search this blog

Monday, February 25, 2019

All quiet on the eastern front


I put together this quick and dirty qpGraph tree just to double check what the Eneolithic trio from the Piedmont steppe (Piedmont_Eneolithic) were roughly made of, and how they related to some of the other populations from the eastern half of ancient West Eurasia. The relevant graph file is available here.


Yep, the tree basically lines up with scientific literature. In other words, Piedmont_Eneolithic appears to be a two-way mixture of populations very closely related to Caucasus and Eastern European Hunter-Gatherers (CHG and EHG, respectively). Good to know.

By the way, please note that Piedmont_Eneolithic is made up of samples from two different locations on the Piedmont steppe, and I occasionally treat them as separate populations labeled Progress_Eneolithic and Vonyuchka_Eneolithic (for instance, see here).

Update 28/02/2019: Below is a new version of the tree designed specifically to investigate whether the ancestry of Piedmont_Eneolithic can be modeled with admixture from Darkveti-Meshoko, a population from the Caucasus roughly contemporaneous with Piedmont_Eneolithic. This doesn't appear to be the case, at least not with this topology, because the mixture edge from the Darkveti-Meshoko-related D7 node to the Piedmont_Eneolithic-related E4 node is marked with a zero. The relevant graph file is available here.


See also...

Big deal of 2018: Yamnaya not related to Maykop

Yamnaya isn't from Iran just like R1a isn't from India

Late PIE ground zero now obvious; location of PIE homeland still uncertain, but...

74 comments:

Andrzejewski said...

CHG + EHG: it's very similar to the Yamnaya and other Steppe populations before the late admixture with Cucuteni Tripolye and GAC farming populations to their immediate west.

PF said...

Can you add MA1?

I’ve been thinking a bit about CHG origins, mostly trying to incorporate the qpGraph and other models from the recent major preprints (Dzudzuana and Yana). Gut feeling is that something weird is going on. What we can gather right now:
- Dzudzuana at ~26k years ago has no CHG. Dzudzuana is less than 50 km from where what we call CHG was found.
- MA1 at ~24K years ago in southern Siberia has something CHG-related, while other ANE samples, both earlier and more recent, apparently do not.
- According to the model in the Yana paper it looks like the non-Basal portion of proto-CHG contributed to MA1.

Taking the model in the Yana paper literally, I’d like to very speculatively propose another Unknown HG population that’s relevant to CHG formation — something that existed in Central Asia a very long time ago. I’ll just crudely call it UCAHG (Unknown Central Asian HG). Just thinking out loud here.

In this framework CHG mostly = UCAHG + ANE + Dzudzuana. Some models in the Dzudzuana paper also require extra ENA for both CHG and Iran_N which is another interesting hint and at the very least points to the east as the relevant location for the formation of these groups.

So CHG is then just the western endpoint along a continuum of some central Asian population that admixed with a Basal-rich population (whether with Dzudzuana or something similar), while MA1 is representative of this population at its eastern end mixing with Yana? This leaves the possibility that there is UCAHG ancestry (CHG minus Basal) all through central Asia / Steppe.

Obviously this all rests on the CHG-related input into MA1 found in the Yana paper which could simply be an artifact of their model. Of course it is related to the to the current discussion: CHG-related admixture could be spread throughout the Steppe since a very long time ago with obvious implications for EHG, Yamnaya, etc. That is, the ghost UCAHG population is inflating CHG estimates for these groups because it needs its UCAHG portion to produce better fits. On the other hand the Yana paper argues that EHG indeed has direct CHG input and consequently low levels of Basal Eurasian. However the amount of CHG in EHG by multiple measures seems like it would produce more Basal than the weak signals actually seen in EHG, thus, it makes the case that there is indeed some CHG-related layer in EHG but with a lower or zero Basal component.

Also, I’m not really well-versed in uniparentals, but what does come to mind is mtDNA X2 which is represented quite heavily in Native Americans and also in specifically CHG-rich populations. That surely needs some explanation.

Slumbery said...

I played a bit with these ancient populations in G25 nMontes just to see what comes up. This is not meant to be a precise ancestry simulation and a fits are sometimes so horrible it is best to not mention them. I was just after some big differences.

Global25 nMontes CHG
AfontovaGora3 0%
Boncuklu_N 50.83%
MA1 43.33%
Natufian 0%
Ust_Ishim 5.83%
Villabruna 0%

Global25 nMontes Hotu_HG
AfontovaGora3 0%
Boncuklu_N 24.17%
MA1 50%
Natufian 7.5%
Ust_Ishim 18.33%
Villabruna 0%

Global25 nMontes EHG:Average
AfontovaGora3 69.17%
Boncuklu_N 2.5%
MA1 0%
Natufian 0.83%
Ust_Ishim 0%
Villabruna 27.5%


What I found interesting (but probably not new for some people here):
1. The ANE ancestry in CHG+Iran HG vs. EHG seems to be very different. CHG-Iran completely rejects AG3 in favor of MA1, while EHG is the opposite.
2. CHG and Hotu HG have difference in relatedness to Ust_Ishim, but not in the way that would be expected based on the above qpGraph tree.

