Armenia or Anatolia.


Also, some people in the comments here are still having problems comprehending the relationship between Eastern European Hunter-Gatherers (Eastern_HG or EHG) and Yamnaya, and how this relates to the issue of the Indo-European expansion from the Pontic-Caspian Steppe into other parts of Eastern Europe.

By and large the vast majority of EHG ancestry among present-day Europeans was mediated via Yamnaya or closely related groups from the steppe. How do I know? Because of the very close relationship between EHG and Yamnaya signals in present-day Europeans, and specifically Indo-European speaking Europeans, including those living in Eastern Europe.


Indeed, most of the EHG in Eastern Europe arrived there from the steppe as a package with Caucasus Hunter-Gatherer (Caucasus_HG or CHG) ancestry. Again, this is especially true for Indo-European speaking Northern and Eastern Europeans, who are generally hugging the line of best fit in these graphs for that reason. Note also that EHG and Yamnaya produce essentially the same results when plotted against CHG.


On the other hand, groups from far Northeastern Europe, where Uralic languages are currently spoken or were spoken until recently, appear to harbor inflated affinity to EHG and deflated affinity to CHG, putting them well above the line of best fit. These are also the same groups that show inflated East Eurasian, and more specifically Siberian, admixture, which is basically lacking in most Indo-Europeans. So I'm guessing that proto-Uralic speakers were mostly a mixture of EHG and East Eurasian, maybe with minor CHG.

Also worth noting is that Southern Europeans show inflated affinity to CHG and deflated affinity to EHG, putting them well below the line of best fit. This is because of their inflated Near Eastern ancestry (relative to Northern and Eastern Europeans) from the Neolithic and later periods, probably including CHG admixture that arrived in Southern Europe independently of EHG and Yamnaya/Yamnaya-related groups.

However, the ubiquitous presence of CHG across Europe today does not confirm Gamkrelidze and Ivanov's Armenian Plateau Indo-European homeland theory. That's because, unlike what their wacky map that I linked to above suggests, Eastern Europeans do not show any signs of ancestry from the South Caspian region (see here), which they clearly should if their Indo-European ancestors migrated en masse from Transcaucasia to the Pontic-Caspian steppe in an anti-clockwise direction around the Caspian Sea.


Rather, it appears that CHG gene flow diffused onto the Eastern European steppe with migrants coming directly from the Caucasus (see here). And, judging by the affinities of CHG and ancient groups in large part of CHG origin, these people were more likely the speakers of Caucasian languages than of Proto-Indo-European.

See also...

Hurrians and the others

152 comments:

Antoni Małkowski said...

Bardzo pouczające. Awar jest w drzewie Pana Sergey Malyshev i jest Z2103

MaxT said...

I agree, it's laughable. EHG & ANE were both intrusive to West Asia aka R1's expansion from the steppes. On top of that, Basal Eurasian ancestry in Yamnaya and Afansievo did not exceed above 14%, which likely came from CHG. we can be sure that PIE did not originate among Basal-rich populations of West Asia.

Ryan said...

"On top of that, Basal Eurasian ancestry in Yamnaya and Afansievo did not exceed above 14%, which likely came from CHG. we can be sure that PIE did not originate among Basal-rich populations of West Asia."

That seems like faulty logic. Based on linguistic evidence I think that's likely the case, but it's not like there aren't countless examples of groups speaking Indo-European languages with less than 14% of their ancestry coming from Yamnaya and Afansievo.

MaxT said...

@Ryan

"Based on linguistic evidence I think that's likely the case, but it's not like there aren't countless examples of groups speaking Indo-European languages with less than 14% of their ancestry coming from Yamnaya and Afansievo."

It's called Indo-Europeanization.

Early male-dominated conquest by Hittites into West Asia is realistic based on all the genetic evidence we have of on expansion of Indo-Europeans from steppes to across various parts of Eurasia.

PIE did not originate in West Asia, their material culture was not West Asian.

Alberto said...

@MaxT

"EHG & ANE were both intrusive to West Asia"

I think you should check the latest Lazaridis et al. paper (2016) and many other information that you can find in this blog about this. We already have 10.000 y.o samples from West Iran that are 40% EHG (and R2). CHG must be about the same EHG/ANE (and Satsurblia is 14.000 y.o).

No one knows who were the PIE, but most people agree these days that the time frame to look for them is around 4000 BC, not 20.000 BC. So intrusive or not, certainly ANE/EHG was there long before any form of PIE ever existed.

MaxT said...

@Alberto

Gets your facts right from Lazaridis et al (2016), there was NO** EHG in Iran but ANE, and that ANE is still intrusive to Iran because it's from Siberia. ANE did not originate in Iran, West Asia or South Asia or East Asia but in Siberia. EHG is from Russia.

Ariel said...

"So intrusive or not, certainly ANE/EHG was there long before any form of PIE ever existed."

And, assuming that EHG always spoke some form of pre-IE, they (ancient iranians, west asians) could have already came into contact with an IE-like tongue. This fact could explain some of those mysterious similarities between west asian languages and IE. At the same time you need less bronze age steppe admixture to explain the high ANE in India and Iran, and cosidering the high amount of CHG/BASAL in Yamnaya (+40%) we will fover wonder if IE have more west asian influences than is generally thought. On a side note, I still think that India have something to do with ANE. Just think about the basal k7, south asians have the most ANE among moderns, and they have much more than europeans and they are not that far from bronze age steppe level. Also they have very basal ANE related y-dna clades like R2. In addition to that, the raw population of India in the past could have been huge compared to much of west eurasia, they probably move out at some point, the proably had a big impact on the rest of west eurasia, and I don't think that they disappeared (it should have been a genocide). Just think about it, in the current model Indians are not much more than Iran neolithic + some west asian and european influence. It doesn't sound right, it might be something like the broken EEF+WHG model for Europeans that was proposed before the discovery of ANE. Anyway we will see in the future. We need something from pre-neolithic India. It's going to be interesting regardless.

Ariel said...

***I wanted to say ANE instead of EHG and pre-PIE instead of pre-IE

MaxT said...

@Ariel

PIE material culture is not West Asian. Their pottery or burial practices is not west asian. Nor did horse-domestication took place in west asia.

Samuel Andrews said...

@MaxT,

Stone age West Asians were not 100% Basal. Yamnaya being 14% Basal doesn't mean they had little West Asian. They were 40-50% West Asian.

Alberto said...

@MaxT

"EHG & ANE were both intrusive to West Asia"

Alberto:

"We already have 10.000 y.o samples from West Iran that are 40% EHG (and R2). CHG must be about the same EHG/ANE (and Satsurblia is 14.000 y.o). [...] So intrusive or not, certainly ANE/EHG was there long before any form of PIE ever existed.

MaxT:

Gets your facts right from Lazaridis et al (2016), there was NO** EHG in Iran but ANE

Lazaridis et al. 2016 Fig. S4.11:

http://imgur.com/a/ORUa8

MaxT:

ANE did not originate in Iran, West Asia or South Asia or East Asia but in Siberia.

Nick Patterson (co-author of the paper), in this same blog:

A problem to ponder is the deep history of ANE.
Ancient Iran/Caucasus of course has plenty. Does this originate
from NorthEast Eurasia (Malt'a boy) or did it originate very anciently
in Iran/Caucasus/Central Asia and spread North East. I don't know.


http://eurogenes.blogspot.com/2016/08/on-enigmatic-early-neolithic-farmers.html?showComment=1470839180551#c5940631630734749783

MaxT, you've entered with the wrong foot in this place. You've been repeatedly warned by someone well respected here, but you insist. Ignorance is an excuse the first time, maybe even a second time. But not a 3rd, a 4th, a 5th...

Samuel Andrews said...

@Everyone,

Can someone post nMonte results for Saami in these models....

EHG, CHG, WHG, AnatoliaN, Siberian(as many as you can add)
Yamnaya, WHG, SHG, AnatoliaN, Siberian(as many as you can add)
Corded Ware Germany, WHG, SHG, AnatoliaN, Siberian(as many as you can add)
Euro(First test use Lithuanian, 2nd Finnish), WHG, SHG, AnatoliaN, CHG.

Ryan said...

@MaxT

"It's called Indo-Europeanization."

Exactly. Don't assume this couldn't have happened in other eras either.

For example, whichever direction the Dene/Yenisian linguistic link goes, it seems unlikely that the genetic link is greater than ~15% of the autosome.


MaxT said...

@Alberto

Again, get your facts right.

Iran_Neolthic and Iran_Hotu had no EHG, for Iran_Neolthic to have EHG, they need to have WHG to have EHG.

That chart you posted does not** show EHG->Iran_Neolthic.

Here is Eurogenes K7 spreadsheet with Iran_Neolthic and Iran_Hotu.

https://docs.google.com/spreadsheets/d/1tFAa7oxWpcNN-OdMMjBdb4NeWKG7EkpKMzZJVW2_MME/edit#gid=1581589557

EHG was not** present in West Asia. It spread with Indo-Europeans from steppe.

MaxT said...

@Alberto

One more thing -

According to (Lazaridis et al. 2016) EHG = (75% ANE & 25% WHG). No Basal Eurasian.

capra internetensis said...

@MaxT

I concur with Alberto. There are plenty of abrasive people here, but at least they have some content in their posts. Yours are only "rah! simplistic version of mainstream view! rah!" Do you wear a little outfit?

Please try harder to contribute something, if not insightful, at least novel.

Probably PIE did originate from the steppe, but your arguments for this are both weak and uninteresting.

MaxT said...

@capra internetensis

"Probably PIE did originate from the steppe"

They did. Good luck proving that wrong.

Alberto said...

@MaxT

And a 6th, and a 7th...

@Capra

Thanks. David stepped up with an initiative to bring some sanity back to this place. Let's see if we can support it and we get some people to hold a bit themselves with their comments.

Ariel said...

If you say that in Finland there is no IE substratum you are saying that the steppe theory is wrong. Too much steppe ancestry and CHG (Lithuania 6,7%, Saami 6,5%) .

Saami
"Eastern_HG" 50.5
"Nganasan" 21.05
"Anatolia_Neolithic" 18.45
"Satsurblia" 6.55
"Villabruna" 3.45
"Itelmen" 0

Saami
"Yamnaya_Samara" 45.35
"Nganasan" 24.3
"Motala_HG" 21.2
"Anatolia_Neolithic" 9.15
"Villabruna" 0
"Itelmen" 0

Saami
"Corded_Ware_Germany" 61.65
"Nganasan" 24.45
"Motala_HG" 13.9
"Anatolia_Neolithic" 0
"Villabruna" 0
"Itelmen" 0

Saami
"Lithuanian" 68.05
"Nganasan" 15.8
"Itelmen" 9
"Motala_HG" 7.15
"Norwegian" 0
"Anatolia_Neolithic" 0
"Villabruna" 0

Lithuanian
"Eastern_HG" 50.4
"Anatolia_Neolithic" 35.05
"Villabruna" 7.85
"Satsurblia" 6.7
"Nganasan" 0
"Itelmen" 0

Lithuanian
"Yamnaya_Samara" 48.3
"Anatolia_Neolithic" 25.5
"Motala_HG" 24.3
"Nganasan" 1.9
"Eastern_HG" 0
"Villabruna" 0
"Itelmen" 0

(SHG lives matter)

Davidski said...