Matt said...

Agree AG3 or MA-1 should probably be used in the model if poss, at the moment EHG modeled by 20:80 IranN:IronGates, or 9:11:80 Basal:OtherHG:IronGates which is a bit odd.

Tried to make a tree that includes AG-3, hopefully the nodes are not wrong: https://pastebin.com/mFcdrezY

No specific drift needed between CHG and Piedmont_En is also a little odd, although I guess maybe this is as the model resolves that via how it specifies Basal Eurasian and C1?

Davidski said...

@Matt

fatalx:
> 2 edges at C

Matt said...

God, I hate writing qpGraphs. OK, that really wasn't a good tree, let's try this again: https://pastebin.com/XDL9Ekaj

Also Davidski, I've tried to replicate a close version of the trees from Wang's supplement, where Maykop gets curiously low levels of Anatolian (90:10 CHG:Anatolian). Changes are, using Darkveti_Meshoko, added a more direct Anatolian reference than GAC, and shifted about some features of the topology that made not much sense to me. Does this run?: https://pastebin.com/XVGDKBQJ

Davidski said...

@Matt

https://drive.google.com/open?id=1VDtOJeWju1Ry-SYLErmRHfpgNqWtV0p3

outliers:
Mbu Iro CHG Pie 0.018223 0.013914 -0.004309 0.001605 -2.685
Mbu Iro EHG Pie -0.016555 -0.019806 -0.003251 0.001411 -2.304
Ust CHG Iro Afo 0.003484 -0.002578 -0.006063 0.002732 -2.219
Ust Iro CHG Gan -0.004000 -0.007775 -0.003775 0.001816 -2.079
Ust Iro CHG Pie 0.015762 0.011300 -0.004462 0.001969 -2.266
CHG Gan EHG Afo -0.000050 0.004528 0.004578 0.002142 2.138
CHG Gan Iro Afo -0.000511 0.005783 0.006295 0.002078 3.029
CHG Iro EHG Iro 0.051345 0.048322 -0.003023 0.001480 -2.043
CHG Iro Afo Pie -0.006015 -0.010632 -0.004617 0.002210 -2.089
CHG Pie Gan Iro 0.022224 0.017756 -0.004468 0.001722 -2.594
CHG Pie EHG Iro -0.006944 -0.010942 -0.003998 0.001550 -2.580
CHG Pie Iro Afo 0.006805 0.015099 0.008294 0.002241 3.700
CHG Pie Iro Pie 0.110419 0.114406 0.003988 0.001740 2.292
Gan Iro EHG Pie -0.025914 -0.029774 -0.003860 0.001430 -2.699
EHG Pie Iro Afo -0.007420 -0.001743 0.005677 0.002195 2.586
Iro Afo Iro Pie 0.079659 0.083808 0.004148 0.001847 2.246
Iro Pie Iro Pie 0.189288 0.193747 0.004459 0.001839 2.425

worst f-stat: CHG Pie Iro Afo 0.006805 0.015099 0.008294 0.002241 3.700

Matt said...

Ah, sigh. OK, let's try reversing the split order for ANE, having it on a clade with the HG element in IranN+CHG clearly breaks the greater affinity of ANE to Piedmont over CHG, even allowing for extra basal in CHG: https://pastebin.com/xx6fLeLC

Did the Darkveti_Meshoko+GAC+Yamnaya+Piedmont fit work at all?

JuanRivera said...

Iran_N ancestry in EHG? Now that's a novel proposal.

Davidski said...

@Matt

https://drive.google.com/open?id=1n5E8Lb28rbg46ej1ZfKaszQeaRVvD0E5

worst f-stat: Los GAC Yam Pie -0.002974 0.004181 0.007155 0.001341 5.335


https://drive.google.com/open?id=1MbH0HAsxtlJiqQ_dFACDVuBRsy2MWlcy

worst f-stat: Gan Iro EHG Pie -0.024194 -0.029774 -0.005580 0.001430 -3.901

Davidski said...

@JuanRivera

Iran_N ancestry in EHG? Now that's a novel proposal.

It's Iran_N-related ancestry, which is basically the same thing as CHG-related ancestry. It's just a sister clade more or less somewhere between the two.

But that edge from C3 to D2 has to be there to differentiate EHG from Iron_Gates_HG, which has much less of that type of eastern influence.

The really surprising thing about this is how it helps to keep the most significant Z score so close to zero, so it has to mean something.

JuanRivera said...

So, it's maybe the CHG-related but with more basal and ANE ghost population proposed here.

Open Genomes said...

@David

How might this Iran Neolithic-like ancestry found in EHG relate to the Iranian Hunter-Gatherer from Hotu Cave?

Isn't there extra Iran-like ancestry in the Eneolithic Steppe that isn't in EHG?

Where was this Iran-like ancestral population, if they weren't in the Caucasus? East of the Caspian?

The distances are "high", but not huge like with CHG KK1, or the Natufians:

Neolithic and earlier ancestry composition of sample: VJ1001 Population: Vonyuchka_Eneolithic Chalcolithic Steppe

Ganj Dareh Neolithic 20.4%

Can you run some formal stats and a qpAdm on VJ1001?

By way of comparison, here is a true fundamental LGM population isolate:

Neolithic and earlier ancestry composition of sample: KK1 Population: CHG Mesolithic Caucasus

The distance is 15.46% for Kotias vs. 4.00% for Eneolithic Steppe Vonyuchka VJ1001.

Here is an EHG:
Neolithic and earlier ancestry composition of sample: I0124 Population: EHG Mesolithic Europe

Distance 3.50%, no Iran-like ancestry, only 1.0% CHG.

Here's one of the Ukraine Mesolithic individuals:

Neolithic and earlier ancestry composition of sample: I1763 Population: Ukraine_Mesolithic Mesolithic Steppe

Several Mesolithic relatives and Neolithic descendants, but no sign of Iran-like ancestry there either.

Chad Rohlfsen said...

Iran and CHG are very different. Add in West Siberia, Anatolia_N and Han. You'll see. They can't be made off the same branch. Adding Caucasus Eneolithic makes it more interesting... ;)

Lee Albee said...

@of

There is no good explanation for x2 in native Americans. Specifically ASO related native american. The x2 in them is not really related to x2 anywhere in Eurasia. It is x2a and x2g.

In chg heavy populations x2f, x2e are heavy.

X2c and x2mn were probably neolithic.

X2bd. Is less well associated maybe paleolithic but could also be bronze age.

X2 did originate in the LGM. Possibly in southern Arabia and could possibly be basal eurasian related.

Davidski said...

@Chad

So did you take a moment to think about Yamnaya_Caucasus SA6010?

This sample is practically identical to Vonyuchka_Eneolithic, and has zero Ukraine_N ancestry. What do you think that means?

A) Repin was like this but with significant Ukraine_N ancestry

B) Repin was like this and with zero Ukraine_N ancestry

I'm gonna go with B, do you know why? Because nothing really stands out about SA6010 as a Yamnaya individual, except that he lacks the MN farmer ancestry that other Yamnaya samples have in varying degrees.

In other words, this is what proto-Yamnaya was like.

Slumbery said...

@JuanRivera

"So, it's maybe the CHG-related but with more basal and ANE ghost population proposed here."

Yeas, it is closer to CHG than to Hotu HG. In nMontes a tree way mixture of WHG + AG3 + CHG gives a better fit than WHG +AG3 + Hotu HG and even just adding Hotu HG as a fourth worsens the fit.

Open Genomes said...

@David

Like Vonyuchka VJ1001, Yamnaya Caucasus SA6010 has a substantial percentage of Iran-like ancestry:

Chalcolithic and earlier ancestry composition of sample: SA6010 Population: Yamnaya_Caucasus Bronze Age Steppe

Total Iran-like ancestry 24%.

This one too is unusual with the presence of some kind of Iran-like ancestry, above and beyond the CHG contribution which is not already present in Khvalynsk Eneolithic.

BTW, he seems to be yet another R1b-V1636 like the two Progress Eneolithic individuals (he derived for 3 R-V1636 derived SNPs, V2146, BY21441, and BY15348, and he's also P297-), and he's mtDNA U5a1g, which is found in Iran among other places.

Davidski said...

@Open Genomes

Yamnaya_Caucasus SA6010 is just more basal and ANE-rich than other Yamnaya, that's why he scores Iran_N with Global25/nMonte.

And I don't think any subclades of U5a are native to Iran. There's no U5a in any ancient Iranian samples.

Lenny Dykstra said...

@PF:

"So CHG is then just the western endpoint along a continuum of some central Asian population that admixed with a Basal-rich population (whether with Dzudzuana or something similar), while MA1 is representative of this population at its eastern end mixing with Yana? This leaves the possibility that there is UCAHG ancestry (CHG minus Basal) all through central Asia / Steppe."