No one ever said that Finns and many other Uralic speakers didn't have any Yamnaya-related ancestry.

And there was no SHG or EHG in the East Baltic prior to the Corded Ware expansions, just WHG. This is a fact that you will have to get used to one day.

So unless you're proposing a post-Corded Ware colonization of Eastern Europe by SHG, then it's time to move on.

Samuel Andrews said...

Thanks a lot Arial. Saami look like a Lithuanian-like, SHG, Ural-Volga mix. Balts, Scandinavians don't appear to have any SHG. But Finnish, Karelians, and North Russians might have some.

Shaikorth said...

Even assuming Saamis have no IE substrate whatsoever doesn't contradict steppe origins of IE if proto-Uralic population had significant CHG too. This remains to be seen since we don't have Bronze Age forest zone samples yet.

Rob said...
This comment has been removed by the author.
Matt said...

@ Davidski, if that's from the massive Baltic dna project then I guess that's reality and that's the way it is. But it's very weird! SHG like people in the Balkans (per Mathieson), EHG in Russia, and SHG in Scandinavia, then none in the Baltic....

Rob said...
This comment has been removed by the author.
Davidski said...

@Matt

It might seem weird going purely by geography, but it's useful to take a closer look at the origins of the forager cultures in the East Baltic and much of Eastern Europe west of Russia. They're not from Russia and certainly not from Scandinavia.

Shaikorth said...

@Rob
The current theory isn't that Uralic is from Bronze-Age Altai-Sayan but from Volga-Kama, and that it's 1000-2000 year more recent than PIE which is why the contacts are with IA and not PIE itself.

If it's from Bronze Age Altai-Sayan we'll probably know from ancient DNA too, no N1c1 should be in the Volga region never mind further west before Seima-Turbino.

Davidski said...

By the way, you guys need to take some responsibility for keeping the comments section sane and tidy.

I can't just delete comments and ban people on a whim. So if you don't like what someone is saying, don't respond, and things will calm down very quickly.

Rob said...
This comment has been removed by the author.
MaxT said...

@Ariel

High ANE component in Iran and India and can be explained by ANE-related R1 & R2 migrations. ANE is intrusive, not native to these regions.

Shaikorth said...

@Rob

For Bronze Age Altai-Sayan to be a plausible Uralic homeland we should expext Bronze Age N1c1 from thereabouts (and in Europe only after Seima-Turbino) but so far there's been R1, Q, C during the time period. Anthony's new book has the most recent synthesis of linguistics and genetics, and he hasn't moved Proto-Uralic from the forest zone where he put it earlier, pretty much the same as Parpola's "Formation of Indo-European and Uralic language families in the light of archaeology".

Rob said...
This comment has been removed by the author.
Shaikorth said...

It wouldn't change Anthony's linguistic model to move PU right now, since it only requires moving the location of IA contact with it.

Can't say what's his thought process for sure, but the issues with the current genetic data are the lack of N1c1 in the Bronze Age Altai and Mezhovskaya (post-Seima-Turbino Bronze Age) Y-DNA being R1a/R1b. These theories will obviously be amended once there is more data.

Rob said...

Actually, most Altain LN-EBA haplogroups are N and Q.
I'm sure some N1c will show up between the Altai and Vilga, as it already in China.

I guess we can just wait for the data to roll in.

In the meantime, we should probably pay attention to more correct interpretation of Eneolithic archaeology , based on factual interpretation rather than forced narratives

Shaikorth said...

I was talking about Bronze Age specifically, since that's the relevant time period http://www.ancestraljourneys.org/copperbronzeagedna.shtml
There was possible N in Baikal but that was not N1c (tested for Tat), more likely N1b.

Dunno about what's the proper interpretation of archaeology, but N1c has at least two Upper Paleolithic splits according to the recent EBC paper, so if the Neolithic Chinese samples aren't ancestral to Europeans it won't be surprising. What is surprising is that all Finnic and Baltic N1c1 shares a more recent ancestor with Itelmen/Chukchi N1c1 (5000 bp) than with Udmurt N1c (8000 bp).

Rob said...

sounds like an differential founder effects (?), as per paper

Ryan said...

@David - "And there was no SHG or EHG in the East Baltic prior to the Corded Ware expansions, just WHG. This is a fact that you will have to get used to one day."

Huh? Isn't the East Baltic the EHG heartland?

Davidski said...

Try the Pontic-Caspian steppe and surrounds.

ArtemisVentus said...

In order for Villabruna to bring CHG to Europe there must have been some CHG population living in Eastern Europe and on the Steppe before the arrival of ANE. It couldn't have come recently from the Caucus because Satsurbila and Kotais contain Basal Eurasian whereas the all encompassing WHG Villabruna cluster does not.

We've got 1 R1a from Karelia, 1 R1a and 1 J (with mtDNA U4) from Yuzhnyy Oleni Ostrov, neither of these places are all that far from the East Baltic and all those samples are from the Mesolithic. I expect the Mesolithic Baltic to have R1a and maybe a little J, or possibly some R1b. Autosomally they will be EHG with probably more CHG than more Eastern EHGs.

Isotopic analysis of the Corded Ware males from Haak showed they were local, and the women were not. Even if they were R1a, they weren't migrants from the Steppe. Just local HGs adapting to the Neolithic way of life.

Davidski said...

I pointed out already that Baltic foragers were WHG. I wasn't debating it. This is what they are.

And Corded Ware most certainly came to the Baltic from the steppe, because some Corded Ware individuals are basically like Yamnaya.

So, we have both WHG locals and Yamnaya-like locals in the Baltic not mixing until the Late Neolithic there? Highly unlikely.

Btw, no idea why you think that Villabruna brought CHG to Europe. Obviously, CHG came via the steppe and later the Balkans, with Yamnaya-related groups and Anatolia Chalcolithic-related groups.

Ariel said...

@the uralicist in this section.

Ok, Proto-Uralic had the same genetic makeup as yamna (EHG+CHG), they weren't far from each other and lived around the same time frame (3500-2000BC). Nope, I'm body and mind against that.
What if Uralic came from the east? Well after PIE and bringed with them "tangible" siberian and central asian admixture? What if...

http://eurogenes.blogspot.it/2015/09/uralic-genes.html

PS: hg N1c doesn't point west either

Ryan said...

@David - I'd think Karelia would qualify as "East Baltic" would it not? And that's where canonical EHG was found... so... yah. I'd imagine a rather wide ranging group overall.

Davidski said...

The Karelia Mesolithic site is not in the East Baltic; it's half way between the Baltic Sea and the White Sea.

And it has no relevance to who was living in the East Baltic during the Late Neolithic, just before the Corded Ware expansion into the region.

ArtemisVentus said...

@Davidski

Directly from the Ice Age Europe Paper:

Europe and the Near East drew together around 14 ka
Beginning around 14,000 years ago with the Villabruna Cluster, the
strong affinity to GoyetQ116-1 seen in El Mirón Cluster individuals
who belong to the Late Glacial Magdalenian culture becomes
greatly attenuated (Supplementary Information section 10). To test
if this change might reflect gene flow from populations that did not
descend from the >37,000-year-old European founder population,
we computed statistics of the form D(Early European, Later European;
Y, Mbuti) where Y are various present-day non-Africans. If no gene
flow from exogenous populations occurred, this statistic is expected
to be zero. Figure 4b shows that it is consistent with zero (|Z|<3)
for nearly all individuals dating to between about 37,000 and 14,000
years ago. However, beginning with the Villabruna Cluster, it becomes
highly significantly negative in comparisons where the non-European
population (Y) is Near Easterners (Fig. 4b; Extended Data Fig. 3;
Supplementary Information section 11). This must reflect a contribution
to the Villabruna Cluster from a lineage also found in present-day
Near Easterners (Fig. 4b).


-----------
This is the important part
-----------


The Satsurblia Cluster individuals from the Caucasus dating to
~13,000–10,000 years ago2
share more alleles with the Villabruna
Cluster individuals than they do with earlier Europeans, indicating that
they are related to the population that contributed new alleles to people
in the Villabruna Cluster, although they cannot be the direct source of
the gene flow. One reason for this is that the Satsurblia Cluster carries
large amounts of Basal Eurasian ancestry while Villabruna Cluster individuals
do not


-------------------------------

So Villabruna didn't have "CHG" exactly as per Satsurbila and Koitas, it had the non-basal eurasian part of CHG. But the J and CHG in EHG is proof enough of CHG in Eastern Europe in the Mesolithic.

Davidski said...

What if Uralic came from the east? Well after PIE and bringed with them "tangible" siberian and central asian admixture? What if...

Mix Corded Ware with some WHG and you basically get modern Balts and northern Slavs. No need for anything from Central Asia or SHG.

Davidski said...

You guys should maybe consider the possibility that I know and understand more than you?

Crazy, I know, but not impossible.

ArtemisVentus said...

Also I believe the Mesolithic Karelia site is relevant. After the ice age, in the Magdalenian Period, Europe was Recolonized from the West to East by the pre-ice age population, and from East to West by ANE, which is why we see that ANE shift in Motala vs Loschbour and Labrana. At a minimum I see the Baltic Mesolithic as SHG, but more likely EHG.

Ryan said...

@David

"The Karelia Mesolithic site is not in the East Baltic; it's half way between the Baltic Sea and the White Sea."

Closer to the Baltic Sea than the Pontic Steppe at least.

"And it has no relevance to who was living in the East Baltic during the Late Neolithic, just before the Corded Ware expansion into the region."

When did we specify this was only about the Late Neolithic?

Davidski said...

You must have missed these...

The results show that the typical European hunter-gatherer maternal lineages are represented exclusively in all individuals from until the Middle Neolithic. From the Late Neolithic on, haplogroups that are associated with European Neolithic farmers are detected. The results indicate genetic continuity of foraging cultures of Mesolithic and early Neolithic backgrounds and a late demic diffusion into the territory of Estonia associated with people of the Corded Ware culture.

http://eurogenes.blogspot.com.au/2016/09/isba7-palaeobarn-abstracts.html

In this paper we aim to show how these processes affected the Eastern Baltic region where the archeological record shows a drastically different picture than Central and Southern Europe. While agricultural subsistence strategies were commonplace in most of the latter by the Middle Neolithic, ceramic-producing hunter-gatherer cultures still persisted in the Eastern Baltic up until around 4000 BP and only adopted domesticated plants and animals at a late stage after which they disappeared into the widespread Corded Ware culture.

http://eurogenes.blogspot.com.au/2016/08/eaa-2016-abstracts.html

None of this should be surprising to anyone who's been reading the posts here for the last few years.