I basically proposed this around 18 months ago in this thread, tho I called this ghost pop "Ancient Central Eurasian" (ACE): https://anthrogenica.com/showthread.php?11513-Pseudo-ANE-derived-from-Ancient-Central-Eurasian-ghost-population&highlight=ancient+central+eurasian

Later on in that thread I think I went off in a tangent speculating about Basal Eurasian. But I still think I was onto something with the OP.

Think of it like this: we have the East-West Eurasian split around 50kya, but you can infer a further split ]of an "Ancient South Eurasian" population from the main East Eurasian trunk shortly after the East-West split, and ANE splits from the West Eurasian branch apparently around 35-40kya, judging by Yana RHS.

So beginning ~35kya we have ANE holding down central and northern Siberia, Ancient West Eurasians (Gravettians) anchored west of the Urals and north of the Caucasus, ASE was probably already dispersed around the Bay of Bengal, Sundaland, and Sahul, and the "main" East Eurasian branch presumably anchored around mainland China (?).

What about the middle of this territory--modern SW Siberia, Kazakhstan, Turan, etc? Afontove Gora is close to this area but he's "too late" at 17,000 kya (after ANE moved south) and Ust Ishim is "too early" at 45kya. Central Asia was probably a transitory zone after the East/West/North/South Crown Eurasian splits. It was not home to one solitary population but a population continuum that was more "Gravettian-like" at its western fringe, more "ANE-like" at its northern fringe, ASE or AEE-like at its eastern fringe etc...

JuanRivera said...

ANE and ACE may have been admixing with each other through tens of thousands of years. In fact, you proposed that ACE was a sibling of ANE. As for ANE, it also was in Southern Siberia (except SW Siberia).

Open Genomes said...

@David - So Iran HG is just CHG-like + Basal + ANE?

Why not say that this kind of Iran HG-like population was also somewhere east and northeast of the Caspian, as well as south of it, like Hotu?

It wasn't as far north and east as West Siberian Neolithic / Botai, because the Basal Eurasian ancestry was missing there, but this Iran HG-like population existing around the entire eastern shore of the Caspian during the LGM seems right to me. Unlike the Caucasus, there were no barriers to movement in either direction.

I think that the most parsimonious solution here is to say that the Iran HG-like population spread out around the south, east, and maybe north Caspian during the LGM, rather than proposing a novel Mesolithic population with an "identical mix" of components. Try modeling everything with Hotu if possible, and see what happens.

Davidski said...

@Open Genomes

Why not say that this kind of Iran HG-like population was also somewhere east and northeast of the Caspian, as well as south of it, like Hotu?

This is unlikely, considering the population history of the Samara region, where people mostly of EHG origin still lived as late as the Eneolithic.

The place to look for the CHG-like and Iran_N-like hunter-gatherer population that contributed most of the ancestry to the Piedmont steppe, Khvalynsk, Yamnaya, Afanasievo and similar groups is the southernmost steppe region, simply because the Piedmont Eneolithic people were still foragers and native to this region.

Their ancestors were probably Mesolithic foragers who lived in that part of the steppe, and belonged to a clade that was closely related to CHG and Iran_N, but with extra ANE.

Open Genomes said...

The Caspian and Arals Seas seem to have been much greater in extent during the LGM.
There was even a narrow land corridor between the Caspian and Aral Seas, narrower than the Caucasus, which could have served as a refugium for an LGM population:

Extent Of Aral/Caspian Seas At The LGM

Davidski said...

Maybe, but the steppes and swamps just east of the Caspian were probably populated by a Botai-like people with varying levels Iran_N ancestry, because there are two Bronze Age outlier samples from Gonur that are like this. They may have been migrants from the nearby Kelteminar culture.

So it's very difficult to see the homeland of the Piedmont steppe population east of the Caspian, even in part.

Samuel Andrews said...

Open Genomes: "BTW, he seems to be yet another R1b-V1636 like the two Progress Eneolithic individuals "
Davidski: "Because nothing really stands out about SA6010 as a Yamnaya individual, except that he lacks the MN farmer ancestry that other Yamnaya samples have in varying degrees."

Which means Yamnaya_Caucasus_o:SA6010 is a continuation of Progress_Eneolithic.

Davidski said...

@Samuel Andrews

Which means Yamnaya_Caucasus_o:SA6010 is a continuation of Progress_Eneolithic.

Not exactly.

The Progress site is on the steppe right near the Caucasus, and this individual is from a site well to the north of the Progress site. But even in this more northern area Steppe Maykop was present before Yamnaya came in.

So SA6010 looks like a return of the pre-Steppe Maykop Eneolithic genotype to the region from somewhere else, probably the Don steppe to the north, which is very likely where Yamnaya originated.

Matt said...

@Davidski:
https://drive.google.com/open?id=1n5E8Lb28rbg46ej1ZfKaszQeaRVvD0E5
worst f-stat: Los GAC Yam Pie -0.002974 0.004181 0.007155 0.001341 5.335


OK, thanks, so in general, proportions look pretty much as expected here:

Darkveti_Meshoko 72:28 CHG:Anatolian
GAC 30:70 WHG:Anatolian
Yamnaya 88:12 Piedmont_Steppe:European_MN

I think that confirms to me that the high CHG in the trees from Wang 2019 was an effect of modelling Maykop as European_MN+CHG, which apparently worked for them but I would guess needed to up CHG to offset WHG balance in European_MN.

Still, it seems like there are some strange long branches in the upper levels of the graph, and I'm not 100% on how it models Basal Eurasian (lack of Ust_Ishim, and Chad has discussed on his blog an issue of unexplained attraction of Anatolian farmers to Mbuti that may cause some slight error in estimation as well).

Another worry is worse stat suggests that this topology over-states the relatedness of GAC and Yamnaya, relative to Loschbour. That suggests GAC is perhaps not the mixing population and may be another with slightly different balance of Anatolian and WHG related streams (which possibly supports ideas from earlier Copper Age SE+Ukraine European populations instead).

Could you try a couple of variant topologies:

A: https://pastebin.com/TUJPuGNd - adds Chimp in place of Mbuti and Ust_Ishim to topology

B: https://pastebin.com/rQzwWZ2S - Model A above with a trifurcation between Anatolian ancestry in Darkveti_Meshoko, GAC and Yamnaya, and separate HG pulses to GAC and Yamnaya's Anatolian ancestry.

If no Chimp, can you try anyway with Mbuti? I'm not sure will get the relative Basal Eurasian levels right but could still show if either the trifurcation of Anatolian ancestry or adding UI helps.

Matt said...

@Davidski:
"https://drive.google.com/open?id=1MbH0HAsxtlJiqQ_dFACDVuBRsy2MWlcy
worst f-stat: Gan Iro EHG Pie -0.024194 -0.029774 -0.005580 0.001430 -3.901 "


It seems like the problems around this one are that the fewer degrees of freedom means that a worse stat is generated wrt the relatedness between Ganj_Dareh and EHG. A pulse from GD's ancestor to EHG maybe doesn't make sense, but this becomes not a problem to resolve in the topologies in the post with more degrees of freedom without AG3.

To add more of those, possibly this will work better: https://pastebin.com/1dcUkPy3 (EHG has a separate Iran/CHG pulse that can have a variable Basal:Main ratio).

Davidski said...

@Matt

https://drive.google.com/open?id=1sX1kM60-pKQfd4-7NWEjgxGSEUPjsdLb

The other two graphs won't run, because...

fatalx:
> 2 edges at Main_Eurasian

Matt said...

There's always one.... Amended:

A: https://pastebin.com/uZ4tqfNm
B: https://pastebin.com/jiVsknXY

Aniasi said...

What technically counts as the Piedmont steppe for this purpose?

Davidski said...

@Matt

https://drive.google.com/open?id=1lWdRW36B6LfDIydoH1dDdgZuUwsyJzPJ

https://drive.google.com/open?id=1lWdRW36B6LfDIydoH1dDdgZuUwsyJzPJ

Davidski said...

@Aniasi

What technically counts as the Piedmont steppe for this purpose?

Refer to the second map in this post. The Progress and Vonyuchka sites are on the Piedmont steppe.

Genetic borders are usually linguistic borders too

Davidski said...

@All

I updated the graph in the blog post. The new graph includes AfontovaGora3.

It's a bit of a messy effort, but generally quite realistic, and it explains a lot, especially about the non-EHG ancestry of the Piedmont Eneolithic samples.

Check it out and let me know what you think. The graph file has also been updated, so if anyone wants to improve the graph or build on it that'd be great.

Slumbery said...

@Davidski

That 2:1 western vs. eastern ancestry in EHG is the opposite of the Villabruna vs. AG3 1:2 ratio given by G25 nMontes. Is this simply because of the difference in the test methods? (Alternatively I misunderstood the graph.)

Davidski said...

@Slumbery

Yeah, the graph doesn't define the mixture sources like they're defined in the Global25.

So a mixture edge from ghost pop E to EHG in the graph doesn't mean the same thing as EHG being modeled directly as WHG.

Davidski said...

So it looks like the basal stream into Piedmont Eneolithic was almost totally basal (and closely related to the basal ancestry in CHG) except with some ANE.

That's hard to comprehend, but it seems legit.

PF said...