But have another look at the stats I posted above. The story is obvious. There was a large scale Indo-Europeanization of Eastern Europe from the steppe by the Corded Ware people, and the change was also genetic.

Davidski said...

It really makes no difference anyway whether EHG, SHG or martians were living in the East Baltic prior to the Corded Ware expansions.

The stats show that EHG and Yamnaya-related ancestry are closely correlated in Indo-European speaking Eastern Europeans.

Who doesn't understand this yet? And why not?

Onur said...

@Artemis

So Villabruna didn't have "CHG" exactly as per Satsurbila and Koitas, it had the non-basal eurasian part of CHG. But the J and CHG in EHG is proof enough of CHG in Eastern Europe in the Mesolithic.

There is no CHG in EHG as EHG lacks Basal Eurasian.

Shaikorth said...

@Ariel

I'm not an uralicist or other kind of professional linguist, the BA Volga urheimat theory is not mine but that of Anthony's and other academia. If it's wrong we'll see it - when ancient DNA becomes more plentiful - because then N1c will postdate Seima-Turbino in Europe.

The Siberian genetic signal in modern Uralic speakers was tested in Busby et al. 2015. with Globetrotter. In Mordvins it's shared with Turkics and dates to Iron Age. In Finland and a lesser degree in Scandinavia the eastern signal was shared with Yeniseians/Selkups. Turkics in Europe and West Asia share the same Tungusic/Eastern Turkic cluster using fine-scale methods, which is why we can pretty safely infer the proto-Turkic signal from moderns, and since this isn't the case with Uralic speakers in Eastern Europe we need to find the ancient DNA to figure it out.

epoch2013 said...

@Ariel

There is an old and deep connection between Proto-Uralic and PIE. It is - to me, at least - clearly a genealogical connection. They share such features as me- for me, I, ne- for negation etc etc. Features that seem the least likely to be loanwords. That means - to me, again - that they may share a common root. But it also clearly shows how far languages can drift apart over the course of time.

Ryan said...

@David - Gotcha.

I thought this was suspected for linguistic reasons for a while? IE that there is an Indo-European substrate in Uralic areas around the Baltic?

epoch2013 said...

@David

Completely offtopic, but should you find some time would you kindly run this?

Mbuti Villabruna Anatolia_Neolithic Israel_Natufian
GoyetQ116-1 ElMiron Anatolia_Neolithic Israel_Natufian

Onur said...

@Shaikorth

The Siberian genetic signal in modern Uralic speakers was tested in Busby et al. 2015. with Globetrotter. In Mordvins it's shared with Turkics and dates to Iron Age. In Finland and a lesser degree in Scandinavia the eastern signal was shared with Yeniseians/Selkups. Turkics in Europe and West Asia share the same Tungusic/Eastern Turkic cluster using fine-scale methods, which is why we can pretty safely infer the proto-Turkic signal from moderns, and since this isn't the case with Uralic speakers in Eastern Europe we need to find the ancient DNA to figure it out.

A big part of the East Eurasian genetic element of Volga Turkics (Chuvash, Tatars and Bashkirs) comes from Uralic peoples as can be seen in ADMIXTURE analyses:

http://i63.tinypic.com/fk1c92.png

Ariel said...

"Completely offtopic, but should you find some time would you kindly run this?

Mbuti Villabruna Anatolia_Neolithic Israel_Natufian
GoyetQ116-1 ElMiron Anatolia_Neolithic Israel_Natufian"

Yes, someone should try Villabruna vs Vestonice/Goyet for Natufians. Also something like Mbuti Vestonice Anatolia_Neolithic Israel_Naufian

Shaikorth said...

ADMIXTURE is confounded by recent drift (see Falush et al preprint etc). The Siberian in Chuvashes is, in fact, largely shared with other Turkics, though there's a good chance that their Uralic neighbours have the same situation as Mordvins going on.

This is consistent over every Globetrotter analysis Busby et al performed.


http://www.cell.com/cms/attachment/2062326182/2064072288/mmc2.xlsx

It's also why Chuvashes, Mordvins and Western Turkics share Han donors in Broushaki et al. fits.

http://science.sciencemag.org/highwire/filestream/681447/field_highwire_adjunct_files/3/Table_S24.xlsx
http://science.sciencemag.org/highwire/filestream/681447/field_highwire_adjunct_files/4/Table_S25.xlsx

Samuel Andrews said...

@Everyone,

Can someone run these?

Udmurd, Russian_North, Karelian
EHG, WHG, Esperstedt_MN, Andronovo, Siberian(as many as you can add)
EHG, WHG, Esperstedt_MN, CHG, Siberian(as many as you can add)

Uygur, Scythian_IA, Kazakh.
Andronovo, IranN, CHG, EHG, Siberian(as many as you can add)

@David,

Scythian_IA shows excess IranN/CHG in many ADMIXTURE tests I've seen. There's IranN/CHG ancestry, beyond what Steppe can explain, in many modern Central Asians I've seen. This is why our guesses on what Andronovo would be were too Caucasus before we got their DNA.

Only explanation is there's CHG/IranN in North Eurasia beyond what Yamnaya can explain. For the Sycthian an answer can be their ancestors spent time in Western Asia. For everyone else I think there's something going on we haven't taken into account.

Shaikorth said...

@Samuel

If you've checked the Lipka Tatar paper in David's other blog, they aren't just a mixture of Eastern Slavs or Volga people with East Turkics but also have a significant Caucasus contribution best represented by Armenians, clearly visible with fine scale analysis.

Onur said...

@Shaikorth

ADMIXTURE is confounded by recent drift (see Falush et al preprint etc). The Siberian in Chuvashes is, in fact, largely shared with other Turkics, though there's a good chance that their Uralic neighbours have the same situation as Mordvins going on.

This is consistent over every Globetrotter analysis Busby et al performed.


http://www.cell.com/cms/attachment/2062326182/2064072288/mmc2.xlsx

It's also why Chuvashes, Mordvins and Western Turkics share Han donors in Broushaki et al. fits.

http://science.sciencemag.org/highwire/filestream/681447/field_highwire_adjunct_files/3/Table_S24.xlsx
http://science.sciencemag.org/highwire/filestream/681447/field_highwire_adjunct_files/4/Table_S25.xlsx


None of those populations are highly drifted. We are not talking about long-isolated populations. The East Eurasian ancestry in Volga Turkics is clearly more similar to the East Eurasian ancestry in Uralics than that found in all other Turkics. This is especially true for the Chuvash, who are the oldest extant Turkic population in the Volga region and have had enough time to mix with their Uralic neighbors.

Gioiello said...

@ Davidski

I apologize for my ignorance in the autosome, but just two quotes from ArtemisVentus:

1 "However, beginning with the Villabruna Cluster, it becomes highly significantly negative in comparisons where the non-European population (Y) is Near Easterners (Fig. 4b; Extended Data Fig. 3; Supplementary Information section 11). This must reflect a contribution to the Villabruna Cluster from a lineage also found in present-day Near Easterners (Fig. 4b)".
2 "The Satsurblia Cluster individuals from the Caucasus dating to ~13,000–10,000 years ago2 share more alleles with the Villabruna Cluster individuals than they do with earlier Europeans, indicating that they are related to the population that contributed new alleles to people in the Villabruna Cluster, although they cannot be the direct source of the gene flow".

That is used from this Caucasian for demonstrating that Caucasus influenced Samara:
"One reason for this is that the Satsurblia Cluster carries large amounts of Basal Eurasian ancestry while Villabruna Cluster individuals do not. So Villabruna didn't have "CHG" exactly as per Satsurbila and Koitas, it had the non-basal eurasian part of CHG. But the J and CHG in EHG is proof enough of CHG in Eastern Europe in the Mesolithic".

But don't you think that from all that the most important thing is that Villabruna influenced Middle East and Caucasus and not the other way around (that always for my theory that at least R1b was in the "Italian or Alpine or Western European Refugium" (with very likely R1a, J and of course at least hg I)?

Aram said...

Davidski
Excuse me for offtop. Is there a way to test whether Armenia MLBA has an extra ancestry ( compared to Kura Arax) from post Yamna EEF rich Steppe populations? I got this impression comparing MLBA to EBA in some gedmatch calculators. Also Lazaridis says that MLBA has higher ENF & EHG and lower CHG compared to EBA.

Shaikorth said...

"None of those populations are highly drifted."

Yes they are Onur, especially Maris and Udmurts which couldn't pass the similarity exclusion of EGDP, like Roma, Igorots and Shors (Figure S2.2.1-VIII of Pagani et al) They are not Neolithic or even Bronze Age isolates, but neither are EGPD Roma. This doesn't matter, it doesn't take that much time to become highly drifted.

This drift confounds their Admixture results and inferences made based on that, as it does with Chuvashes. With methods that get around this, like Globetrotter, it becomes obvious that the bulk of Siberian contribution in Chuvash is similar to that of other Western Turkics. This is why they get such a clear Han donation in Broushaki's fits as well. Since Mordvins also have this ancestry, it's very likely that Maris etc. do as well.

Aram said...

As for that map. To be fair the first half of that migration is realistic. There is evidence of migration from Iran Chl to Central Asia. But off course it doesn't turn back and has no effect on Europe.
This map of G1 y dna speaks itself.

http://journals.plos.org/plosone/article/figure/image?size=large&id=info:doi/10.1371/journal.pone.0122968.g002

G1 was found in Iran Chalcolithic. But not in Europe or Yamna steppe.

P.S. That hotspots in Central Asia are recent founder effects.

Onur said...

@Shaikorth

Those ADMIXTURE components do not show significant influence of recent drift. No matter where you look, you will find out that the East Eurasian ancestry of Volga Turkics is closer to that of Uralics than it is to that of other Turkics. Here is another illustration of this fact:

http://4.bp.blogspot.com/_Ish7688voT0/TNY8EEnjDJI/AAAAAAAAC18/DL4nRsu8hy0/s1600/ADMIXTURE7.jpeg

Note that, like the Uralic peoples and unlike the non-Volga Turkic peoples in the analysis, the Chuvash have an excess of the "Nganassan" component over the "Yakut" component.

Davidski said...

@Ryan

Not sure, but as far as I know there's a consensus that Uralic replaced Indo-European in Estonia and nearby areas of Finland, probably when climate put Corded Ware herders/farmers out of business there.

As per the abstracts linked to above, typical European Hunter-Gatherer mtDNA is seen in the cultures preceding Corded Ware in the East Baltic, so this has to be WHG-like mtDNA, not anything unusual, like the EHG mtDNA at the Mesolithic Deer Island site in Karelia.

When Corded Ware reaches the East Baltic, it brings with it lots of farmer mtDNA, and, as per the recent paper on the Corded Ware and Unetice X chromosomes, farmer girl enriched X chromosomes, plus some EEF autosomal ancestry mixed in with its mainly Yamnaya-related ancestry.