@Lenny

Ancient Central Eurasian (ACE) is a much better acronym than I proposed so let’s stick to that. :)

Yeah, we had much of the same thoughts. It’s hard to say how much of the phenomenon is a result of an unidentified specific early Eurasian branch and how much of it is clinal relationships between known branches. To cop out a bit I’d say it’s both? But very generally these seem to be the Paleolithic dynamics at play:

Early West Eurasian (Kostenki-related?) <—> Yana (ANE) east-west cline across Eurasia
Villabruna-like <—> Basal north-south cline
ENA influences from East Asia emanating both west and north


@Matt @Davidski

The weird thing in the last graphs made for Matt and the updated graph is how they work in the absence of AG3 input into CHG and Iran (none in Matt’s graph and only 6% into Ganj in the updated graph in the post). I think this speaks to my earlier point about another central Asian ghost pop modulating the relationship (D1 in the post’s graph, though it doesn’t make really sense since it’s equally related to ANE as E).

Regarding the issue at hand, Piedmont_Neolithic, in the updated graph it comes out as (Basal+ANE)+EHG and totally ignores CHG! Again, I can’t help but consider a ghost Central Asian pop here, which also may be explaining the Iranian signals being picked up by Open Genomes with nMonte/G25. Indeed as OG suggested I think there was something east of the Caspian that had both Basal and ANE and affected CHG, Iran, and ultimately Piedmont_Neolithic.

To make the graph more complex and tease out some these ancient relationships I’d add Kostenki, Villabruna, and Tinyuan and see what happens…

Open Genomes said...

@David

Okay, I agree that the Kelteminar culture east of the Caspian was probably the source of the Botai/West Siberian Neolithic ancestry in the Gonur outliers.

The thing is that we see a rather substantial increase in Iran-like ancestry in Kura-Araxes, Early Bronze Age Iran, and even among the Amorites of Middle-Late Bronze Age Levant North (Sidon). Is this too just another ANE incursion that bypassed CHG? So how did they get their extra Basal Eurasian aside from the Anatolian and Levantine Neolithic?

What we do know is that the Amorites came from the north. ERS1790733 is a basal "Arabian" J1-FGC11, and it may be that the J1s in BMAC are in fact an early branch of J1-P58, which is found in the region today among Punjabi Jatts, among others.

Chalcolithic and earlier ancestry composition of sample: ERS1790733 Population: Levant_BA_North Bronze Age Near East

Here we see 12.2% Iran-like ancestry.

Looking at Neolithic and earlier, we see 21.0% Iran-like ancestry, aside from the 9.8% additional CHG not present in the Levantine Chalcolithic. A tiny amount of ANE-related Siberian Neolithic also shows up:

Neolithic and earlier ancestry composition of sample: ERS1790733 Population: Levant_BA_North Bronze Age Near East

Some of this extra Iran-related ancestry is coming from Darkveti-Meshoko, but some is from something else entirely.

Open Genomes said...

There's a big missing piece here:

What would the Ubaid (pre-Sumerian) and Uruk (Sumerian) Mesopotamians have looked like?

I think we can say that I1671 from Godin Tepe c. 5800 BCD was a Halafian, because the child was found in a Halaf pottery bowl, and Halaf influence extended to Godin Tepe. I1671 Iran Late Neolithic looks quite Iranian, but with 12.2% extra Levant Neolithic.

Neolithic and earlier ancestry composition of sample: I1671 Population: Seh_Gabi_LN Neolithic Iran

I1674 from Godin Tepe c. 3800 BCE may represent Ubaid ancestry.

Chalcolithic and earlier ancestry composition of sample: I1674 Population: Seh_Gabi_ChL Chalcolithic Iran

Now we have a hefty dose, 21.4%, of Hajji Firuz Chalcolithic ancestry, with the addition of 8.2% Kura-Araxes, and 7.6% of additional Anatolian Neolithic-like ancestry.

In Neolithic terms, this is the introduction of 26.6% Anatolian Neolithic ancestry compared to the Iran Late Neolithic, along with 9.6% CHG.

Neolithic and earlier ancestry composition of sample: I1674 Population: Seh_Gabi_ChL Chalcolithic Iran

@David

Is there some relationship between what's going on in the south in Western Iran, presumably under Mesopotamian and Northern Fertile Crescent influence, and what's going on in the north with the extra Basal Eurasian ancestry mixed with CHG?

Slumbery said...

@Open Genomes

As far as I know we have zero samples from the Eastern Caucasus or generally from the Western Caspian region before Kura-Araxes. That is a significant gap, so there is no need to look after exotic scenarios for now. That entire region was probably populated by a variety of related groups in the terminal UP + Mesolithic.

Samuel Andrews said...