This pretty much stays that way in much of Eastern Europe (refer to the models in Haak et al., which show Balts and Slavs as Yamnaya, EEF, WHG mixtures), but not in Finland, where a fair amount of extra EHG and also Siberian admixture probably arrives there after the above mentioned Corded Ware downturn.

@Epoch & Arial

Mbuti Villabruna Barcin_Neolithic Israel_Natufian -0.0375 -8.389
GoyetQ116-1 ElMiron Barcin_Neolithic Israel_Natufian -0.0119 -1.891
Mbuti Vestonice16 Barcin_Neolithic Israel_Natufian -0.0383 -8.726

@Samuel

Not really sure which D-stats you want. You'd have to make a specific list for me to run the correct ones.

By the way, Scythians were very active in the Kuban steppe near the Caucasus. They were very mobile, so they could have moved between South Siberia, the Kuban steppe and the Pontic steppe, spreading around all sorts of ancestry.

@Aram

Both Armenia_Chalcolithic and Aremnia_MLBA have European ancestry, including EHG, that Armenia_EBA seems to lack. This is discussed in Laz et al.

I can add Armenia_MLBA to the sheet above later today, so you can run some graphs. It should be pretty easy to see the European influence by plotting Armenia_MLBA against Armenia_EBA/Caucasus_HG.

Ryan said...

I think it would be really helpful to just see a PC of the various peoples on the steppe and everywhere north.

I think the most obvious vector for East Asian ancestry would be via the spread of Pit-Comb Ceramics from somewhere near the Liao river. Right haplotypes (almost) and right archaeology.

http://bmcevolbiol.biomedcentral.com/articles/10.1186/1471-2148-13-216

Though from 10kya on NE Asia seems to have been in utter chaos. You have Indo-Europeans intruding via the steppe, Uralics and Yukaghir and N1c expanding, agriculture spreading to Korea and Japan, C3 somehow spreading all the way across to North America, Q was hanging out there somehow... and none of these groups seem to correlate with each other.

Shaikorth said...

Onur, Nganasan and Yakut "components" ADMIXTURE creates are clear examples of recent drift and endogamy of these populations. Western Turkics' - Chuvash included - Siberian ancestry comes from more southern populations, as does that of Mordvins etc.

These samples were in the cluster that was the best match for eastern admixture in Mordvins, Chuvash and Turks:
buryat x15 kyrgyz x14 altai x13 tuva x13 mongolian x12 daur x9 oroqen x7 hezhen x6
burya x2 nganassan x1 uygur x1

Similar populations formed the East Eurasian contribution in Lipka Tatars.

And yes, before you ask, more Nganasans and Yakuts were in Busby's study. They were in the regular "Siberian" cluster save a few outliers including the one mentioned above.

Alberto said...

@Aram

Using nMonte with a D-stats spreadsheet:

Armenia_MLBA
"Armenia_EBA" 77.55
"Yamnaya_Samara" 17.7
"Sintashta" 4.3
"Satsurblia" 0.45
"Anatolia_Neolithic" 0
"Loschbour" 0
"Eastern_HG" 0
"Armenia_Chalcolithic" 0
distance%=1.1007 / distance=0.011007

But ENA populations get underfitted, and adding Ulchi changes the balance towards Sintashta:

Armenia_MLBA
"Armenia_EBA" 73.25
"Sintashta" 18.1
"Yamnaya_Samara" 5.1
"Ulchi" 2.2
"Satsurblia" 0.8
"Anatolia_Neolithic" 0.55
"Loschbour" 0
"Eastern_HG" 0
"Armenia_Chalcolithic" 0
distance%=0.9241 / distance=0.009241

So definitely Armenia_MLBA has less Basal Eurasian and a pull towards the steppe. Whether more towards Yamnaya or more towards Sintashta might depend on what is the real meaning of that extra ENA affinity.

BTW, Armenia_EBA from Armenia_ChL:

Armenia_EBA
"Armenia_Chalcolithic" 45.45
"Iran_Chalcolithic" 21.55
"Satsurblia" 18.35
"Anatolia_Neolithic" 10.95
"Levant_Neolithic" 2.95
"Yamnaya_Samara" 0.75
"Loschbour" 0
"Eastern_HG" 0
"Ulchi" 0
"Israel_Natufian" 0
"Iran_Neolithic" 0
distance%=0.7368 / distance=0.007368

Onur said...

@Davidski

Can you run these stats?

Mbuti Chuvash Yakut Nganasan
Mbuti Tatar_Kryashen Yakut Nganasan
Mbuti Bashkir Yakut Nganasan
Mbuti Mari Yakut Nganasan
Mbuti Mordvin Yakut Nganasan
Mbuti Finnish Yakut Nganasan
Mbuti Russian_Kargopol Yakut Nganasan
Mbuti Saami Yakut Nganasan
Mbuti Udmurd Yakut Nganasan
Mbuti Mansi Yakut Nganasan
Mbuti Nenets Yakut Nganasan
Mbuti Turkish Yakut Nganasan
Mbuti Uygur Yakut Nganasan
Mbuti Kazakh Yakut Nganasan
Mbuti Turkmen Yakut Nganasan
Mbuti Uzbek Yakut Nganasan
Mbuti Altaian Yakut Nganasan
Mbuti Shor Yakut Nganasan

Alberto said...

@Samuel

The spreadsheet that Davidski posted in this thread does not have the ancients as rows, but only as columns. So they can't be used to make those models that you ask with nMonte.

Shaikorth said...

David, if you're doing D-stats run this too, there's a thing about them I want to check:
Mbuti Chuvash Latvian Mordvin
Mbuti Tatar Latvian Mordvin

Davidski said...

@Aram

Yeah, clearly, Armenia_MLBA does have North European-related ancestry that Kura-Araxes (Armenia_EBA) lacks. Kura-Araxes is instead much more Northeast Caucasian.

https://4.bp.blogspot.com/-eC3uQRfKTrc/V_N7AVvaLZI/AAAAAAAAE8k/MmKWNY6dj-k0hLhb7TPr79T2UERJtHA8wCLcB/s1600/Armenia_MLBA_vs_EBA.png

Here's the updated datasheet.

https://drive.google.com/file/d/0B9o3EYTdM8lQQ3M5cUpCZlU0czg/view?usp=sharing

Ariel said...

"Mbuti Villabruna Barcin_Neolithic Israel_Natufian -0.0375 -8.389
Mbuti Vestonice16 Barcin_Neolithic Israel_Natufian -0.0383 -8.726"

Ok, this doesn't work, at all. Barcin is too close to Natufian and has already WHG like admixture. Using anatolia isn't a great idea. Maybe Iran neolithic could be more informative. If someone want to try these..

Mbuti Villabruna Iran_Neolithic Israel_Natufian
Mbuti Vestonice Iran_Neolithic Israel_Natufian
Mbuti AG3/MA1 Iran_Neolithic Israel_Natufian (as a sanity check)

Davidski said...

D-stats...

Mbuti Chuvash Yakut Nganasan 0.0046 1.364
Mbuti Tatar_Kryashen Yakut Nganasan 0.0025 0.737
Mbuti Bashkir Yakut Nganasan 0.008 2.415
Mbuti Mari Yakut Nganasan 0.011 3.46
Mbuti Mordovian Yakut Nganasan -0.0043 -1.343
Mbuti Finnish Yakut Nganasan -0.0002 -0.069
Mbuti Russian_North Yakut Nganasan -0.0027 -0.713
Mbuti Saami Yakut Nganasan 0.0102 2.841
Mbuti Udmurd Yakut Nganasan 0.0058 1.779
Mbuti Mansi Yakut Nganasan 0.0223 5.548
Mbuti Nenets_Forest Yakut Nganasan 0.0437 10.16
Mbuti Nenets_Tundra Yakut Nganasan 0.0398 9.993
Mbuti Turkish Yakut Nganasan -0.0054 -1.756
Mbuti Uygur Yakut Nganasan -0.0013 -0.294
Mbuti Kazakh Yakut Nganasan 0.0068 1.925
Mbuti Turkmen Yakut Nganasan 0.0047 1.359
Mbuti Kyrgyz Yakut Nganasan 0.01 2.773
Mbuti Altaian Yakut Nganasan 0.0134 3.76
Mbuti Shor Yakut Nganasan 0.0079 2.056

Mbuti Chuvash Latvian Mordovian -0.0107 -4.848
Mbuti Tatar_Kryashen Latvian Mordovian -0.0086 -3.833

Mbuti Villabruna Iran_Neolithic Israel_Natufian 0.0222 3.513
Mbuti Vestonice16 Iran_Neolithic Israel_Natufian 0.0053 0.817
Mbuti AfontovaGora3 Iran_Neolithic Israel_Natufian -0.0408 -5.055

Shaikorth said...

Thanks David, this pretty much highlights how homogenous populations can have extra pull in D-stats, enough to create statistically significant "odd" results. Also, how ADMIXTURE components can be ancestral blurs. Turkmens for example had significant "Yakut" and no "Nganasan" in the EBC Turkic paper's ADMIXTURE run yet D-stats don't show any such preference. Not that it matters since that paper pinpointed Tuvan/Buryat type pops as the best proxies for eastern Turkic, which agrees with later fine scale analysis from Busby etc.

Onur said...

Mmm... the D-stat results are hard to make sense of, so I prefer the ADMIXTURE results in this case. Thanks anyway, David.

Shaikorth said...

https://twitter.com/3rdreviewer/status/714864074568384512

Back to D-stats, is there a pattern with these?

Mbuti Natufian Villabruna Kostenki14
Mbuti Natufian Villabruna Goyet_Q116
Mbuti Natufian Villabruna Vestonice16
Mbuti Natufian Villabruna ElMiron
Mbuti Natufian Villabruna Loschbour
Mbuti Natufian Villabruna EHG
Mbuti Natufian Villabruna AG3

epoch2013 said...

@Ariel

O but this yielded something interesting. El Miron is modeled as 33% WHG (in other models in Fu at al even more). These D-stats mean that the part of WHG that is admixted in Magdalenians did not admix in Middle-Easterners at all or did so evenly, whereas something looking like WHG and Gravettians admixted far more in Anatolia than in the rest of the Middle East. That, to me, means that the part of WHG in El Miron is *different* from the part WHG in Anatolia, and that the Anatolian part looks like Vestonice. And that IMHO means that WHG is a mixture of Gravettians and another entity, found in Magdalenians as well.

I would like to combine that with the fact that there is something as the Levantine Aurignacian. They were originally though as predecessor of the European Aurignacian but timing issues made it clear that it emerged *later* that European Aurignacian. I would like to suggest a Aurignacian group spreading south being the source of both this unknown part as well as the source of the strange Middle-Eastern affinity in the Villabruna group.

Whatever it is: I think these D-stats show that Villabruna was an admixture of something unknown, Gravettian and a small dose of ANE.

Onur said...

@Shaikorth

What you fail to understand is that most ADMIXTURE components are not the result of drift in a single population but collective results of many of the populations used in an ADMIXTURE analysis. Components peaking in Yakuts and Nganasans surely cannot be explained solely by drift in those populations given their widespread distribution in Siberia-related populations and the logicalness in the pattern of their distributions.

Davidski said...

Mbuti Israel_Natufian Villabruna Kostenki14 -0.0385 -6.155
Mbuti Israel_Natufian Villabruna GoyetQ116-1 -0.0281 -4.165
Mbuti Israel_Natufian Villabruna Vestonice16 -0.0347 -5.453
Mbuti Israel_Natufian Villabruna ElMiron -0.0187 -2.802
Mbuti Israel_Natufian Villabruna Loschbour -0.0061 -1
Mbuti Israel_Natufian Villabruna Karelia_HG -0.0313 -5.012
Mbuti Israel_Natufian Villabruna AfontovaGora3 -0.0537 -6.479

Rob said...

@ Epoch

"And that IMHO means that WHG is a mixture of Gravettians and another entity, found in Magdalenians as well. "

I think the Fu paper commented on that, didn't it ?

" a Aurignacian group spreading south being the source of both this unknown part as well as the source of the strange Middle-Eastern affinity in the Villabruna group.."

That is what I originally thought. But we don;t have an Epi-Gravettian samples from SEE / Black Sea to exclude this, other, archaeological possiblity.

Shaikorth said...

@Onur

Those populations get their own components when K is increased because they are drifted, not because they are some ancestral relics (in fact they are recent). That these components appear in others doesn't matter - you'll see examples of that in the Falush preprint. It's not unlike MA-1 getting sigificant Kalash at one K and generic SC-Asian at other.

This is why methods that ignore this drift can show something quite unlike inferences made from ADMIXTURE. But these results make sense, including historical sense unlike Yakutia being the origin of Turkic expansions. They may look illogical at a glance but only because they contradict ADMIXTURE-based inferences which shouldn't have been made in the first place. In Broushaki's modeling using the Human Origins set Yakuts are haplotype donors to Dolgans and their Tungusic and Yukaghiric neighbours, that's a sensible result.

Onur said...

@Shaikorth

I have already explained why I do not agree with you on the drift issue. Anyway, I do not think Yakutia is the origin of the Turkic expansions. I think populations such as Tuvans, Mongolians and Buryats best represent the original Turkic type. It is harder to make a guess for the original Uralic type and Uralic homeland.

Ryan said...

Random question here - how relevant is your old K7 ANE test on Gedmatch these days. It has me at over 60% WHG, which seems to be outside the range of the populations in your table, and wasn't sure how off this was.

Onur said...

@Ryan

Some of the Eurogenes calculators on GEDmatch are obsolete now. Personally I see no reason to still keep them on GEDmatch, but it is of course up to Davidski to decide on their fate as the creator and owner of those calculators.

Shaikorth said...

@Onur

Frankly speaking a "widespread" Yakut component looks like a Siberian analogue of Pagani et al 2012's Ari Blacksmith component which not only made them look like some "ur-East Africans" but was spread beyond Aris and significantly present in pops like Afars and Somalis. Falush et al's retesting of course showed what we knew from van Dorp et al 2015: that the Blacksmiths and Cultivators are the same population divided by drift (due to marginalization of the Blacksmiths), and also that Somalis and Afars don't have Blacksmith or Cultivator admixture.

Such issues are part of ADMIXTURE/STRUCTURE quirks:
"Specifically, groups that are numerically small with respect to other groups in the
sample or have undergone little population-specific drift of their own are likely
to be fit as mixes of multiple drifted groups, rather than given their own ancestral population.
...when a source population is put into a dataset with two or more drifted sink populations. The source can be represented as a mix, even though there is no mixture within its history."

Davidski said...

The ANE K7 isn't compatible with the Basal-rich K7 and it's not really designed for estimating overall WHG ancestry, but it's still OK for estimating relative ANE levels.

Alberto said...

@epoch2013

"I would like to combine that with the fact that there is something as the Levantine Aurignacian. They were originally though as predecessor of the European Aurignacian but timing issues made it clear that it emerged *later* that European Aurignacian."

Yes, and then there is the Zagros Aurignacian, which was thought to be later, but now new dates make it earlier than the Levantine one. I have no idea how this "Aurignacian" broad term would correlate with genetics, but I would love to know it (though human remains are mostly absent, so probably it won't be easy).

https://www.academia.edu/1046320/New_radiocarbon_dates_for_the_Zagros_Aurignacian_from_Yafteh_cave_Iran

https://www.academia.edu/3281693/Aurignacian_in_the_Zagros_Region_Test_excavation_at_Yafteh_Cave_Lorestan_Iran

Onur said...

@Shaikorth

ADMIXTURE is a great tool for inferring population structure as long as you know how to use it and interpret its results. That is exactly what I do. I do not treat ADMIXTURE components as absolutes and instead focus on their relative distributions across populations. This helps you arrive at robust conclusions about population structure. For instance, the Uralic to Altaic ratio in the East Eurasian components of Turkic populations is above 1 only among Volga Turkic populations and this indicates higher Proto-Uralic ancestry than Proto-Altaic or Proto-Turkic ancestry among Volga Turkic populations. This is a robust conclusion and can be replicated easily using different ADMIXTURE analysis strategies.

Gill said...

There's something similar to Bronze Age Armenia in South Asia.

From the West Eurasian components, if you separate out the following:

1. The typical Gedrosia stuff, likely of Iranian Neolithic ancestry. In older terms, it's high ANE, high ENF, no WHG.

2. The Steppe stuff: WHG, and its associated ANE/ENF (pointing to a high ANE, high WHG ancestor)

And then separate out a third component: The ASI stuff: ASE/Onge/Oceanian/East Eurasian...

That leaves something left over which resembles Bronze Age Armenia or Chalcholithic Iranian, peaks in modern Iran and continues through Afghanistan and into North India in a relatively smooth gradient. I've been calling it ANI.

This assumes that all the East Eurasian in South Asians is from ASI and not the Steppe (because it often registers as Siberian, Amerindian, Beringian, etc in calculators).

There's also some archaic stuff that shows up as African left over.

Shaikorth said...

@Onur

It's a solid correlation of ADMIXTURE clusters - that are not ancestral so bluntly put about as useful as measuring Ari Blacksmith admixture in Somalis and Maasai. D-stats also showed that it didn't translate into meaningful differences between Turkic groups (Chuvash and Turkmen show identical results re: Yakut/Nganasan).

Nganasans and Yakuts are sinks, not sources. Once you account for drift the source of eastern ancestry in Chuvashes and Mordvins and Turkmens alike is not from their group (never mind the Yeniseian group) but more southern - regardless of language.

epoch2013 said...

@Alberto

Thanks. Maybe the road into ME, but the paper states very old dates, up to 42k BP. However, recently European dates have been pushed back as well.

Let's for the sake of the argument go like this: Europe > ME.

@David

Could you do this? The idea being that we could play with the WHG part of Magdalenians and compare to CHG, which was reported to have a WHG signal as well.

GoyetQ116-1 ElMiron Satsurblia Israel_Natufian
GoyetQ116-1 ElMiron Kotias Israel_Natufian
GoyetQ116-1 ElMiron Satsurblia Iran_Neolithic
GoyetQ116-1 ElMiron Kotias Iran_Neolithic

Onur said...

@Shaikorth

Of course Yakuts and Nganasan are sink populations. But their relative isolation or driftedness informs us about the underlying genetic structure of them and the related populations to them by way of the algorithms of ADMIXTURE. Contrary to what you say, ADMIXTURE components are not garbage.

Shaikorth said...

@Onur

The Nganasan and Yakut components of ADMIXTURE tell they are drifted and given that once you account for drift they are not sources of ancestry outside their immediate neighbourhood, probably in a similar situation as Ari Blacksmiths who also got such a component.

It's a drifted mix. No one's saying that's garbage, but you should really try not to ascribe it continent-spanning meaning. If you want to try to understand their underlying genetic structure beyond that they are drifted, using ADMIXTURE and adding K's until they get their own components isn't the way either.

Onur said...

@Shaikorth

So what? Before the ancient autosomal tests of Europe, many people (including me) thought that the widespread ADMIXTURE component peaking in Sardinians was the result of high driftedness of Sardinians and was garbage. But we now know that Sardinians are the modern population that has best preserved the Neolithic Europan genetic structure, they are like a relic from the Neolithic times (not entirely so but close). Maybe Yakuts and Nganasan too preserve very old genetic structures that have been more diluted elsewhere over time.

Shaikorth said...

They aren't analoguous to Sardinians - once drift is accounted for we see that Sardinians still look like neolithic relics, while other pops which easily form their own components like Kalash or Nganasans look like a mix of their modern neighbours. If there's preserved old genetic structure in Eastern Siberia it's logical to assume it's best represented by paleosiberians, and that's what the little data we have suggests too: in Flegontov et al. 2016 Saqqaq clustered with some Yukaghirs in Finestructure using the Human Origins set and with Chukotko-Kamchatkans using their Illumina set.

Onur said...

@Shaikorth

I agree with you regarding the Kalash. But the Kalash component is never as widespread as the Yakut and Nganasan components and does not show any relation with certain language families.

Shaikorth said...

Yet those "relations", such as the Turkmen v Chuvash one, turned out quite illusory once tested with D-stats and didn't show up if you correct for drift just like the Ari Blacksmith component in Somalis, so we don't have a particular reason to care about them.

A component peaking in Kalash can be all over South Central Asia and India, and so can a Burusho component. When Hazaras or Sindhis show a Burusho component we can check whether they really have Burusho admixture with Finestructure. Broushaki's fits with Human Origins suggest they don't.

In line with the Falush preprint "use [of algorithms like STRUCTURE] should represent the beginning of a detailed demographic...analysis, not the end". If a more detailed look reveals a component doesn't actually represent a contributing population there's no reason to keep chasing such ghosts.

Matt said...

I think it's still tricky to know you've accounted for all the drift. Another example than the Sardinians is that lots of us here thought the Gedrosia and Baloch components that showed up in some high K were just drift. But then the Baloch and Makrani turn out to actually really to seem to have a slight increase in affinity to the Iranian Neolithic. Of course the Baloch component wasn't a real ancient population however it seems likely the excess of the Iran Neolithic in the populations rich in the component and not just drift alone was why that component was generating.

To account for drift, there are methods we can use, like explicitly looking at drift through Runs of Homozygosity. Just still seems hard to be 100% without the ancient dna.

Shaikorth said...