@Davidski,
"So SA6010 looks like a return of the pre-Steppe Maykop Eneolithic genotype to the region from somewhere else, probably the Don steppe to the north, which is very likely where Yamnaya originated."

SA6010 is no closer to Yamnaya than Progress Eneolithic is. He doesn't have Yamnaya-specific drift in G25 PCA. Plus, he has the same Y DNA as Progress Eneolithic. I doubt he's recent relative of Yamnaya.

Andrzejewski said...

Is there any Botai in Sintashta? Does Andronovo contain any BMAC?

And did they finally find out the ethnolinguistic and autosomal dna of the Ubaidian/Halafians v. The Sumerians?

Matt said...

@Davidski, think those are both the same link, thanks for running. Strange that adding Ust_Ishim into the graph seems to drive it bonkers like that. From a worst score involing Yamnaya, GAC, Loschbour and Piedmont to one involving an f3 stat of some kind with CHG, Darkveti_Meshoko and Barcin. Not really sure what's going on there..

Andrzejewski said...

https://www.biorxiv.org/content/early/2016/06/16/059311.full.pdf

Lazaridis says that Levantine farmers derive 2/3 of their ancestry from Natufian and 1/3 from Anatolian farmers. Is that why they cluster quite closely on the PCA chart?

Davidski said...

@Andrzejewski

If you're going to post here please get the basics right.

There are no instances of R1b-L51 or R1a-M417 in the samples from the Khvalynsk site.

The oldest R1a-M417 is this one...

Ukraine_Eneolithic_outlier I6561, Sredny Stog II, Alexandria, 4045-3974 calBCE

Davidski said...

And R1b-L51 hasn't been found yet in any Eneolithic/Bronze Age steppe samples, hence the continuing online controversy about its ultimate origins.

Although I suspect that it will pop up somewhere in the same area and time as that Eneolithic R1a-M417.

Andrzejewski said...

I think I made a mistake, I was probably referring to Khvalynsk R1b-L23

Davidski said...

@Samuel Andrews

I'm not sure what you mean when you say that SA6010 doesn't share drift with Yamnaya.

As far as I can see, when looking at various PCA data, including Global25, SA6010 appears to be sitting within a fairly gradual cline of genetic diversity that runs from Piedmont_Eneolithic to the most western-admixed Yamnaya individuals.

So I'm not seeing any controversy here. Simply, the Yamnaya population was a mixture, in varying degrees, between Piedmont_Eneolithic-related populations and other Eneolithic steppe groups with western admixture. And even as late as 2,700 BCE, some Yamnaya individuals, and perhaps groups, were basically still identical to Piedmont_Eneolithic.

Davidski said...

@Andrzejewski

I think I made a mistake, I was probably referring to Khvalynsk R1b-L23.

No such thing.

The R1b from Khvakynsk is probably R-V1636.

Dragos said...

@ Davidski
Sorry OT:can you add Kumtepe 6 to your G25 (not the best coverage, but > 100,000 SNPs).

Davidski said...

@Dragos

I don't know. I'll have a look later today.

Davidski said...

@Matt

It's pretty easy to incorporate Yamnaya into my new qpGraph topology, because the model doesn't require EEF admixture in Yamnaya. This can be wholly substituted with Iron_Gates_HG-related ancestry.

Also, the model clearly works better with Yamnaya_Samara, and screws up a bit with Yamnaya_Caucasus, possibly due to the more heterogeneous genetic character of Yamnaya_Caucasus, and/or its relatively lower quality.

Of course, it'd be nice to include EEF in the model, but that would require quite a bit of work.

Piedmont_Eneolithic qpGraph + Yamnaya Samara

Piedmont_Eneolithic qpGraph + Yamnaya Caucasus

Samuel Andrews said...

@Davidski,
"So I'm not seeing any controversy here. Simply, the Yamnaya population was a mixture, in varying degrees, between Piedmont_Eneolithic-related populations and other Eneolithic steppe groups with western admixture. And even as late as 2,700 BCE, some Yamnaya individuals, and perhaps groups, were basically still identical to Piedmont_Eneolithic."

He carried R1b-V1636 so he probably wasn't Yamnaya but instead from a different Steppe population related to Progress Eneolithic.

Ric Hern said...

@ Matt

"Strange that adding Ust_Ishim into the graph seems to drive it bonkers like that." Is this not because of the significant Neanderthal admixture in Ust Ishim ?

Davidski said...

@Samuel Andrews

What's the bet that eventually whole Yamnaya populations rich in R1b-V1636 will be discovered?

SA6010.A0101 (BZNK-008/3), kurgan 2, grave 13, typical Yamnaya burial position with a flint artefact. Dating: 2884-2679 calBCE (4185±23BP, MAMS29821)

Wang et al. supp info page 49

Halfalp said...