@Matt


Checking for RoH can tell how drifted someone is but can't tell whether they're special in their ancestral proportions. Sardinians have lots of RoH, but so do Sauris villagers in NE Italy etc

The Broushaki fits actually do a good job of pinpointing the specialness of Balochi/Brahui/Makrani in having excess Iran_N once you take out the Pakistani donors - preventing the three from donating to each other - and try to model them as ancients+moderns. They get Iran_N way in excess to anyone else then, just like Sardinians get ENF.

Matt said...

Re: Sauris, eh, I mean they're not that similar in RoH to Sardinians. If we're talking the isolates in those Italian villages, which are what form ADMIXTURE components and not the General Pop in those regions, which don't. Those isolates have massive huge RoH and LD compared to Sardinians - https://archive-ouverte.unige.ch/unige:34640/ATTACHMENT01. If you look at the figure 5, the RoH just isn't even close. Sardinians are practically the same as the mainstream of Europe in comparison. The GPop of Sauris, who have a reasonably similar RoH and LD to the Sardinians basically didn't form a component in that paper (none of the GPops did). So I think using LD and RoH would work a lot better than I think the impression you're giving for removing really outsized effects of drift on ADMIXTURE.

I do see what you mean about the donation fits being able to identify that the population is distinctive because of donation from Iran_N, but in this case we do have the adna. Imagine if we didn't, we would probably not have a clue, and would still think that a component tends to form in ADMIXTURE that donates to them just because of recent drift and not also to a degree unusual ancient structure.

Lukasz M said...

Little offtop Davidsky,
Do you use "Nonlinear fit" function from PAST program to generate those plots?

Davidski said...

I usually just run Model > Generalized Linear Model.

Other options might be more useful though depending on what you're looking at and how much detail you want.

Kurti said...

MaxT said

"Gets your facts right from "

Says the guy who claims the material culture of Yamna is not West Asian. I am not going to join this debate here. But your claim couldn't be more wrong, from herding and Bronze weapons to Kurgans and Pit graves anything looks to be intrusive from West Asia. It's like you can mention it over and over again but the ignorance is too high, the oldest Kurgans to date are from the leyla Tepe culture.

Rob said...

@ KurTi

Leila-Tepe dates to the same period is Majkop, so the two are part of the same epi-phenomenon (c. 3800 BC); and at present we haven clearly established directionality.
There is a lot more to kurgans than a a pile of dirt, and for a thorough comparison you'd need to include body positioning, grave goods, etc. As far as Yamnaya goes, some of the rituals can be traced back to the Skelya culture (4500 - 4000 BC), and the odd burial mounds also appear very early in northeast Balkans.

FrankN said...

@Rob: Latest research suggests that the directionality is from Leyla-Tepe to Majkop, and Leyla-Tepe has been taking up influences from the Iranian Plateau, Central Asia (Namzaga I) and w. Pakistan(Merhgarh, Mundigak). See my extensive comment here:
http://eurogenes.blogspot.de/2016/08/a-few-mito-genomes-from-maikop.html?showComment=1471358981706#c8063887656440690719

Such a genesis of Leyla-Tepe also corresponds to Alberto's nMonte results above that show substantial Iran Chalcolithic inflow into Armenia EBA:
Armenia_EBA
"Armenia_Chalcolithic" 45.45
"Iran_Chalcolithic" 21.55
"Satsurblia" 18.35
"Anatolia_Neolithic" 10.95
"Levant_Neolithic" 2.95
"Yamnaya_Samara" 0.75


I think there is sufficient linguistic evidence for early 4th mBC IE (Euphratic) influence on Sumerian, possibly out of W. Iran (Central Zagros). Iranian Plateau->SW Caspian Sea (Leyla-Tepe)->Majkop->Steppe is a plausible trajectory for the spread of (P)IE, of course along a different route than suggested in that dubious Gamkrelidze-Ivanov map.

If we accept a W. Iranian Homeland of PIE, Kura-Araxes should most likely also have spoken some form of early IE, which provides a trajectory into Anatolia and beyond, something a Steppe homeland cannot provide, at least not timely enough to explain Hittite.

Rob said...

@ Frank

Im aware of and certainly agree about the links you've described , although it's curious how Majkop pales anything south of it; and the material culture is wholly original, with numerous differences to K-A (although these are later, potentially different "people")
I was merely pointing out the complexities of Kurgan rite
I'm happy to wait for more aDNA for final conclusions; but needless to say, paradigms will fall ;)

Davidski said...

@Frank

You're basically describing some trade routes. That's exceedingly flimsy evidence for language change.

You need a migration, and evidence for this migration. You don't have that evidence, or even any signs of such a thing.

Davidski said...

@Rob

Paradigms will fall.

Some idiots will try to fight the said paradigms and they will fail, as they have in all attempts to date.

The data will do the talking.

Rob said...

Ha that's right Daveo, it already is

FrankN said...

@Dave "You're basically describing some trade routes. That's exceedingly flimsy evidence for language change."

To the opposite. It is trade that creates need for a lingua franca - English today, French in the 17-19th century, Latin well into the Medieval, Arab around the Indian Ocean from Dar-Es-Salam to Indonesia, Spanish in Latin America, Siouan and Algonquin languages in the N.American Woodland period, Akkadian and Aramaic in the NE, etc. pp.

Asides, Alberto has just provided genetic (nMonte) evidence for migration from Iran_Chalc into Armenia_EBA (though I am not sure about the dating, this may have happened after Leyla-Tepe). Your new K12 has Arm_Chalc at 37-44% Iran_Neol, going up to 59-62% in Arm_EBA (Kura-Araxes).

"The data will do the talking." It does!

Davidski said...

Armenia_EBA doesn't have any ancestry from Iran. It just shares CHG ancestry with Iran_Chalcolithic and CHG-related ancestry with Iran_Neolithic.

CHG admixture moved into Iran from the west, first during the Chalcolithic, so we see the clear western shift from Iran_Neolithic to Iran_Chalcolithic, and then with Hurrians, so Iron Age Iranians become even more similar, and in fact basically identical, to Armenia_EBA.

The data sure does talk, but you gotta know what it's saying, and it's not what you think.

Chad Rohlfsen said...

Uni-parental markers from Iran and Maykop do not match Yamnaya or Europeans. The nMonte is also not "evidence" of any specific people moving anywhere. More likely, it is also an Anatolian related people mixing with CHG-related pops in Armenia, rather than a pop replacement from the Zagros. Similar groups with similar ancestors... nothing more.

Rob said...

Well, I admit I'm wanting some surprises, otherwise what are we all going to debate about ?

Shaikorth said...

@Matt

Sauris was for demonstrative purposes, the genpops don't form their own components in that run because there's only so many K's and a more drifted subpop present to form it first, as per Falush et al.

If we had no ancient DNA it would be very difficult to discern a pattern about even Sardinians being so different from just IBD.

http://blogs.discovermagazine.com/gnxp/files/2012/08/01.jpeg

Davidski said...

@epoch

GoyetQ116-1 ElMiron Satsurblia Israel_Natufian 0.0009 0.097
GoyetQ116-1 ElMiron Kotias Israel_Natufian 0.0016 0.177
GoyetQ116-1 ElMiron Satsurblia Iran_Neolithic -0.0036 -0.449
GoyetQ116-1 ElMiron Kotias Iran_Neolithic -0.0064 -0.831

@Rob

Plenty to discuss apart from the basic location of the PIE homeland, and there will be plenty of big shocks I'm sure.

But we're still bogged down on many basic issues with people claiming, for instance, as per above, that any pile of dirt close to a favorable zip code from the Bronze Age must have been put up by a Proto-Indo-European. Just ridiculous.

Aram said...

Davidski and Alberto

Thank You very much. This meets my expectation. I will start to experiment with nMonte instead of Gedmatch.

As for Leyla Tepe that is discussed here. The Leyla Tepe is contemporary to that Armenia Chl samples. It is only on the other side of Syunik-Karabakh mountain range some 200-300 kilometers east. Leyla Tepe sites are more connected to Kura valley. So if Arm Chl has EHG then Leyla Tepe will also have EHG, probably even more because Kura valley is more open to Eurasian plains.

Matt said...

@ Shaikorth
Sauris was for demonstrative purposes, the genpops don't form their own components in that run because there's only so many K's and a more drifted subpop present to form it first, as per Falush et al.

It's difficult for me for you to, I think, assert that ADMIXTURE definitely would've been misled by the Italian general populations, when that didn't happen in their run, and that this is a problem of ADMIXTURE even without these really, really extreme outliers. I could just as easily say, no actually they wouldn't form a cluster, so ADMIXTURE is much better than you think at picking up real populations. :)

If we had no ancient DNA it would be very difficult to discern a pattern about even Sardinians being so different from just IBD.

Well, I'm not saying you would ever think Sardinians are unusual *just* from RoH or IBD, I'm saying more that you would combine that with clustering like ADMIXTURE or PCA, then also look at them from a RoH perspective and go "They tend to cluster very unusually; the RoH patterns aren't very different from recent people so this is unlikely to be due to extreme drift; therefore they're likely to represent an unusual ancient structure". That's the way people should be doing it rather than "Oh, ADMIXTURE's pointless for data exploration because drift can change frequencies" or "They form a cluster in ADMIXTURE therefore they automatically were an ancient population which has survived", without looking at the degree of drift. People who investigate should use the methods we have for checking if a population is actually subject to strong recent drift, and if it's not, then it probably does represent an interesting degree of ancient structure. (Another way of looking at drift is to look at the differentiation from an outgroup, usually African, but that's less perfect).

Shaikorth said...

"It's difficult for me for you to, I think, assert that ADMIXTURE definitely would've been misled by the Italian general populations, when that didn't happen in their run, and that this is a problem of ADMIXTURE even without these really, really extreme outliers. I could just as easily say, no actually they wouldn't form a cluster, so ADMIXTURE is much better than you think at picking up real populations. :)"

Those aren't "general" as in Tuscans, as per the RoH chart even most of the general FVG pops were clearly more drifted than other regions of Italy or Europe generally. Clusters pop up according to drift or sample size as per Falush. Not unreasonable to expect they'd get their clusters before North Germans or Bulgarians or French. There could also be a Jewish cluster or something like that.


If we look at the HGDP Pakistani populations, Pathans and Sindhi, Balochi and Brahui, Makrani show high recent drift based on long RoH - in fact Kalash have less extreme pattern of long RoH. But we would be mistaken to reject Baloch & co based on that as ancient DNA has shown. Without ancient DNA we could even say they get their own components often because they are oversampled (three very similar overlapping populations) which can cause cluster creation. Hence the "be sure to oversample your favourite group, if it doesn't get its own cluster increase K until it does" joking in the Falush preprint.

Lukasz M said...

I made nice PCA from those datasheet
https://s16.postimg.org/xs3j5a81v/pca.png

Kurti said...

Rob Said

"Leila-Tepe dates to the same period is Majkop, so the two are part of the same epi-phenomenon (c. 3800 BC); and at present we haven clearly established directionality. "

Not true, where did you got that from? This is outdated information but since that time new archeological sides have been found and Leyla Tepe is dated to 4500 BC. It is significantly enough older than both Maykop and Kura_Araxes and is said to be predecessor of these two.

Kurti said...

For example the Soyuqbulaq Kurgans have just been found in 2006. Having up to date informations would avoid these kind of false assumptions.

Davidski said...

It is significantly enough older than both Maykop and Kura_Araxes and is said to be predecessor of these two.

Right, so you're saying they were proto-Hurrians. I agree.

Kurti said...

SOme guys still asking for "evidences" of migration from the Iranian Plateau/Mesopotamia to the Steppes like broken records not realzing (or not having enough knowledge of archeology) that there is evidences of influx from NW Iran/Ubaid Mesopotamia to Leyla Tepe and from Leyla Tepe to Kura Araxes and Maykop while even Maykop Kurgans predate those in Yamna and look just like Maykop copies. There are several studies flying around the net about this some of them appearing just lately. But ok you always see what you want to see.

If you say PIE evolved in the Steppes ok that is your opinion and a valid and strong theory. But don't ignore the archeological facts.



If you mean that R1b is actually not only PIE but also Proto Hurrian fine with me. Maybe they even got that marker from the Proto Hurrians?!

Kurti said...

I have been pointing out the similarities between Hurrian and Indo Europeans for quite some time now. But somehow not many listen. It starts with the Deities such as Teshub.

Davidski said...

I'm not waiting for any evidence that there was a migration from Iran to the steppes. I already know that there wasn't.

We now have the ancient data that clearly says that there wasn't.

Matt said...

@ Lukasz: Yep, nice PCA and very nicely labeled.

Using just a subset of populations with a high average score (relatively West Eurasian):

http://i.imgur.com/8DByuye.png - with D-stats all
http://i.imgur.com/522OPyf.png - with just Barcin, CHG, Levant, Loschbour, Iran, Yamnaya, EHG

Kinda typical but with some displacements of populations in unusual places (that don't quite match pairwise Fst between them but that may relate to population movements more recent than Bronze Age that have linked populations together).

Davidski said...

If you transpose the datasheet you can get a very nice PCA of the ancient samples.

Rob said...

@ Kurti

I usually know what I'm talking about

Eg try "Ancient Metallurgy in the Caucasus From the Sixth to the Third Millennium BCE"
& "Late Chalcolithic Kurgans in Transcaucasia. The cemetery of SoyuqBulaq (Azerbaijan)"

Leila Tepe horizon is dated here from 3800 BC, the Sioni is 4500 BC. These kurgans appear to date specifically to 3800 BC

Moreover, they appear to be different to those in Majkop
Clearly, they are part of a similar phenomenon, but causality is not demonstrated

curiously, they appear after the disappearance of Shuvaleri-Shomu horizon, as our friend Olympus informs

Lukasz M said...

@Matt
The same subset as your with (suppoosed)cluster's borders
https://s10.postimg.org/y66d6uv4n/pca2.png

Onur said...

Back again after several days of break from any blog activity!

@Shaikorth

Yet those "relations", such as the Turkmen v Chuvash one, turned out quite illusory once tested with D-stats and didn't show up if you correct for drift just like the Ari Blacksmith component in Somalis, so we don't have a particular reason to care about them.

Like I said, the Proto-Turkic genetic element is much better represented by Mongols and Tuvans than by Yakuts, so those D-stats do not demonstrate anything. Also, it is dubious that Nganasans represent the Proto-Uralic genetic element any better.

A component peaking in Kalash can be all over South Central Asia and India, and so can a Burusho component. When Hazaras or Sindhis show a Burusho component we can check whether they really have Burusho admixture with Finestructure. Broushaki's fits with Human Origins suggest they don't.

The Kalash component, when found in non-Kalash populations, is usually found
in very small levels in non-Kalash populations (most of which are geographically close to Kalash).

http://blogs.discovermagazine.com/gnxp/2012/02/the-kalash-in-perspective/#.V_szLOWLSM-

https://violentmetaphors.files.wordpress.com/2014/05/structure1.jpg?w=636

The Yakut and/or Nganasan components, on the other hand, are usually found in relatively high levels in non-Yakut and non-Nganasan Altaic and/or Uralic populations of Siberia and Central Asia and show some correlation with Altaic and Uralic ancestry respectively. That hardly indicates that the distribution of those components can be totally explained by relatively recent genetic drift.

http://oi63.tinypic.com/fk1c92.jpg

http://4.bp.blogspot.com/_Ish7688voT0/TNY8EEnjDJI/AAAAAAAAC18/DL4nRsu8hy0/s1600/ADMIXTURE7.jpeg

In line with the Falush preprint "use [of algorithms like STRUCTURE] should represent the beginning of a detailed demographic...analysis, not the end". If a more detailed look reveals a component doesn't actually represent a contributing population there's no reason to keep chasing such ghosts.

On that issue Matt has already replied to you and I agree with his assessment.



Shaikorth said...

Onur

" That hardly indicates that the distribution of those components can be totally explained by relatively recent genetic drift."

But it can be, just like with Ari Blacksmith admixture in Somalis. The difference between relation to these two groups in geographically distant populations that have different proportions of their specific has been shown to be insignificant in light of D-stats and Chromopainter-based analysis, it's irrelevant whether they are supposed to represent proto-populations or not. ADMIXTURE analysis that shows them as widespread is in all likelihood not showing real ancestry, as is the case with the Cultivators or Burusho component in a K=15 run here:


POPULATION
SE_Asian
Burusho
Balochi
Siberian
Kalash
Caucuses
SW_Asian
E_Asian
Paniya
E_African
WHG
NE_Indian_Tribal
EEF
Onge
W_African

Armenian
0.13%
4.49%
5.27%
0.07%
0.94%
49.85%
21.80%
0.20%
1.67%
0.02%
1.21%
0.47%
13.18%
0.70%
0.00%


Sindhi
0.64%
28.67%
28.07%
0.52%
2.85%
8.65%
1.98%
0.62%
1.33%
0.39%
4.54%
19.57%
1.44%
0.64%
0.12%

etc.

Obviously Armenians and Sindhi have no Burusho ancestry, for example something like Baloch+Gujarati is a way better fit for Sindhi.

I replied to Matt about the matter already so I don't repeat that. Suffice to say that from Flegontov et al. 2016 and Broushaki table S24-I and II it looks like Nganasans, Yakuts and Dolgans are in fact mostly similar populations as Dolgans donate almost all the former's ancestry, with a bit less than 10% Selkup contribution in Nganasans reflecting their language's Samoyedic origin being the main difference. If they form a full differentiation in ADMIXTURE, the most likely reason is drift.

Onur said...

@Shaikorth

But it can be, just like with Ari Blacksmith admixture in Somalis. The difference between relation to these two groups in geographically distant populations that have different proportions of their specific has been shown to be insignificant in light of D-stats and Chromopainter-based analysis, it's irrelevant whether they are supposed to represent proto-populations or not. ADMIXTURE analysis that shows them as widespread is in all likelihood not showing real ancestry, as is the case with the Cultivators or Burusho component in a K=15 run here...

ADMIXTURE results are not complete trash unlike what you and your guru Daniel Falush make them out to be, that is all I am saying. I am not an ADMIXTURE literalist, never have been.

Suffice to say that from Flegontov et al. 2016 and Broushaki table S24-I and II it looks like Nganasans, Yakuts and Dolgans are in fact mostly similar populations as Dolgans donate almost all the former's ancestry, with a bit less than 10% Selkup contribution in Nganasans reflecting their language's Samoyedic origin being the main difference.

How can Dolgans contribute to Yakuts and Nganasans when it is well known historically that Dolgans emerged from Yakuts and are a younger population than both Yakuts and Nganasans? You certainly should not take haplotype-based analysis (e.g., fineSTRUCTURE and ChromoPainter) results literally. I do not take ADMIXTURE results literally but instead try to interpret them in the most rational way as possible, you should do the same for haplotype-based analysis results.

If they form a full differentiation in ADMIXTURE, the most likely reason is drift.

The most logical explanation for their ADMIXTURE results is drift + ancient genetic structure, not just drift.

Shaikorth said...

"ADMIXTURE results are not complete trash unlike what you and your guru Daniel Falush make them out to be, that is all I am saying."

So all you're saying is aimed against some strawman? If ADMIXTURE results were complete trash they could never be expected to differentiate Pygmies from Bantus, and no one is claiming something like that.

"How can Dolgans contribute to Yakuts and Nganasans when it is well known historically that Dolgans emerged from Yakuts and are a younger population than both Yakuts and Nganasans? You certainly should not take haplotype-based analysis (e.g., fineSTRUCTURE and ChromoPainter) results literally. I do not take ADMIXTURE results literally but instead try to interpret them in the most rational way as possible, you should do the same for haplotype-based analysis results."

Dolgans get major donation from Yakuts too, this kind of analysis doesn't tell which came first just like ADMIXTURE doesn't. The rational interpretation is that these donate to each other because they are similar populations whose differentiation in ADMIXTURE is mainly based on drift. It's similar to how Armenian subpopulations donate mainly to each other, or how Anatolian Turkish subpopulations donate mainly to each other. If someone oversamples either of those to create multiple ADMIXTURE clusters, they'd be also primarily drift-based.

Onur said...

@Shaikorth

So all you're saying is aimed against some strawman? If ADMIXTURE results were complete trash they could never be expected to differentiate Pygmies from Bantus, and no one is claiming something like that.

Differentiating Pygmies from Bantus is a quite easy task. That is not something enough to make a genetic analysis method worth implementing. You are still underestimating ADMIXTURE's abilities.

Dolgans get major donation from Yakuts too, this kind of analysis doesn't tell which came first just like ADMIXTURE doesn't. The rational interpretation is that these donate to each other because they are similar populations whose differentiation in ADMIXTURE is mainly based on drift. It's similar to how Armenian subpopulations donate mainly to each other, or how Anatolian Turkish subpopulations donate mainly to each other. If someone oversamples either of those to create multiple ADMIXTURE clusters, they'd be also primarily drift-based.

Good to see that you do not take haplotype-based analysis results literally.

Shaikorth said...

Onur

"You are still underestimating ADMIXTURE's abilities."

Nope. If you know what its limits are you won't be ascribing undue significance to Ari Blacksmith, Burusho, Nganasan or Yakut "components" appearing outside their immediate neighbourhood.

Onur said...

@Shaikorth

Nope. If you know what its limits are you won't be ascribing undue significance to Ari Blacksmith, Burusho, Nganasan or Yakut "components" appearing outside their immediate neighbourhood.

Like I said many times, the Nganasan and Yakut components do not indicate that populations that carry them have ancestry from Nganasans or Yakuts or something similar to them, but rather they point to some more distant shared origins after we account for the parts that can be explained by recent drift.

Shaikorth said...