Yes it's interesting how an almost dead modern lineage such as R1b-V1636 could have played an historical role into the Steppe formation. Taking into mind that R1b-V1636 is generally an ethnic Turkish or Armenian argument to explain that R1b came from South Caucasus because it's " basal " or something, even if it's just the brother of P297 in the phylogenetic tree from now. Where could be in the Yamnaya horizon places that are yet not very sampled? South Ukraine ( Crimea )? Repin Culture?

Davidski said...

@Halfalp

Where could be in the Yamnaya horizon places that are yet not very sampled? South Ukraine (Crimea)? Repin Culture?

Not much of a mystery, considering that 1/2 Yamnaya_Caucasus males belongs to R1b-V1636.

There will be plenty more of where that came from.

Bob Floy said...

Where exactly does V1636 sit on the R1b tree? Where is it in relation to V88, for example?

Davidski said...

@Bob Floy

https://yfull.com/tree/R-V1636/

https://yfull.com/tree/R-L389/

Bob Floy said...

So it's just north of P297, thanks.
Weren't the Samara culture samples P297?

Grey said...

"What's the bet that eventually whole Yamnaya populations rich in R1b-V1636 will be discovered?"

if i had to guess i'd guess at least one distinct but related ydna clade per major river with sub-branches along the tributaries.

Davidski said...

@Bob Floy

I'm not aware of any Samara culture samples, but 1/3 of the Khvalynsk_Eneolithic samples belongs to R1b, and this is probably R-V1636.

a said...

Blogger Davidski said...
"@Halfalp

Where could be in the Yamnaya horizon places that are yet not very sampled? South Ukraine (Crimea)? Repin Culture?

Not much of a mystery, considering that 1/2 Yamnaya_Caucasus males belongs to R1b-V1636.

There will be plenty more of where that came from."


Are you sure about this?
It looks like the more successful Yamnaya branch[90%] was not
R-L389+>V1636, but alternative branch R-L389+>R-P297+>M-269+>L23+>Z2103+>Z2106+>Z2108+KMS-67+
https://www.yfull.com/tree/R-Z2108/

Bob Floy said...

Those 7000 ybp samples from the Haak paper in 2015, weren't they Samara culture?

Matt said...

@Davidski, cheers. I think integrating Barcin_N, Caucasus_En and European_MN is kind of the interesting bit though with Iran_N less so, but that's worth showing.

Had a closer look today and lots of weird stuff going on in my adapted version of Wang's graphs adding Ust Ishim. Seems to zero drift drift lengths from root-> Basal Eurasian split->Ust Ishim split, basically rooting the whole tree with Eurasians, and also a trifurcation between BEu, Ust_Ishim and "NW Eurasian" (remaining groups)? Then a trifurcation within "NW Eurasian" of WHG+CHGnonBasal+ANE? Just super strange and not behaving normally at all.

@ric, I don't think it's Neanderthal in Ust_Ishim as we don't usually see that adding UI to trees.

Arza said...

@ Matt Re: Ust_Ishim

It seems that there is a cline like this:

MA1:MA1
AfontovaGora3:I9050.damage 78.5%
Ust_Ishim:Ust_Ishim 21.5%
Distance 3.6826%

You can try this tree as a starting point:
https://i.postimg.cc/0vBQmN5W/ustimix.png

@ Davidski

Can you try this tree? Thank you in advance!

https://drive.google.com/file/d/1S6f7qpjNS-di17Ql6gTCr3MTRcHflUP5/view?usp=sharing

Davidski said...

@Arza

fatalx:
bad addvertex call

Ric Hern said...

@ Matt

"Then a trifurcation within "NW Eurasian" of WHG+CHGnonBasal+ANE?" Could this be that Haplogroup K2b stayed in contact with WHG, CHG-nonBasal longer than Haplogroup K2a which maybe migrated earlier towards the North/NorthEast ?

Arza said...

@ Davidski

if (vertname == NULL)
fatalx ("bad addvertex call\n");


Label at the root node was missing. Sorry! I've updated the file under the same link.

Davidski said...

@Arza

https://drive.google.com/open?id=1bR4R-PAYJJSTUyzeVtAWJQs1JYc49ct0

Davidski said...

@All

Please check out the update. I posted a new version of the tree to test whether Piedmont_Eneolithic can be modeled with admixture from Darkveti-Meshoko.

It's a bit messy, with a few too many mixture edges, but it does the job. Note the zero mixture edge going from the Darkveti-Meshoko-related node to the Piedmont_Eneolithic-related node.

Arza said...

@Davidski
Thanks one more time. Awesome result (and I don't have the Z-score in mind).