Neither D-stat comparison of "Yakut"-showing Turkmens and "Nganasan"-showing Chuvash or haplotype analysis pointed towards them having unique affinities related to these two Siberian pops beyond drift. This indicates that Nganasans and Yakuts are similar to each other and that their components mask the actual South Siberian affinity of Western Turkics (just like Ari components in Afars and Somalis hide other East African affinity and Burusho in Sindhis other South Asian). John Novembre has a good related paragraph in his new article about STRUCTURE (also applicable to ADMIXTURE).

"Another challenge is that STRUCTURE has become, in some sense, a victim of its own success. It is applied by default in most studies without consideration of whether the underlying model is relevant. *For example, if applied to a geographic continuum, the method will infer source populations that are vaguely spatial but have no real interpretation as source populations in an admixed sample* (e.g., Witherspoon et al. 2007). A recent paper captures the care needed with its colorful title: “A tutorial on how (not) to over-interpret STRUCTURE/ADMIXTURE bar plots” (Falush et al. 2016). All this being said, the need for careful interpretation is ubiquitous in population genetics, and the extra attention on the STRUCTURE method is warranted because of its widespread use." (emphasis mine)

Onur said...

@Shaikorth

Of course ancestors of Nganasans and Yakuts came from more southern parts of Siberia, of course they got admixed and drifted in their new lands, and of course they do not represent the Proto-Uralic and Proto-Turkic populations respectively, I have never denied these. But they did not come from the same region of southern Siberia and do not currently live in the same region of Siberia either. Nganasans are from a more western part of Siberia and this reflects the fact that Proto-Uralics lived in a more western region than Proto-Turkics. The ADMIXTURE results show this distinction clearly and have some correlation with the language family, so there is no reason for saying that they are totally garbage, at least in this case.

Shaikorth said...

Onur, this isn't about Nganasans and Yakuts representing any proto-populations, but their supposed components' relevancy in more distant regions. It's clear from GLOBETROTTER of Busby et al. that the eastern ancestry of Chuvash and Turkmens (and that of Mordvins etc.) came from the same region of Southern Siberia, and drifted Siberian clusters like Nganasan/Yakut in ADMIXTURE obscure this just like a Burusho cluster can obscure Sindhi population's ancestry or Ari Blacksmith cluster can obscure Afars.

Onur said...

@Shaikorth

The differences in the East Eurasian ADMIXTURE components of Uralic and Turkic populations cannot solely be explained by recent drift in some populations. I have a suggestion to you: do an ADMIXTURE run up to meaningfully high K levels without the populations that you think are highly drifted and then share the results (for each K level) here.

Shaikorth said...

Onur, you've already been given several examples of drift-bases admixture clusters popping up, and examples of scientists explaining that this can happen. ADMIXTURE can't disprove formal stats or haplotype-based methods.

Look up Chuvash, Turkmens and Mordvins, and their admixing source pops here:
http://www.cell.com/cms/attachment/2062326182/2064072288/mmc2.xlsx

Onur said...

@Shaikorth

Onur, you've already been given several examples of drift-bases admixture clusters popping up, and examples of scientists explaining that this can happen. ADMIXTURE can't disprove formal stats or haplotype-based methods.

It is you who is seeing a contradiction between the ADMIXTURE results and the haplotype-based method and formal stat results, not me, and I do not try to explain away the ADMIXTURE results solely by recent drift.

Look up Chuvash, Turkmens and Mordvins, and their admixing source pops here:
http://www.cell.com/cms/attachment/2062326182/2064072288/mmc2.xlsx


Look up Finns in the same spreadsheet, Mordovians are among the contributing populations to them in some of the GLOBETROTTER analyses. You should of course not take those results literally, but they give you some ideas about the East Eurasian ancestry in Uralic populations. Mordovians are surely not a recent population and did not acquire their East Eurasian ancestry largely or totally from Turkics, rather, they inherited it from their Proto-Uralic ancestors by and large.

Shaikorth said...

"It is you who is seeing a contradiction between the ADMIXTURE results and the haplotype-based method and formal stat results, not me, and I do not try to explain away the ADMIXTURE results solely by recent drift."

You said yourself you couldn't make sense of the Chuvash and Turkmen D-stat comparison. That's because there is a contradiction, and methods less sensitive to drift take precedence.

"Mordvins are surely not a recent population and did not acquire their East Eurasian ancestry largely or totally from Turkics, rather, they inherited it from their Proto-Uralic ancestors by and large."

Mordovians donate more to the Ukrainian cluster than to Finns, they also donate to Norwegians, Orcadians etc in similar amounts, they're mostly North European. The Siberian donor for Finns, and to a lesser degree in Norwegians is the NorthSiberian (Ket/Selkup) cluster in all Busby runs.

The Volga populations and Turkmens have theirs from the South Siberia-centric Mongolian cluster. The dates for the admixture in Mordovians implies it's recent, just as recent as for Chuvash. It definitely isn't Bronze Age, if it's not related to the first Turkic expansions it's something that happened shortly before it and was similar genetically.

Onur said...

@Shaikorth

You said yourself you couldn't make sense of the Chuvash and Turkmen D-stat comparison. That's because there is a contradiction, and methods less sensitive to drift take precedence.

It is because Yakuts and Nganasans are not main contributors or close in genetics to the main contributors to other Turkic or Uralic populations. They are surely new, admixed (in Yakuts' case, they mostly admixed with Tungusic populations, who have a similar East Eurasian ancestry due to their common Altaic connections) and drifted populations, but those are not enough to mask their ancient connections with other Turkic or Uralic populations. ADMIXTURE has the sensitivity to detect those ancient connections once we account for the part of the ADMIXTURE results explainable by other factors such as drift.

Mordovians donate more to the Ukrainian cluster than to Finns, they also donate to Norwegians, Orcadians etc in similar amounts, they're mostly North European. The Siberian donor for Finns, and to a lesser degree in Norwegians is the NorthSiberian (Ket/Selkup) cluster in all Busby runs.

The Volga populations and Turkmens have theirs from the South Siberia-centric Mongolian cluster. The dates for the admixture in Mordovians implies it's recent, just as recent as for Chuvash. It definitely isn't Bronze Age, if it's not related to the first Turkic expansions it's something that happened shortly before it and was similar genetically.


Mordovians appear along with Germans among the contributing populations to Finns in those GLOBETROTTER analyses. So obviously Mordovians represent all of the East Eurasian ancestry of Finns along with other ancestries given the fact that Germans lack East Eurasian ancestry.

Shaikorth said...

"ADMIXTURE has the sensitivity to detect those ancient connections once we account for the part of the ADMIXTURE results explainable by other factors such as drift."

No, it just creates Nganasan and Yakut drift clusters that hide their primarily shared ancestry.

For the record, Mordovians are primarily North European and it's impossible to confuse their linked haplotypes for West Siberians or East Siberians. Mordovian-related admixture events detected in some runs represent some kind of intra-North European admixture (especially given that they're dated to 1500's in Orcadians that get them and 1700's in Finland) and Siberian-related ones are represented by the appropriate Siberian cluster.

Angantyr said...

Regarding Mordvins, from http://terheninenmaa.blogspot.se/2016/10/worldwide-admixture-analysis-based-on.html posted today:

"Mordvins seem to differ from other Volga-Finnic populations and belong to Balto-Slavic ancestry and they probably are language shifters from a Baltic to a Volga-Finnic language."

Onur said...

@Shaikorth

No, it just creates Nganasan and Yakut drift clusters that hide their primarily shared ancestry.

You have not demonstrated that the distribution of those clusters are TOTALLY due to drift in Nganasans and Yakuts.

For the record, Mordovians are primarily North European and it's impossible to confuse their linked haplotypes for West Siberians or East Siberians. Mordovian-related admixture events detected in some runs represent some kind of intra-North European admixture (especially given that they're dated to 1500's in Orcadians that get them and 1700's in Finland) and Siberian-related ones are represented by the appropriate Siberian cluster.

Never said Mordovians are not primarily North European in ancestry. What part of my statement "Mordovians appear along with Germans among the contributing populations to Finns in those GLOBETROTTER analyses. So obviously Mordovians represent all of the East Eurasian ancestry of Finns along with other ancestries given the fact that Germans lack East Eurasian ancestry" is so hard to understand?

@Angantyr

Regarding Mordvins, from http://terheninenmaa.blogspot.se/2016/10/worldwide-admixture-analysis-based-on.html posted today:

"Mordvins seem to differ from other Volga-Finnic populations and belong to Balto-Slavic ancestry and they probably are language shifters from a Baltic to a Volga-Finnic language."


To a large extant they are. But they also have some Proto-Uralic ancestry that they share with other Uralic populations and non-Uralic populations with some Uralic ancestry (e.g., northern Russians, Chuvash, Tatars, Bashkirs).

Shaikorth said...

Onur,

"You have not demonstrated that the distribution of those clusters are TOTALLY due to drift in Nganasans and Yakuts."

D-stats show no difference between Turkmen and Chuvash relation to those pops, Globetrotter shows no difference. That's more than enough evidence that their distribution in geographically distant regions is drift-based, perhaps you just don't like what you see.

Also, you don't undestand the underlying methods in the Busby paper. When available donors include populations from Europe and Siberia, Mordovian haplotypes do not and can not proxy for Siberian, only for North European. Siberian-related ancestry events are represented by Siberians and they are listed on the sheet. Mordovian-German event in Finland (8%) in 1700's and Mordovian-Orcadian event of Orcad5 (44%) in 1500's represent intra-North European population structuring, and not admixture significant to population history, not related to Siberians or even actual Mordovians who never were in Finland, not to mention the Orkneys. Same for the Lithuanian-Spanish event in Finland dated to 1800's or the Sicilian-Norwegian event in Orkneys dated to 1900's. You shouldn't make crazy theories about these.

Onur said...

@Shaikorth

D-stats show no difference between Turkmen and Chuvash relation to those pops, Globetrotter shows no difference. That's more than enough evidence that their distribution in geographically distant regions is drift-based, perhaps you just don't like what you see.

Also, you don't undestand the underlying methods in the Busby paper. When available donors include populations from Europe and Siberia, Mordovian haplotypes do not and can not proxy for Siberian, only for North European. Siberian-related ancestry events are represented by Siberians and they are listed on the sheet. Mordovian-German event in Finland (8%) in 1700's and Mordovian-Orcadian event of Orcad5 (44%) in 1500's represent intra-North European population structuring, and not admixture significant to population history, not related to Siberians or even actual Mordovians who never were in Finland, not to mention the Orkneys. Same for the Lithuanian-Spanish event in Finland dated to 1800's or the Sicilian-Norwegian event in Orkneys dated to 1900's. You shouldn't make crazy theories about these.


I have already explained to you in detail why I do not agree with you on these points, so I won't repeat. Plus, this discussion has been going on for too long (in fact, we made a similar discussion several months ago in this blog) with no prospect of compromise between you and me anytime soon (at least until new ancient DNA data favor either one of the two sides), so we'd better give it a rest for now.